Keith Copsey - Work

 

Summary:

Name: Keith Copsey

Location: Worcestershire, England

Email: kcopSPAMPROTECTIONDELETECAPITALSsey@gmail.com

Role:
Statistical Pattern Recognition and Data Mining Scientist


Specialities:
Statistical data analysis, statistical pattern recognition, data mining, Bayesian modelling,  machine learning, artificial intelligence, C/Mathematica/Matlab programming, technical authorship, technical bid writing.

Statistical Pattern Recognition, Third Edition:

Webb, A.R. and Copsey, K.D., Statistical Pattern Recognition, Third Edition, John Wiley and Sons, Chichester, 2011.  To appear.

http://eu.wiley.com/WileyCDA/WileyTitle/productCd-0470682280.html 

http://www.amazon.co.uk/Statistical-Pattern-Recognition-Andrew-Webb/dp/0470682280

Outline:

Statistical pattern recognition concerns the use of statistical techniques for analysing data measurements in order to extract information and make justified decisions.  It is a very active area of study and research, which has seen many advances in recent years. Topics such as data mining, web searching, multimedia data retrieval, face recognition, and cursive handwriting recognition, all require robust and efficient pattern recognition techniques.

This third edition provides an introduction to statistical pattern theory and techniques, with material drawn from a wide range of fields, including the areas of engineering, statistics, computer science and the social sciences.  The book has been updated to cover new methods and applications, and includes a wide range of techniques such as Bayesian methods, neural networks, support vector machines, feature selection and feature reduction techniques.  Technical descriptions and motivations are provided, and the techniques are illustrated using real examples.

Statistical Pattern Recognition, 3rd Edition:

  • Provides a self-contained introduction to statistical pattern recognition.
  • Includes new material presenting the analysis of complex networks and basic techniques for analysing the properties of datasets.
  • Introduces readers to methods for Bayesian density estimation and looks at new applications in biometrics and security.
  • Provides descriptions and guidance for implementing techniques, which will be invaluable to software engineers and developers seeking to develop real applications
  • Describes mathematically the range of statistical pattern recognition techniques.
  • Presents a variety of exercises including more extensive computer projects.
The in-depth technical descriptions make the book suitable for senior undergraduate and graduate students, in statistics, computer science and engineering departments.  Statistical Pattern Recognition is also an excellent reference source for technical professionals.  Chapters have been arranged to facilitate implementation of the techniques by software engineers and developers in non-statistical engineering fields.

PhD:

Title: Bayesian approaches for robust automatic target recognition

University: Imperial College of Science and Technology

Department: Statistics

Supervisors: Prof D. Hand, Imperial College, Dr A.R. Webb, QinetiQ Malvern.

Download: (kcopsey_phd_version_1_2.pdf)

Abstract:

This thesis documents studies into robust automatic target recognition (ATR), with an emphasis on ATR from radar measurements. The underlying aim of the research is to develop ATR systems that are robust to the variety of conditions that will be faced in operational use. Although motivated by ATR applications, the majority of the research is applicable to generic classification (discrimination) problems.  A variety of approaches have been proposed, all following a Bayesian formalism.

The initial research focuses on the use of Bayesian mixture modelling, to estimate the conditional distributions of the sensor measurements for each class of target. A procedure based on altering hyperprior distributions is investigated as a mechanism for improving the generalisation properties of the mixture model classifiers. Specifically, attempts are made to design a classifier that is robust to changes in target configuration and that can generalise to other targets of the same generic class.

This thesis argues that the development of robust ATR systems requires more than just application of classification algorithms designed using limited amounts of training data, and applied to single sensor measurements.  Specifically, it is argued that full use must be made of any extra information or data that is available.

A Bayesian inverse model based algorithm is developed to enable a classifier designed using data gathered from one sensor to be applied to data gathered from a different sensor (provided that physical and processing models for the sensors are known), thus addressing the issue of insufficient training data for operational sensors. Bayesian networks are proposed for incorporating contextual information and domain specific knowledge alongside the sensor measurements in an ATR system.  Finally, a Bayesian framework is developed that utilises measurements from two sensors, separated in time, to classify relocatable targets in a scene.

Journal Papers: 

  • Copsey K D, Webb A R, Bayesian gamma mixture model approach to radar target recognition. IEEE Transactions on Aerospace and Electronic Systems 39(4), October 2003. 

  • Copsey K D, Gordon N J, Marrs A D, Bayesian analysis of Generalized Frequency-Modulated signals. IEEE Transactions on Signal Processing 50(3), pp 725 –735, March 2002.

Selected Conference Papers: 

  • Varga M, Ducksbury P, Green A, Copsey K, Warner E, Ellis I O, Green A R, Hanka R, An automated breast cancer grading demonstrator - Pathscore.  10th Nottingham International Breast Cancer Conference, 18-20 September 2007.
  • Copsey K D, Automatic target recognition using both measurements from identity sensors and motion information from tracking sensors. Proc. of SPIE International Symposium on Defence and Security, Automatic Target Recognition XV, 28 March - 1 April 2005, Florida, USA, March 2005. 
  • Copsey K D, Lane R O and Webb A R,. Designing NCTR algorithms when operating sensor conditions differ from training conditions. Proc. of Radar 2004, International Conference on Radar Systems, Toulouse, France, October 2004. 
  • Copsey K D, Webber C J S, An adaptive unified algorithm for both detection and recognition. Proc. NATO RTO SET Symposium SET-080, Target Identification and Recognition using RF Systems, Oslo, Norway, October 2004.
  • Copsey K D, Lane R O, Manchanda S and Webb A R, Bayesian Approach to Recognising Relocatable Targets. Proc. NATO RTO SET Symposium SET-080. Target Identification and Recognition Using RF Systems. Oslo, Norway, October 2004. 
  • Copsey K D, Webb A R, Classifier design for population and sensor drift. Structural, Syntactic and Statistical Pattern Recognition, Springer Lecture Notes in Computer Science, Vol. 3138, pp 744-752, August 2004.
  • Lane R O, Copsey K D, Webb A R, A Bayesian approach to simultaneous autofocus and super-resolution. Proc. of SPIE Vol. 5427, Algorithms for Synthetic Aperture Radar XI, ed. E G Zelnio, F D Garber, April 2004.
  • Copsey K D and Webb A R, Bayesian networks for incorporation of contextual information in target recognition systems. Advances in Pattern Recognition, Springer Lecture Notes in Computer Science 2396, pp 709-717, August 2002.
  • Copsey K D and Webb A R, Bayesian approach to mixture models for discrimination. Advances in Pattern Recognition, Springer Lecture Notes in Computer Science 1876, pp 491-500, August 2000. 
  • Britton A, Copsey K D, Maskall G T, Webb A R, West K, Nonlinear feature extraction and Bayesian mixture model approaches to target classification using MMW ISAR imagery: a preliminary study.  Proc. of SPIE Vol 4033, Radar System Technology V, AeroSense 2000, 2000.