[PDF Download] Machine vision algorithms in java techniques and implementation whelan paul f molloy
Machine Vision Algorithms in Java Techniques and Implementation Whelan Paul F Molloy Derek
Visit to download the full and correct content document: https://textbookfull.com/product/machine-vision-algorithms-in-java-techniques-and-im plementation-whelan-paul-f-molloy-derek/
More products digital (pdf, epub, mobi) instant download maybe you interests ...
Exploring BeagleBone Tools and Techniques for Building with Embedded Linux 2nd Edition Derek Molloy
Vision Systems Laboratory, School of Electronic Engineering, Dublin City University, Dublin 9, Republic of lreland
ISBN 978-1-4471-1066-8
DOI 10.1007/978-1-4471-0251-9
ISBN 978-1-4471-0251-9 (eBook)
British Library Cataloguing in Publicat ion Data
Whelan, Paul
Machine vis ion algorithms in java : techniques and implementation
1.Computer vis ion 2. java (Computer program language)
3.Computer algorithms
I.Titiie II. Molloy, Derek
006.3'7
ISBN 978-1-4471-1066-8
Library of Congress Cataloging-in-Publication Data
Whelan, Paul E, 1963Machine vision algorithms in java: techniques and implementation / Paul E Whelan and Derek Molloy. p.cm.
ISBN 978-1-4471-1066-8 (alk. paper)
1. Computer vis ion. 2. Computer algorithms. 3. java (Computer program language) 1. Molloy, Derek, 1973- II. Title
TA1634.w542000
006.3 '7 -dc21 00-030072
Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced, stored or transmitted, in any form or by any means, with the prior permis sion in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms of licences issued by the Copyright Licensing Agency. Enquiries concerning reproduction outside those terms should be sent to the publishers.
Originally published by Springer- Verlag London Berlin Heidelberg in 2001 Softcover reprint of the hardcover 1st edition 2001
The use of registered names, trademarks etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant laws and regulations and therefore free for general use.
The publisher makes no representation, express or implied, with re gard to the accuracy of the information contained in this book and cannot accept any legal responsibility or liability for any errors or omissions that may be made.
Typesetting: PostScript files by authors
34/3830-543210 Printed on acid-free paper SPIN 10743121
To Caroline, Hannah and Sean. (PFW)
To Sally and parents. (DM)
Preface
For many novices to the world of machine vision, the development of automated vision solutions may seem like a relatively easy task as it only requires a computer to understand basic elements such as shape, colour and texture. Of course this is not the case. Extracting useful information from images in a laboratory environment is a difficult process at the best of times, but to develop imaging systems with the reliability and repeatability required for industrial, medical and associated imaging applications increases the complexity of the design task. The aim of this book was to produce a self contained software reference source for machine vision designers which takes such issues into account. To that end Machine Vision Algorithms in Java contains explanations of key machine vision techniques and algorithms, along with the Java source code for a wide range of real-world image processing and analysis functions.
A number of texts have been written over the last few years, which have focused on implementing image processing and to a lesser extent image analysis functions, through coded examples (i.e. for a range of software languages and environments). So, why do we need another book on the subject? Firstly, Machine Vision Algorithms in Java concentrates on examining these issues from a machine vision perspective. It focuses on the image analysis and general machine vision design task, as well as discussing the key concepts relating to image processing. In addition, we have made available (via the Internet) a comprehensive machine vision development environment, Neat Vision, which allows the algorithms and techniques discussed throughout this book to be implemented by the reader.
The range of machine vision techniques and applications has grown significantly over the last decade and as such it would be difficult for a single text to cover them all. Therefore, this book concentrates on those algorithms and techniques that have found acceptance within the machine vision community. As is in the nature of putting a book like this together, certain areas receive greater attention reflecting our experience and the nature of our own research.
This book has grown from a number of different elements. Many of the ideas outlined are based on the postgraduate modules Computer and Machine Vision (EE544) and Object-oriented Programming (EE553) developed by Dr. Paul Whelan and Derek Molloy respectively. Both modules are available in
traditional form and via the Internet as part of Dublin City University's Remote Access to Continuing Engineering Education (RA CeE) initiative l
Another key element was the development of Neat Vision, a Java based visual programming environment for machine vision. It provides an intuitive interface which is achieved using a "drag and drop" block diagram approach, where each image analysis/processing operation is represented by a graphical block with inputs and outputs that can be interconnected, edited and deleted as required. NeatVision was designed to allow the user to focus on the machine vision design task rather than concerns about the subtlety of a given programming language. It allows students of machine vision to implement their ideas in a dynamic and easy to use way, thus reinforcing one of the key elements of the educational experience, interaction. In conjunction with the publication of this book a fully functional 'shareware' version of NeatVision has been made available via the Internet 2
We have also included an introduction to Object-oriented Programming (OOP) and the Java programming language, with particular reference to its imaging capabilities. This was done for those readers who may be unfamiliar with OOP and the Java programming language. It includes details relating to the design of a Java based visual programming environment for machine vision along with an introduction to the Java 2D imaging and the Java Advanced Imaging (JAI) Application Programming Interface (API). A wide range of illustrative examples are also included.
Having successfully digested the ideas outlined in this book the reader will:
• Be familiar with the essential elements of machine VlSlOn software and have an understanding of the problems involved in the development and implementation of machine vision systems from a software perspective.
• Have the ability to design and implement image processing and analysis techniques.
• Be able to evaluate emerging machine vision algorithms and techniques.
• Be familiar with the Java programming language and its application to image analysis.
• Be able to develop machine vision solutions using a Java based visual programming environment (i.e. Neat Vision).
This book is aimed at senior undergraduate and postgraduate students in engineering and computer science as well as practitioners in machine vision, who may wish to update or expand their knowledge in the field. We have tried to present the techniques and algorithms of machine vision in a way
1 http://www.racee.ie/
2 See http://www.NeatVision.com/ for further details on downloading and installing this software. This web site also contains details of the NeatVision environment along with a summary of its techniques. A number of working examples along with sample files and associated visual workspaces are also available.
Preface IX
that it will be understood not only by specialists familiar with the field, but also by those who are less familiar with the topic. Care has also been taken to ensure that we have provided adequate references to supporting work. This should aid readers who wish to examine the topics covered in more detail.
The organisation of the book is as follows. Chap. 1 introduces the general field of machine vision systems engineering. Chap. 2 outlines the key concepts behind the Java programming language. As well as giving a brief history, description and layout of the Java language, this chapter outlines the properties of Java that make it useful for image processing and analysis. The purpose of Chap. 3 is to detail some of the basic techniques and algorithms used in the development of machine vision systems. Key elements of the image processing and analysis functions introduced in this section are also implemented in Java and form the basis of the NeatVision visual programming environment. Chap. 4 follows on from this basic introduction by examining mathematical morphology, a key framework in which many machine vision algorithms can be placed. Chaps. 5 and 6 expand our discussion of imaging techniques to include key elements of texture and colour image analysis. Chap. 7 details the design and implementation of the Neat Vision visual programming environment. Appendix A outlines the graphics file formats that are used by NeatVision and Appendix B details the NeatVision imaging Application Programming Interface (API) specification. Finally, Appendix C summarises the range of operators available in the NeatVision machine vision software development environment. A range of sample applications implemented in NeatVision are highlighted throughout the book.
For updates, corrections, colour images and sample visual programmes please refer to the book web site at http://www.eeng.deu.ie;- j avamv /.
This book has benefited from the comments and suggestions of a wide range of people, including our colleagues with whom we have had many fruitful discussions and collaborations. Numerous students have also contributed, both directly and indirectly. The most important contributions coming from members of the Vision Systems Laboratory (VSL) at Dublin City University (DCU), namely Ovidiu Ghita, Alexandru Drimbarean and Pradeep PP. We would particulary like to express our gratitude to Robert Sadleir (VSL) for a fruitful collaboration in the development of Neat Vision. Robert also contributed to our discussion on the Neat Vision development environment, specifically in Chap. 7. We would also like to thank all the members of the VSL for their comments on the various drafts of this book.
We would like to thank Prof. Charles McCorkell, Head of the School of Electronic Engineering, DCU, for his support of the VSL and this book project. We would also like to acknowledge all our academic colleagues for our many discussions on computer and machine vision, especially Prof. Bruce Batchelor (machine vision systems engineering) and Dr. Pierre Soille (mathematical morphology). We would like to acknowledge Xuemei Zhang (Department of Psychology, Stanford University), Mark Graves (Spectral Fusion Technologies) and Tim Carew (Technology Systems International) and thank them for their permission to use some of the material cited in this book.
Special thanks are due to Nicholas Pinfield and Oliver Jackson of SpringerVerlag for their commitment to this book and maintaining its schedule. Machine Vision Algorithms in Java: Techniques and Implementation was prepared in camera-ready form using the IbTEX text processing environment and Paint Shop Pro image editing software.
Finally, the authors would like to thank their families for their constant support and encouragement during this project. We would like to thank the readers in advance for comments and suggestions aimed at improving and extending the present book and its associated NeatVision software. Any errors, of course, remain our own responsibility.
• Systematic or multiple-copy reproduction or republication; electronic retransmission to another location; print or electronic duplication of any NeatVision material supplied for a fee or for commercial purposes; or altering or recompiling any contents of NeatVision and its associated materials are not permitted.
• By choosing to use Neat Vision and its associated materials, you agree to all the provisions of the copyright law protecting it and to the terms and conditions established by the copyright holder.
• The authors cannot accept responsibility for any loss or damage caused by the use of the source code presented in this book.
Trademarks
• Sun, Sun Microsystems, Solaris, Java and all Java-based trademarks are trademarks or registered trademarks of Sun Microsystems, Inc. in the United States and other countries.
• Netscape and Netscape Navigator are registered trademarks of Netscape Communications Corporation in the United States and other countries.
• Microsoft, Windows, Windows NT and/or other Microsoft products referenced herein are either trademarks or registered trademarks of Microsoft Corporation.
• IBM is a registered trademark of IBM Corporation in the United States and other countries.
• Paint Shop Pro is a registered trademark of Jasc Software, Inc.
2.4
3.4.4
4.3.4
5. Texture Analysis
5.1
5.5
5.6
5.7
5.8
5.9
5.12
5.13
6.3
6.8
6.10
6.11
6.12
6.2.2
1. An Introduction to Machine Vision
The purpose of this chapter is to introduce the reader to the basic principles of machine vision. In this discussion the differences between computer, machine and human vision are highlighted. In doing so, we draw attention to the key elements involved in machine vision systems engineering. While this book concentrates on the software aspects of the design cycle, this task cannot be taken in isolation. Successful application of vision technology to real-world problems requires an appreciation of all the issues involved.
This chapter is aimed at readers with minimal prior experience of machine vision and as such more experienced readers will find most of this material familiar. We also briefly introduce the reader to the concepts behind the Neat Vision visual programming environment. This is discussed in more detail in Chap. 7.
1.1 Human, Computer and Machine Vision
"There is more to machine vision than meets the eye!" (Batchelor & Whelan 1994a)
Computer and machine vision involve the automatic extraction, manipulation, analysis and classification of images or image sequences, usually within special or general-purpose computing systems. The purpose of which is to obtain useful information about the world with a view to carrying out some task.
Many researchers use the terms computer and machine vision interchangeably, although there are key differences. Machine vision systems are generally used in an industrial environment. The design of machine vision systems requires a broad spectrum of techniques and disciplines. These include electronic engineering (hardware and software design), engineering mathematics, physics (optics and lighting), mechanical engineering (since industrial vision systems deal with a mainly mechanical world) as well as the system engineering aspects of developing reliable industrial systems. Machine vision generally involves the engineering of solutions to specific problems and can be seen as a
P. F. Whelan et al., Machine Vision Algorithms in Java
subset of the general systems engineering task. The formal definition 1 of machine vision put forward by the Automated Vision Association (AVA) refers to it as:
"The use of devices for optical, non-contact sensing to automatically receive and interpret an image of a real scene in order to obtain information and/or control machines or processes." (AVA 1985).
Computer vision is a branch of computer science which concentrates on the wider scientific issues relating to the automated interpretation of images. Computer vision often deals with relatively domain-independent considerations, generally aiming to duplicate the effect of human vision by electronically perceiving and understanding an image. Research in this area often deals with issues involving perception psychology and the analysis of human and other biological vision systems.
Of course there is a danger in over-emphasizing the differences between computer and machine vision. Both fields embrace common ground (Fig. 1.1), but each has its own non-overlapping areas of expertise (Braggins 2000). The key difference seems to lie in the motivation behind the development of the imaging solution.
Fig. 1.1. Key aspects of computer and machine vision.
In general, machine vision engineering is about solving specific engineering problems and not trying to develop 'human-like' visual systems. The concept of modelling an industrial vision systems design on the human visual system can often seem like a good starting point for novices to the field. This is not the case (Hochberg 1987) and can often be counterproductive. We should not confuse the two types of vision, the danger in relying on human
1 Note that this definition does not specifically refer to cameras or computer systems
driven approaches is that simpler and perhaps more elegant solutions may be overlooked. The human vision system is qualitative in nature, dealing with a broad range of imaging tasks in varying conditions. The human system can pick out a familiar person in a crowded room, but would find it difficult to give the exact dimensions of that person. Machine vision On the other hand is quantitative, i.e. it can deal with making precise measurements at high speed, but lacks the key elements of the human vision system. Machine vision systems are good at repetitive tasks and provide fast and reliable decision making. If a human operator was to examine a colour web material over a shift, they would find it difficult to detect a subtle colour change due to the human's vision process ability to adapt to such gradual changes. The fact that the human vision process is prone to subjective considerations such as fatigue and boredom, which interfere with consistent evaluations, must also be taken into consideration (these can be reliably handled by a machine vision system).
Machine vision systems for industry first received serious attention in the 1970s (Parks 1978), although the proposal that a video system be used for industrial inspection was first made in the 1930s. Throughout the early 1980s, the subject developed slowly, with a steady contribution being made by the academic research community, but with only limited industrial interest being shown. In the mid-1980s there was serious interest being shown in vision systems by the major automobile manufacturers, although this was followed by a period of disillusionment with a large number of small vision companies failing to survive. Interest grew significantly in the late 1980s and early 1990s, due largely to significant progress being made in making fast image processing hardware available at a competitive price. Throughout this period academic workers have been steadily proving feasibility in a wide range of products, representing all of the major branches of manufacturing industry.
Machine vision systems nOw appear in every major industrial sector, including such areas as electronics (PCB inspection, automatic component recognition), car manufacturing (inspection of car bodies for dents, dimensional checking), food (inspection and grading of fruit and vegetables, inspection of food containers) and the medical industries (tablet quality control, detection of missing items in pharmaceutical packets). The main application areas for industrial vision systems occur in automated inspection and measurement and robotic vision. Automated visual inspection and measurement systems have, in the past, tended to develop faster. In fact, quality control related applications such as inspection, gauging and recognition currently account for well over half of the machine vision market. This has been mainly due to the lower cost and the ease of retrofitting such inspection systems onto existing production lines, compared to the large capital investment involved in developing a completely new robotic work cell and the extra uncertainty and risks involved in integrating two new complex technologies. As manufacturing technology becomes more complex, there is a growing requirement to
1. An Introduction to Machine Vision
integrate the inspection process more closely with the overall manufacturing process (McClannon et al. 1993). This moves the application of automated inspection from a quality control to a process control role, that is from defect detection to defect prevention. The reader is referred to Batchelor & Whelan (1994b), Chin (1988), Chin & Harlow (1982), Davies (1996) and Freeman (1987) for further details on a wide range of industrial implementations of machine vision systems.
As well as reducing scrap and rework costs, product quality can also be improved by using vision to aim for a higher quality standard. Machine vision can be used to determine the cause of "abnormal situations" and provide early warning for potential hazards on a production line, for example detecting a warped metal plate before it is fed into a stamping press and potentially damaging the press. This technology can also provide extensive statistics of production process and in some applications we may have no alternative to using automated vision, as the environment may be hazardous to humans.
Machine vision systems are not perfect, they contain two significant types of error. System errors will be familiar to all engineers as there will always be a trade-off between the cost and functionality of the systems components. System errors can often be reduced by using components with higher specifications. Statistical errors are not as easy to handle. It can be quite difficult to decide exactly where to draw the line which determines a good from a bad product. If the image analysis process places the product in this grey area then we have either false rejects or worse still we may pass a faulty product. This type of error can also be confounded by a lack of communication between the vision engineer and the customer.
Significant progress has been made over the last decade due, in part to the falling cost of computing power. This has led to a spread in the technology and has enabled the development of cheaper machine vision systems. This, in turn, has enabled medium-sized manufacturing companies to consider the option of using machine vision to implement their inspection tasks. To a lesser extent, the availability of a well educated work-force, a small proportion of which has an awareness of machine vision, has also aided the growth and acceptance, of industrial vision systems. The main reason, however, for this growth is strategic. There is a growing realisation within many industries that machine vision is an integral component of a long term automation development process, especially when one considers the importance of quality in manufacturing. This fact, combined with the legal liabilities involved in the production and sale of defective products, highlights the strategic case for the use of machine vision in automated inspection. A similar argument applies to the application of vision to robotics and automated assembly.
1.2 Vision System Hardware
The imaging engine in many low-cost machine vision systems consists of a host computer working in conjunction with single or multiple plug-in boards. The most commOn example of these systems consists of a personal computer, or workstation and a frame-store card, which allow an image to be captured from a standard CCD camera (array image format) and displayed. Many of the current range of frame-store cards also offer on-board processing. Plug-in accelerator cards which enable certain functions to be implemented in realtime are available as daughter boards for many frame-stores. Some framestores have slow-scan capabilities and the ability to interface to line-scan cameras. When used in conjunction with the current range of high speed personal computers, such a vision system is an attractive option for small to medium applications, of low complexity.
Certain personal computers/workstations nOw offer direct video input without the need for additional plug-in frame-store cards. With the growth in multimedia applications, it is likely that this will become more commOnplace On commercial computers. Such systems offer a number of significant advantages, most important of which is their relatively low cost. Additional advantages include their ease of use and familiarity. This is especially the case when used in conjunction with standard personal computers, which have become commonplace both in the home and the workplace. The fact that the host computer for the imaging system is a widely available commercial product also widens the base for software applications.
For greater speed and ability, engineers often turn to plug-in boards which have a specific functionality, such as real-time edge detection, binary correlation and convolution. Typically, the host computer for such boards would be a VME rack fitted with a CPU card. Quite often, such special-purpose boards are pipelined, i.e. they perform different operations On an image in a sequential manner. This allows a new image to be captured while the previous image is still undergoing processing. The main advantage of such systems is their speed and the ability to increase the systems image throughput rate by the addition of extra plug-boards. The disadvantage of such systems is that they can be difficult to program and quite often require programmers with highly specialist skills. There is also a significant cost factor involved in the capital equipment, along with the application development costs. While the majority of dedicated plug-in boards for pipelined systems are tuned to deal with array CCD cameras, newer systems have appeared On the market that are specifically designed for line-scan cameras.
Some system manufactures have taken the option of designing specific machine vision engines which are not tuned for a specific application, but rather designed for their general functionality. Such systems may be totally self-contained and ready to install in an industrial environment. That is, they contain the imaging optics, camera, imaging engine and interfaces for various mechanical actuators and sensors. They differ from turn-key systems in
1. An Introduction to Machine Vision
that the software is supplied with the self-contained system has yet to be moulded into a form that would solve the vision application. Such systems have significant advantages, the main one being speed. The majority of selfcontained systems are custom designed, although they may contain some plug-in boards and are tuned to provide whatever functionality is required by the application. The self-contained nature of the mechanical, image acquisition and display interfaces is also a significant benefit when installing vision systems. However, it can be difficult to add further functionality at a later date without upgrading the system.
Turn-key vision systems are self-contained machine vision systems, designed for a specific industrial use. While some such systems are custom designed, many turn-key systems contain commercially available plug-in cards. Turn-key systems tend to be designed for a specific market niche, such as paper inspection. So, not only is the hardware tuned to deal with high-speed image analysis applications, it is also optimised for a specific imaging task. While the other systems discussed usually require significant development to produce a final solution for an imaging application, turn-key systems are fully developed, although they need to be integrated into the industrial environment. This should not be taken lightly, as it can often be a difficult task. Also, it may not always be possible to find a turn-key system for a specific application.
1.3 Vision System Software
There are a large number of image processing and analysis packages currently available. Several of these packages are freely available over the Internet, see Machine Vision Resources2 for general reference material relating to the design and development of commercial machine vision systems. Some of these packages are tightly tied to a given vision system, while others are compiled for a number of host computers and operating systems. The majority of the software packages have interactive imaging tools that allow ideas to be tested prior to customizing the software for a given application. A number of tutorial books have also been written which implement a large number of image processing functions and to a lesser extent image analysis functions, in a range of software languages (Parker 1997, Klette & Zamperoni 1996, Pitas 1993, Myler & Weeks 1993, Lindley 1991).
For more information on the hardware and software aspects of real-time imaging, including a survey of commonly used languages, see Dougherty & Laplante (1995).
In this section we outline the key components found in the majority of machine vision systems, Fig. 1.2. While the application itself will determine the relevant importance of each of these tasks, successful implementations will have taken all these issues into consideration during the system design. The software aspects of these design stages are expanded upon throughout the course of this book.
Image Interpretation and Mechanical Interface
Image Processing and Analysis
Image Sensor
lighting and Optics
Part Feedi ng and Mechan ical Interface
Fig. 1.2. Machine vision system components.
1.4.1 lInage Acquisition
Image acquisition generally involves the capture of the analogue image signal (although digital cameras can also be used) and the generation of a onedimensional (I-D) or two-dimensional (2-D) array of integers representing pixel brightness levels. In the majority of machine vision applications a solid state sensor based camera is employed. This is made of discrete elements that are scanned to produce an image. Examples of such sensors are Charge Coupled Devices (CCD) and Charge Injection Devices (CID). CCD based cameras are the most commonly used solid state sensor due to their small size, low power consumption requirements, wide spectral response and robustness. These cameras are available in both line scan and area array configurations. Line scan cameras are suitable for high resolution scanning applications where the object moves beneath a fixed camera position, for example a paper web. Area array cameras produce a medium to high resolution 2-D snapshot of a scene (similar to a TV image).
It is worth noting that there have been major advances in other component technologies, specialised lighting units, lenses and advisor programs, which guide a vision engineer through the initial stages of the design process.
Oala Flow
FeedbaCk P alh
8 1. An Introduction to Machine Vision
While the acquisition of image data may not be seen as directly related to image analysis and processing, the design decisions involved in this stage (i.e. camera type, lighting and optical system design) have a fundamental bearing on the systems software. For example, in Fig. 1.3 we have applied two different lighting techniques to the same scene. Given the task of counting the number of candies in the image, which lighting configuration should be adopted? While the image in (b) gives a cleaner image, we would have to apply complex image segmentation techniques (such as the one discussed in Sec. 4.4.2) to separate the touching items prior to counting. Whereas the image in (a) may initially seem to be of little use due to the bright reflections evident on each candy piece. In fact the lighting unit was specifically designed to produce this effect. This lighting configuration simplifies the image processing and analysis tasks as all that is required of the software is to isolate and count these reflections (Chap. 3), both relatively straightforward imaging tasks.
Fig. 1.3. Different lighting techniques applied to candy sweets. (a) Ring Light (Light Background). (b) Back-lighting, this produces a silhouette of the scene.
1.4.2 Image Representation
This relates to the means of representing the image brightness data in a form that enables the image processing and analysis design process to be implemented efficiently. We shall first consider the representation of monochrome (grey scale) images. Let x and y denote two integers where 1 :::; x :::; m and 1 :::; y :::; n. In addition, let f (x, y) denote an integer function such that 0:::; f(x, y) :::; W (W denotes the white level in a grey scale image). An array F will be called a digital image, where an address (x, y) defines a position in F, called a pixel, or picture element. The elements of F denote the intensities within a number of small rectangular regions within a real (i.e. optical) image. Strictly speaking, f(x, y) measures the intensity at a single point, but if
(a)
(b)
the corresponding rectangular region is small enough, the approximation will be accurate enough for most purposes. The array F contains a total of m x n elements and this product is called the spatial resolution of F (Fig. 1.4).
For example, when m = n 2: 128 and W 2: 64, we can obtain a good image of a human face. Many of the industrial image processing systems in use nowadays manipulate images in which m = n = 512 and W = 255 (Fig. 1.5). This leads to a storage requirement of 256 Kbytes/image. A binary image is one in which only two intensity levels, black (0) and white (1), are permitted (Fig. 1.5(d)). This requires the storage of m x n bits/image.
Fig. 1.4. A digital image consisting of an array of m x n pixels. The pixel in the xth column and the yth row has an intensity equal to f(x, y).
An impression of colour can be conveyed to the eye by combining different colours of light. The signal from a colour television camera may be represented using three components (red, green and blue): R = r(x, y); G = g(x, y); B = b(x, y), where R, G and B are defined in a similar way to F. The vector (r(x,y),g(x,y),b(x,y)) defines the intensity and colour at the point (x,y) in the colour image. Colour image analysis is discussed in more detail in Chap. 6. Multispectral images can also be represented using several monochrome images. Video sequences are used to represent and analyse movement within or across scenes. A video sequence is, in effect, a time-sampled representation of the original moving scene (see Blake & Isard (1998) for details of one approach to image sequence analysis). Each frame in the sequence is a standard colour, or monochrome image and can be coded as such. Thus, a monochrome video sequence may be represented digitally as a sequence of 2-D arrays [F1 ,F2 ,F3 FN].
While it is beyond the scope of this book to detail the various image coding techniques that are now available (see Rao & Hwang (1996) for further details on video coding), two basic coding mechanism have been employed in machine vision, namely run-length and chain coding (these will be discussed in Chap. 3).
Fig. 1.5. Image resolution and binary threshold. (a) to (c) Bottle-top grey scale image at varying resolutions. (d) Binary version of the bottle-top image.
1.4.3 Image Processing
This is an image to image operation. It aims to produce images that will make the analysis stage simpler and robust. Key to the image processing task is the ability to segment (Fig. 1.6) and enhance the features to be analysed. This can be a difficult task when dealing with complex scenes. The two main classes of operations applied to images during processing are point to point and neighbourhood operations. A wide range of image processing functions are detailed in Chaps. 3, 4, 5 and 6.
1.4.4 Image Analysis
Image analysis involves the automatic extraction of useful information from an image. This information must be both explicit and useful in the subsequent decision making process. Common image analysis techniques include
Fig. 1.6. Image processing applied to a paper watermark image. (a) Original image. (b) Extraction of the watermark from (a).
template matching, pattern recognition using feature extraction and descriptive syntactic processes.
Template matching involves comparing an ideal representation of a pattern or object, to segmented regions within the image. This can be carried out on a global basis, or locally whereby we can use several local features, such as corners (Fig. 1.7). The key problem with this approach is that it cannot easily deal with scale and/or orientation changes, therefore it can be computationally expensive. This approach to image analysis was popular in the early days of machine vision, but had gone out of favour due to the problems outlined. Although the introduction of faster computing power and custom VLSI devices has caused many researchers to reevaluate the possibility of using template matching for industrial imaging tasks.
Fig. 1. 7. Local template matching. In the image of the rectangle each of the local templates are associated with corners within the scene. When the image suffers a transformation the allocation of the local templates to the scene becomes a more difficult task.
In feature extraction based pattern recognition, images are described in terms of their representative features. This has been a dominant form of image analysis in machine vision over the last ten years. This involves segmenting the image into its constituent parts. Once this has been completed the key features (e.g. the corners within an image) are extracted and then classified. See Chaps. 3, 4, 5 and 6 for details on a wide range of image analysis functions.
12 1. An Introduction to Machine Vision
Descriptive syntactic processes involve a linguistic approach to image analysis. They model an object by a set of basic elements and their (spatial) relationship. A key problem with this approach is that it lacks precision and as such has found little favour with machine vision designers.
Fig. 1.8 summarises some of the key stages involved in a machine vision system. Once an image is captured, processing is applied to highlight the features of interest and reduce the amount of data that has to be processed. The key features, i.e. the star tips in this example, are then extracted. Finally, a scalar value is generated to indicate the number of star points. This will be compared to a predefined value to determine if the part has been manufactured correctly.
Fig. 1.8. A typical sequence of operations found in machine vision. The aim of the algorithm is to count the number of points on the star so that it can be compared to a predefined value and hence a decision on its suitability can be made. (a) Original binary star image. (b) Image processing: The image is thinned (see Sec. 3.3.2) a predefined number of times to reduce the points on the stars to a single pixel width. (c) Image analysis: Limb end analysis (see Sec. 3.3.2) is then applied to isolate each point (these are shown as small white squares). The points are then counted (see Sec. 3.3) prior to classification.
1.4.5 Image Classification
Objects are classified based on the set of features extracted. Key to the image classification task is the fact that similarity between objects implies that they contain similar features, which in turn form clusters. A number of image classification techniques including the Nearest Neighbour, K-Nearest Neighbour, Maximum-likelihood and Neural Networks classifiers are discussed. We also examine the Confusion Matrix and Leave One-out methods of evaluating the performance of these classifiers.
But, how do we know we have identified suitable features? Ideally extracted features should have four characteristics (Castleman 1979).
(a)
(b)
(c)
Discrimination: Each feature should take on significantly different values for objects belonging to different classes. For example, object size in sorting A4 and A5 sheets of paper.
Reliability: Features should take on similar values for all objects of the same class. For example, an apple may appear green or red, depending on ripeness, therefore features may look different even though they appear in the same class.
Independence: Various features used should be uncorrelated with each other. For example, the diameter and weight of fruit could be highly correlated, both reflecting the same property, i.e. size. Highly correlated features may be combined (i.e. averaging to reduce sensitivity to noise) but they should not be used as separate features.
Small feature set: Complexity of pattern recognition system increases rapidly with the dimensionality (i.e number offeatures used) ofthe system. Also, the number of objects required to train and measure the performance of the classifier increases exponentially with the number of features present. Adding more features which are noisy or highly correlated can degrade system performance.
There are two main approaches to the problem of classification: supervised (i.e. when all the classes are known and can be used to train the classifier) and unsupervised classification (i.e. when the number of classes are unknown).
Supervised Classification. Supervised classification requires a priori information about the input data patterns. Usually this information is based on an underlying statistical distribution of the patterns, or on a set of patterns with the correct classification, the latter being produced by a human expert who supervises the choice of learning cases (Weiss & Kulikowski 1990). There are two phases in the supervised classification process: learning (or training) and testing. During the training stage patterns are sequentially presented to the classifier, which modifies its internal properties accordingly. Extracting the training set of patterns to achieve optimal training is generally a difficult task and it is very data dependent. The training set should be fully representative of the populations from which it is drawn (Davies 1996). Once the training step has been completed, the system can be used in the testing phase where the classifier is presented with new test patterns. This set of patterns is called a validation set. Examples of this form of classification include Nearest neighbour, Maximum-likelihood and the application of Artificial Neural Networks.
Nearest Neighbour Classifier (NNC): Classifies an object based on the closest cluster that the unknown object lies near. This is also referred to as the Maximum Similarity Classifier (Vernon 1991). This approach is simple to implement and hence popular, although incorrect classification is possible. It is also sensitive to noise. Assuming an object Q described by
the vector (Xl, X 2, ... , X n), then a set of m reference vectors, descri bing archetypal objects of each class, denoted by Yl , Y 2 , Y m , where Yi = (Yi,l, Yi.2, ... , Yi.n). Similarity between two objects represented X and Yi can be assessed by measuring the Euclidean distance (D (X, Yi)) between them.
The larger D(X, Yi) is, the smaller the similarity is between the patterns they represent. Thus, an object of unknown type and which is represented by a vector X can be attributed to an appropriate class by finding which of the Yl , Y 2 •.• , Y m values are closest to X.
The K-Nearest Neighbour (KNN) is an extension to NNC which tries to obtain the k nearest neighbours in the training set and then labels them using majority voting, i.e. assigns a class C to a pixel X if the majority of the known neighbours of X are of class C. This approach is more robust to noise than the NNC. k is usually chosen to be quite small (k = 3,5 are commonly used) and tends to be dependent on the amount of data that is available.
Maximum-likelihood Classifier: Also referred to as Bayesian classification (Vernon 1991). Assume two classes of objects, C l and C 2 , with a single feature to distinguish these classes (e.g., the number of holes, x). This technique involves finding the Probability Density Function (PDF) for each class and measuring the probability that an object from a given class will have a given feature value. This is found by measuring x for a large number of samples of each class. Let P[Cn ] be the probability of class n occurring (in the range [0, 1]). But we require the probability that an object belongs to class n given a measured value of x, i.e. P[Cnlx]. An object belongs to class 2 if P[C2 Ix] > P[Cllx], where P[xIC2 ] is the probability that feature x will occur given the object belongs to class 2. Using Bayes Theorem and assuming that P[x] (normalisation factor) is the same for both classes and P[C l ] = P[C2 ], P[xIC2 ] > P[xIC l ]. If this occurs then the object belongs to class 2, based on the measurement of the feature x.
Artificial Neural Networks (ANN): Artificial Neural Networks are relatively straightforward to design and are capable of dealing with pattern complexities not discernible by other classifiers. During the training phase we present the input data to the network, calculating its outputs. The outputs are then compared with the expected results. The weights of the network are then adjusted. This should force the network to converge and thus learn how to discriminate varying inputs as they are presented. Each cycle of presenting the input, calculating the output, comparing results and changing weights is referred to as an epoch. Unfortunately the training phase can take a lot of time and we have no guarantee of
Another random document with no related content on Scribd:
can wear red, it is a colour so full of life, but it does not suit me at all.”
For a second she had looked puzzled, but she is too thorough a woman of the world to hang out her emotions, and she replied with her usual suavity:
“If I could wear pink and dead white, and brilliant blues as you can I should resign red without a struggle. I have never seen anything so beautiful as you are with a pink rose in your hair and another at your throat.”
But I had had as much of this as I could stand and I asked her abruptly if she had visited many of the lakes.
“This is my first visit to the Adirondacks, I am ashamed to say. I was taken abroad every summer when I was a girl and I have gone almost every year since from force of habit and because I really have such a delicious time. It is one’s duty to see all the beauties of the old world, don’t you think so, Lady Helen?”
But I did not want to talk about Europe. “I hear that Spruce Lake is so beautiful,” I said. “Could not we—a party of us—walk over there some day? It is only seven miles, and Mr. Nugent says the trail is very good.”
“Seven miles! Dear Lady Helen. Twice seven are fourteen. Remember—we—alas!—are not English.”
“They might put us up for the night——”
“Oh, quite impossible. We do not know any of them. They are business people from Buffalo and Utica and all those provincial towns.”
“In trade, do you mean?”
“That is the way you would express it. It is the same class—people who keep stores or make things.”
“And they have the same tastes as yourself?” I asked, puzzled at this new American facer. “They are—sportsmen? They lead the same life up here as you do?”
“I really don’t know anything about them. I suppose there are certain national characteristics; several lakes in the Adirondacks are owned by people of that sort. I am told that there was an encampment of commercial travellers just off the borders of this property last year.”
“But I don’t understand. Your lines of caste are very marked, it has seemed to me. Why should the leisure class and the commercial traveller have the same tastes. It is very odd.”
But she refused to take the slightest interest in the subject, and that afternoon as I was walking to the lake of the water-lilies with Mr. N. I asked him for enlightenment.
“Oh, Eastern men are keen sportsmen,” he said. “That is to say, most—wherever there are mountains and woods and lakes. It is an instinct inherited from the old hunters and trappers—from the days when the settlers shot game for food and were as familiar with the wilderness as the farm. These settlers were the ancestors of men who are in all classes of life to-day. And you must remember that there is no ‘Continent’ to run over to for the yearly vacation. You can travel an immense distance here and pay a good deal of money only to hear a change of accent. But the forests of New York and Maine mean rest, reinvigoration, and the complete happiness of the sportsman. These men up here go in the woods every year as naturally as they keep their nose to the grindstone for the remaining ten or eleven months.
“And they are first-class sportsmen.”
“As good as any in England.”
“Men that—that—sell hats?”
“Carlisle is neither keener nor better.”
“Certainly your country is wonderfully interesting and sometimes I feel as if I were groping about in the neighbourhood of the true democracy. Do they also play golf?”
“They do, indeed.”
“The tradesmen? People who keep retail shops.”
“In the small interior towns many of them have achieved sufficient prosperity and leisure, and they are very keen about it. But in the large towns it is usually the wealthier class that goes in for it; the families of business and professional men, successful on a large scale.”
And then I saw the lilies.
I must tell you that Mrs. Van Worden often goes into the kitchen and sits in a rocking chair by the window and talks to Mrs. Opp, and that sometimes, when the men are out, she invites her into the livingroom. It appears that unless these people were treated with a certain amount of consideration they would not remain. A city servant is a servant, but in the country they appear to have studied the Declaration of Independence, and doubtless they all know that Abraham Lincoln’s sister “lived out.” Mrs. Opp is quite insensible of her noble blood but she is as proud as Lucifer all the same, and because her untainted Americanism teaches her that she is “as good as anybody.” She is willing to work for hire and envies no one, but the slightest display of “airs,” an unthinking snub, and she would pack her bundle and march over the mountain with a majesty the self-conscious American of high degree never will achieve. It is truly delightful and I love her. I often go out and sit in the rocker and watch her great bulk move lightly about the exquisite kitchen, and listen to her kindly drawl emphasised by little gracious bends of the head. She tells me the gossip of the mountains, and alluded the other day to the cook at Boulder Lake as “a lovely woman.” I told her about Jemima, and she said:
“Poor child. I guess she was right, but she didn’t know how to take it. Of course you folks nat’rally wants ter eat with yourselves, and the hired help as is used to farms and little country towns don’t just see how it is at first. Different people has different ways and all we ask up here is to be treated right, we don’t expect the hull earth. I’ve always knowed that, because I’ve lived so much to hotels, but Jemima, I guess she’s pretty green yet.”
When I told her about Jemima wanting to see “a dook,” she laughed heartily.
“Well, I guess I’d like to see one, myself,” she admitted, “not that I’d expect them to be so different from other folks, but just because I’ve read so much about ’em. That’s it. That’s it. I’m glad he’s gittin’ on so nice. He had orter drink plenty of milk.”
Curiously enough, that evening I received a letter from Bertie saying that another “eminent doctor” had put him on a milk diet and promised him complete health in one year if he would be faithful to it and the Adirondacks for that period.
“Its beastly uninteresting diet, Nell, and required all the will I’ve got to make up my mind to it,” he wrote; “but I want to get back to England and be alive once more, so I’ve plunged in—literally enough. I’ve leased an Inn on one of the big public lakes from October till June, so we’ll have a big old-fashioned house, they tell me, and not a care, for the proprietor will ‘run it.’ Rogers has promised to come up twice a month and I have written to Nugent and asked him to come often and bring all the friends he likes. I fancy from your letters that I should like the men—and women—over there better than these— Rogers excepted. I believe he is in love with you, Nell, and so is Nugent; but you mustn’t marry an American. By the way Roddy Spencer is coming over here—wrote me to expect him any day, and that he’d look me up at once. He has just succeeded—old Landsburghe died last month, and left Roddy all his personal property. That must amount to three or four hundred thousand and with the estates will set up Roddy as well as he could wish—and his debts must have been a pretty penny.”
I shall be rather glad to see Roddy. He was always with Bertie when they were boys, but I have not seen much of him of late years. Didn’t he go to South Africa in the hope that Rhodes would put him in the way of making a fortune—after he had loaded himself too heavily with debts to remain in London. I forget the details. The legacy must have been a pleasant surprise, for the old Marquis was very eccentric and had refused to pay his debts. Well, I shall be glad to see him and suppose he is as good looking as ever.
H .
Letter IX
From the Lady H P to the Countess of E and R .
Chipmunk Lake, August 11th
Dearest Polly:
I AM rather put out, and have been so irritable for two days that I hardly know myself. Still, thank heaven, nobody suspects it. I never have been more amiable.
The other night a half dozen of the party were playing Bridge in a corner of Mrs. Van Worden’s living-room. I detest gambling and was trying to interest myself in a book when I happened to glance out of the window and saw—Mrs. Coward and Mr. N. on their way down to the lake. Now, I don’t pretend to be in love with the man, Polly, but I do feel that while he is pretending devotion to me it is little short of an insult for him to sneak off with another woman—and an arrant coquette—for a row at nine o’clock at night—it is scandalous and I never have heard any one utter so many virtuous platitudes as Mrs. C. If I thought he was trying to make me jealous I should merely dismiss him from my mind with the contempt he would deserve, but he really is incapable of such pettiness, and I happen to know he was only too frightened I’d find it out.
Polly, I cannot pretend to describe to you my sensations when I saw those dark shapes steal through the spruce grove before the house —the branches are cut so high that it is really a grove of slender trunks and you see the lake plainly. For the moment I felt as if my heart were sinking and I involuntarily pushed my hand underneath it, while my breath shortened and my face burned and then went cold. I had an impulse to rush out and see if it really were true and to prevent it. And then I fell into a rage. How I wished Roddy Spencer were here. He is such a splendid looking creature that he could be made to set another man wild with jealousy. Suddenly I bethought myself of Carlisle. He was playing, but the game was nearly over. I made up my mind in an instant; I got up and moved about the room as if I were getting bored and impatient, and in a few moments I caught his eye. I sent him a glance of coquettish appeal, and it had
the desired effect. The moment the game was over he was at my side and we ensconced ourselves on the three-cornered divan under the swinging rose-coloured lamp and never moved till twelve o’clock. N. and Mrs. C. returned, looking half-frozen and too silly, for they were obliged to get almost inside the chimney. We never noticed them. I coquetted, Polly, as I never coquetted before, and Mr. Carlisle is a flirt whose accomplished depths it is interesting to explore. For fear he should think I was animated by pique—although he knew nothing of the row—I contrived to intimate that I was rather bored and on the verge of making an excuse to return to Boulder Lake. At the same time I made him feel what a triumph he would achieve if he renewed the fascinations of Chipmunk Lake for me. Nothing would induce me to leave. I shall stay and prove to this selfsatisfied American flirt that I can make myself twice as interesting as herself. I’ll employ her own weapon, flattery, and make her platitudes apparent. When I have sufficiently punished N. I’ll take him back and keep Carlisle besides. I am sure she wants to marry N. She has too large a fortune of her own to be tempted by Carlisle’s, and N.’s possibilities appeal to her inordinate ambition and vanity.
This morning, of course, N. tried to be as devoted as usual. But I dismissed him with an absent smile, which became brilliantly personal the moment C. appeared. We went off for a walk in the forest and I never shall forget the expression of N.’s face. I almost relented. But he deserves punishment. I will have all or nothing. In the afternoon Mrs. Coward drifted about majestically for a half hour or so—her face expressing nothing—while Carlisle and I read a novel together on the divan in the corner. She tried to get N. into her pocket but he merely glowered into the fire and took no notice of her. Presently she drifted away, and in the afternoon I saw her fishing with Mr. Van Worden! If I were in such desperate straits I would give out that I was writing a book, and keep to my room. I fancy she wove a net of flattery for Mr. Latimer, but he is a faithful soul. Mrs. Van W. often looks sad, by the way, brilliant as her normal spirits are. It must be an unsatisfactory roundabout way of trying to be happy. I am more than ever determined to make no mistake when I do marry, and to consider one thing only. I am convinced there is no other happiness.
13th
I have restored N. to favour but now give him only half my time, that he never may be quite sure of me again. Mr C. apparently is quite as high in my good graces, and while he is merely stimulated and on the verge of becoming serious, N. shows a curious mixture of alarm, anger and energetic determination—and has taken no more rows or walks with Mrs. C. I have managed to convey to him that I will accept no divided homage, and he is now only too eager to give me the whole of it, and keeps out of Mrs. C.’s way. I must say she is a thoroughbred. She has never betrayed jealousy or pique by the flutter of an eyelash. Perhaps I’ll restore Mr. Carlisle to her presently, for I am rather tired of him. There is none of the quality of the unexpected about him. He is a well-proportioned mass of good points and good fortune—all trained outward; he never has had the necessity—I believe his family has had wealth and position for four generations—nor the inclination to look very far into himself, consequently the crust has deepened and the personality diminished. Mr. N., on the other hand, while as well-born, has had all his faculties sharpened by a struggle with the impinging forces that array themselves against the young man seeking to conquer them without money. But first of all he received a college education and distinguished himself by his talents and hard study. Given the illuminating education first and the struggle of the wits for mastery afterward, and adding to both the advantages and the principles of a gentleman—and the inner life, the soul, of such a clever man could not fail to be developed, complex and interesting. I have had only glimpses of it, but it has excited my curiosity so that I naturally could not watch another woman carry him off with equanimity. But I don’t think there is any danger of another lapse. He has renewed his efforts to interest me, and—it certainly is clever of him—has not so much as addressed me with his eyes again. That is to say there is no appeal in them. But there are other things, Polly, and there is something about the man that fearfully suggests the impossibility of failure. But the weather is so heavenly just now that I wish everybody in the world could have everything he wanted. It is like living in a crystal dome and being bathed by invisible waves of soft stimulating perfumed air; with splendid masses of rich and tender greens, of
amber-browns and turquoise blue, and the golden glory of sunsets for the eye, and a vast uplifting silence. It is a sort of voluptuous heaven, virtuously seductive.
The other day several of us walked down the mountain to one of the farms to see the haying. It was a grand valley among the mountaintops. From the farm we visited we looked over rolling wooded hills, dotted with houses, cattle, and a solitary white church spire, to a great irregular chain of green mountains, encircling the horizon; other peaks, faint and blue and distant, showing beyond their depressions; and the forest, the forest, everywhere beyond the clearings of the farmers.
The hay had been cut and the mechanical rake was gathering it into heaps when we arrived, while men pitched it into a wagon where another man stood with a pitchfork pressing it down. The sweetness of that air! I never shall forget it; I was doubly glad I never had used perfumes. It was drenched with the sweetness of newly mown hay and it almost intoxicated me. I fancy that if Mr. N. had seized the occasion to press his suit—however, I do not know. He did not, and as there were some ten people in the field it would not have been so romantic, in spite of the fragrance. They gave me a hand-rake and I raked quite a good deal of the pretty green stuff that it seems shocking the farm-yard cattle should eat.
I was very much disappointed in the appearance of these mountain farmers. How few things in life resemble the traditions of them—we have been so victimized by poets and romanticists. I expected great brawny muscular fellows, with enormous legs, brown skins, and deep chests. But they are pale and thin and stooping, not one looks as if he would see sixty or as if he got the least pleasure out of life. Mr. N. explained that hard work while they were young and no cessation of it thereafter had broken their constitutions. In winter they work on the roads which they are under contract to keep open, and of course it snows and freezes heavily and frequently. During these same severe months they also go in the woods and help to “draw” the logs the lumbermen have cut during the summer for the pulp mills. These logs have to be piled on sleds and drawn to the creek, to be floated down to the valley after the spring rains. The men rise
at three in the morning, working by torches till sunrise, and seldom get to bed before eleven at night! No wonder they are old men at forty. And I don’t pretend to say how many cords of wood they cut a year, or how many stone fences they have built, or how many thousand stones they dug from the ground before they could farm. When all the hay had been removed the field looked like a great green lawn—brilliantly green under the five o’clock sun. Beyond was a dip, then the thick masses of the dark green woods, touched into richer green by that blaze of sunshine, then the mountains, sombre and faintly blue. It is a beautiful land, Polly, but it depresses me to think that while it means new blood and new life and all cure for the outsider of leisure, it sucks jealously back into its own store of vitality the little store of its struggling children. It is an unnatural and snobbish mother, after all.
Mr. N. calls me “Maud Muller,” since I raked the hay. Did you ever come across that Quaker poet, Whittier? He lived here in America, but I am told he is the poet of the English quakers as well. Mr. N. recited “Maud Muller” to me, as we stood apart in the field—after I had tired of raking. It really is a beautiful and musical poem, but I could not see anything quakerish about it.
When I write “I am told,” Polly, you may assume that my authority is Mr. N. As far as my limited comprehension can perceive he knows everything.
16th
Mr Carlisle was called suddenly to Newport last night by the illness of his mother. A man rode thirty miles with the telegram, left his horse at a farm house and walked the trail—which is so full of rocks logs and mud-holes that no strange horse could cover it without breaking a leg, and, likely as not, his neck. It was just after dinner and we were all grouped in the little spruce grove before the camps watching the sunset, when the man, looking so hot and tired, came hurrying out of the woods. I am sure every one of us had a fright when we saw the yellow envelope, a telegram is such a rare interloper in the peace of these mountain camps; and when the man said “Mr. Carlisle,” I am equally sure that every one of us wanted to
hold Mr C.’s hand. He went rather pale, but said it was doubtless a false alarm as his mother had been very nervous ever since the war —during which her nerves had been on the rack between apprehension of Cervera’s fleet and his own demise—and hastened away to get his things together. The keeper was sent out to bring a buckboard in and Mr. C. left at three this morning.
I am sorry to say, Polly, that before he left I had rather a painful interview with him. I tried to avoid it, but these American men are very determined, my dear, and he managed to detain me on the veranda after the others had gone in. How I hate men to be serious! It hurts my conscience so that I don’t get over it for weeks. I had not the slightest intention of making him love me, I only wanted to punish that woman for her contemptible conduct in regard to Mr. N. Now I feel quite as bad myself and wish that I had simply contented myself with showing her that I had a string tied to Mr. N. Mr. C. is really a fine manly fellow and I felt like petting him as I would Bertie and telling him not to mind, but of course I couldn’t. He vows he’ll come back the moment he is free, and gently insinuated contempt for a suitor who was ten years too old to win me. How funny these men are! What I am to do with them all I am sure I cannot imagine, but I would rather he went away hoping, and accepted his fate by degrees; for heaven knows I do not wish to add too heavily to his troubles. Not that I gave him any encouragement. Heaven forbid. But they are so determined, these Americans.
The buckboard awoke me at three o’clock, and I got up and peered out of my high window. The woods looked so grey and ghostly, filled with mist that was like a wet cobweb. The keeper was driving and Mr C. sat on the back seat muffled in a winter great coat. Mr Latimer went out with him to the end of the trail; and presently they disappeared into the forest and the mist; and the silence was as if the world were dead.
But I have not told you of the new arrival. She has come to spend the last of our camping days here with Mrs. Wilbur Garrison, and she is quite the most imposing, nay, overwhelming person I have met in this extraordinary jumble of democracy and caste, known— infelicitously, I gather—as the United States. She is Mrs. Earle wife
of —— —— —— ——,[A] and of course, a personage of vast importance in Washington. But she is no mushroom; she has belonged for heaven knows how many generations—eight, perhaps —to the haute noblesse of the country and was born into an equally imposing number of dollars. But, Oh, Polly! she is so cold, so haughty, so frozen! Her handsome little head is set so far back on her mountainous body, her backbone is so rigid and her upper lip so proudly curled, there is such a touch of icy peremptoriness in her manner, as if it were her daily task to dismiss pushing aspirants for social recognition—that I feel I have looked upon the walking embodiment of the aristocratic idea as it is interpreted by Americans. Heaven knows she has been sweet to me, she has even invited me to spend a month with her in Washington next winter; but I know that after a consecutive month of that chill presence I should return to Bertie a sort of hysterical iceberg, my marrow frozen and my humour on the verge of insanity. And, although not in the least clever, she would be quite an agreeable woman were it not for her tragic selfconsciousness, for she must know the world of Washington like a book; and it is a book I should like to read. But she will not talk. All she has said of Washington is to intimate her scorn of all “newcomers,” her boundless ennui of the duties of her official position, her sacrifice of her own inherited desire for segregation from the common herd to the interests of her distinguished husband. And yet she is not ill-natured. She is as placid as Chipmunk Lake, and, I am told, an exemplary wife and mother, if not a radiant and fascinating hostess. Her only fault is—well—her aristocracy! I tried to interest her in the vivid people of Boulder Lake, in the farmers of the valley, in Jemima, in the various strange beings I have met in this strange country. All by way of experiment, and in vain. Her mind could not respond to the fact of their existence. They had not been born in her original circle nor thrust upon her by the exigencies of public life. She betrayed a flicker of interest in Mrs. Opp and remarked vaguely that she should have imagined blood would count for more than that, then curled her lip and relapsed into silence. Polly, what are we? I am oppressed sometimes with the suspicion that those countries which are not the United States are like diseases in the creed of the Christian Scientists—they do not exist. We merely imagine we are,
they know we are not, but tolerate our whim. We have lost caste because we have lost our consciousness of birth, and are therefore degenerate. Upon my word, Polly, I begin to think that the snobs who run after us in England are the truest Republicans, after all. However, I have nothing to say against my other friends of Chipmunk Lake—always excepting Mrs. Coward—for they are wholly charming, and unaffected, and not afraid to let you see they know the world. Miss Page does not, and thinks well of all the world—happy, happy girl.
There is another point on which these people greatly differ from those of Boulder Lake—that is a certain homogeneity. Men and women, they are like one large family and evidently have been brought up together; and Mrs. Van Worden says that with most of her set it is quite the same. They all call each other by their first names, and there is that same utter absence of formality as with us when we are quite among ourselves. In the set at Boulder Lake there is a formality that never relaxes, they all seem to have an abnormal respect for each other, and I vow I never heard the first name of one of them. They have accumulated each and all with care, but their set is stamped with the heterogeneity of a new incident in civilisation. And they have a bourgeois timidity about expressing their real opinion—if they have any—against the ruling opinion—or fad. Each thinks as the other thinks—how often I have disconcerted them!
Mrs. Coward, by the way, preserves her unruffled demeanour and has never so much as put out a little claw and scratched me. A woman of twenty-seven with that amount of self-control should be capable of great things. She never has overdone it, never for a moment. (I wonder if she has anything up her sleeve.)
I did not tell you, she informed me the other day that she is a “Colonial Dame,” and has her family tree—with two Presidents and ten statesmen on collateral twigs—framed and hung in her library at Newport. Polly, what do you make of that. I ventured to speak to Mrs. Van Worden about it, and she said:
“Rot. Fads. We all think that the Almighty made Heaven first and the United States just after—pickling it till Europe and Asia were old
enough to appreciate—but some of us have the decency to do less talking than thinking. Nettie Coward is fairly mum as a rule, but she can’t help showing off to you—wants to impress you with the fact that you’ve not got a monopoly on all the blood there is. She’s a clever woman, but everybody makes an ass of himself one way or another. When we’ve got twenty generations to the good we’ll be just as unconscious about it as you are. But aristocracy will be a sort of itch with us till then. Quantities of idiots have their family trees framed.”
I find her very refreshing, Polly.
I have not written much about Mr. N. lately. But I’ve talked with him! —hours and hours and hours. It is no use trying to avoid him—and he certainly is interesting. Well—heaven knows.
H .
9. . .—Your letter has just come. It seems years since I last heard from you. I know how you feel but I can’t help being glad that he has gone. Nothing will happen to him. Don’t be foolish. He is your manifest destiny and you will be married to him this time next year. If Freddy really has hurt himself and is suffering I can’t help feeling sorry for the wicked little beast—I have grown so soft about such things in the last two years. But the circumstances were disgraceful and you were wise to treat his summons to his bedside as a trick to compromise you and hamper the proceedings. What an enigma even a miserable little degenerate can be. Who can say whether he really is fascinated by you still—he is incapable of love—and honestly desires a reconciliation, or whether he wants to prevent your marriage with V. R.—or, who knows?—perhaps he is afraid that woman will want to marry him. Well, I do wish that the evidence could have been gathered more quickly and that it were over.
From the Lady H P to the Countess of E and R .
Chipmunk Lake, August 19th
Dearest Polly:
IT is eleven o’clock . ., and I have been in bed and asleep since half-after seven. I foresee myself wide-awake for two hours and giving you an account of the last two days. How flat that sounds—but wait! And otherwise you might never hear of them, for I return to Boulder Lake to-morrow, and in this country events are so quickly crowded into the past.
I wrote you—did I not?—that the subject of a camping expedition had been mooted more than once, but put off from time to time on account of threatening weather and various other causes. I longed to go; “camping out” in the “Adirondack wilderness” being pitched upon a most adventurous and romantic note; and finally I begged Mr. Nugent to arrange it. He went “straight at it” in the energetic American way, and in two hours it was all arranged: Opp drove out in the buckboard for another guide, and Mrs. Opp was making so many good things at once that all the other cooks had to come over to help her. Then Mr. Nugent and Mr. Van Worden packed the big packbaskets, and everybody was ready to start at nine o’clock the day before yesterday.
The original plan was that all of us should go, but the actual party were Mr. and Mrs. Meredith Jones, Miss Page, Myself, Mr. Nugent, Mr. Latimer, and Mr. Van Worden. The others “backed out” on one excuse or another, and happy it was for them and us that they did.
This colony is only two years old, and, as it happened, none of the men ever had camped out in this part of the Adirondacks before, and as they found their lake and surroundings quite sufficient there was not a tent on the place. However—and the expedition was avowedly got up for my benefit—I insisted that I wanted a genuine rough camping experience, and we all took Opp’s word for it that he knew the very spot—where there was fishing, a clearing, and an “open camp,” erected by other wood-loving spirits. It is true he grinned as
he assured me that I would get a good taste of the “genuine article,” but I suspected nothing. What imagination, indeed, would be equal to it!
Mrs. Coward kissed me good-bye quite affectionately, for she expected to “go out” before I returned, and even Mrs. Earle stood on the shore in the little spruce grove and waved her handkerchief with the others as we rowed down the lake.
It was one of those crystal mornings when life seems the divine thing of those imaginings of ours when we have lost for a little the links that hold them to facts. I never felt happier, I was almost excited. It seemed such a delightful thing to float off into the unknown like that, to go in search of adventures, with the certainty that six strong men, one of them your devoted slave, would take the best of care of you. It was all so undiscovered—that rough mountain world beyond the lake—so unimaginable—well, I know all about it now.
We were a very picturesque party, my dear. The men wore white sweaters, corduroy breeches, and top boots. I wore hunter’s green, a short skirt of covert cloth just above my boot tops, a linen blouse the same shade and a little bolero to protect my back and arms from the mosquitoes. Miss Page, who is very dark, wore a bright red skirt and cap and a red and white striped “shirt waist” with a red tie. Mr Nugent said she looked exactly like a “stick of peppermint candy,” and I am sure I shall recognise that indigestible the first time I enter a “candy store.” Mrs. Meredith Jones, who has golden hair and blue eyes, wore a dark blue skirt and cap and the inevitable “shirt waist”; but hers was striped with blue; and the jauntiest little cape hung from her shoulders. Of course we all wore canvas leggins as a further protection from the mosquitoes, which are the least of Adirondack charms.
Well, the moment we stepped on shore our troubles began. We were landed on to a big slippery stone, then handed across several others and a few rotten logs into a swamp. Before us was an impenetrable thicket as high as our heads and wet with dew. We stood staring at it until the guides had shouldered their packs and picked their way over rocks and logs to take the lead.
“That’s all right,” said Opp, “there ain’t bin any one in here for two years and the road’s growed over, but it’ll be all right in about a mile. Good trail then. We’ll go first and break the road. Wimmin folks’d better bring up in the rear.”
So we started; crashing through the wet bushes over the wetter ground until we came to a narrow rocky trail sidling along the inlet. This is a gentle stream in a wild setting. Its rocks are so many and so big that the wonder is the water can crawl over them, and the mountain beside the path is as precipitous as a cliff. None of us paid much attention to the beauties of Nature; we did not dare take our eyes off the path, which had given way in places and was swampy in others. Where it was safe it was rocky. Nor could the men help us much; the trail was too narrow. Single file was a necessity, but Mr. Nugent was just behind me and gave me occasional directions, besides surrounding me, as usual, with an atmosphere of protection. So, slipping, and bending and clutching at trees, we picked our way along until at last the trail turned up hill, and if no less rough was free of the worst element of danger. In another half hour we had passed a lumber camp and were on a level trail along the crest of the mountain. The forest was more open here, so much “lumbering” had been done, but only the spruce were gone—not all of those—and high on one side and down in a valley on the other was the beautiful leafy forest, full of the resinous odor of spruce gum, the spaces rather a welcome change after the forest densities of the last two months. And our procession was very picturesque. The guides with their big pack-baskets strapped to their shoulders were in the lead, almost trotting, that they might outdistance us and have an occasional rest. All our men carried small packs and strode along looking very supple and free, with the exception of poor Mr. Van Worden who is rather stout and must have felt the irksomeness of his pack. But he was enjoying himself, no doubt of that; and indeed, so were we all. Mr. Latimer, who had looked a little consciencestricken as he said good-bye to Mrs. Van Worden, whistled as gaily as a school-boy on a runaway lark. And it was so cool and fresh in the woods, who wouldn’t be happy? Not that there was one minute of easy walking—nor an opportunity for sentiment. When we followed the narrow trail through the brush we had to stoop and
overlook every inch before we put a foot down. When we were on the long stretches of corduroy, built by the lumbermen to haul their logs over, Mr. Nugent held my hand, but he might have been his ghost for all the impression he made on me, so many were the holes and so rotten some of the logs. Conversation was impossible. We exchanged an occasional remark, but we were all too intent on avoiding sprained ankles and broken tendons—you cannot imagine the painfulness of walking too long on log roads—to be interested in any one but ourselves.
There were four hours of this, and good a walker as I am I was beginning to feel tired, when Opp, who had gone for ahead, came in sight again, looking sheepish, rather.
“Be gosh!” he remarked to Mr. Van Worden as we met, “here’s a fine lay out. One of the camps is burned. Them last campers done it, I reckon. I seen ’em go round by way of Spruce Lake.”
I heard Mr. Van Worden swear softly under his breath, and saw an expression of blank dismay on Mr. Nugent’s face. Mr. Latimer burst into a peal of boyish laughter. But Mr. Meredith Jones said sharply,
“Well let’s go on and cook dinner. That is all that concerns us now. We can decide what to do later.”
“Are we there?” I asked, hopefully, for I longed to give my poor bruised feet a rest.
“Yes’m,” said Opp, “we’re there, all right.”
And in a moment, Polly, we “were there.”
Have you wasted any time, my dear, imagining what an “open camp” is like? I hope not, for it were a waste of good mental energy. The briefest description will fit it. Three sides and a sloping roof, all of bark. The front “open” in the exactest interpretation of the word. Inside—nothing. Twelve feet long and not quite the depth of Mr. Meredith Jones, who is six feet two.
This mansion stood on the edge of a clearing, across which lay a big felled tree. Against this we immediately all sat down in a row. Beyond was a charred ruin and near the log a rude table. Does that sound
romantic? I wish you could have seen it. But we all laughed and were happy, and we women, even then, did not realise the true inwardness of the situation. The forest, the beautiful forest, rose on three sides of us; beyond a stream, concealed by alders, was a high sharp ridge of mountains; and we were hungry.
The guides immediately set about making a fire. There seemed to be plenty of logs and they soon had a roaring blaze. Opp found a limb with a forked top, which he drove into the ground just beyond the fire and in the fork transfixed a long curving branch which held a pail of water above the flames. Mr. Nugent and Mr. Van Worden unpacked the baskets, Mr. Meredith Jones set the table, and Mr. Latimer fought off the hornets which swarmed at the first breath of jam and gingernuts. When we finally sat about that board, on logs or “any old thing,” we eat that excellent luncheon of fried ham and hard boiled eggs, mutton cutlets and fried potatoes, hot chocolate and cake, with a grateful appetite, I can assure you. Mr. Van Worden fried the ham and potatoes and made the chocolate, and we all coddled his culinary pride. All my fatigue vanished, and Mrs. Meredith Jones looked equally fresh and seemed prepared to take whatever might come, with the philosophy of the other sex. But poor Miss Page looked rather knocked up. She has never gone in for walking and her very cap had a dejected air; her fine colour was almost gone, but she looked very pretty and pathetic and all the men attempted to console her.
“I wouldn’t mind it,” she said with a sigh, “if we didn’t have to go back.” Then, as if fearing to dampen our spirits with the prospect of carrying her out, she added hopefully, “But it’ll be two days hence. I reckon I’ll be all right by that time. I’ll just lie about and rest.”
When luncheon was over Mr. Latimer made her a comfortable couch of shawls, with a small pack-basket for pillow, and she soon fell asleep. The guides washed the dishes, then immediately felled two young spruce-trees, and, with the help of Latimer and Mr. Meredith Jones, shaved off the branches and covered the floor of the cabin. This was our bed, my dear, and it was about a foot deep. When it was finished they covered it with carriage robes, and all preparations for nightly comforts were complete. By this time it had dawned on
