Showing papers on "Contextual image classification published in 1983"

An appraisal of a decision tree approach to image classification

Abstract: This paper investigates the applicability to a shape-recognition problem of a concept learning algorithm which generates decision rules from examples. A comprehensive analysis of this algorithm applied to an industrial vision problem is described. This problem has no obvious 'best' solution and much effort has been devoted to performing a realistic appraisal of the algorithm by making a detailed set of comparisons with the performances of appropriate alternative classifiers. Results presented show the algorithm to be comparable in performance with the alternative classifiers but superior in terms of both the cost of making a classification and also the intelligibility of the solution.

Pattern Recognition Experiments in the Mandala/Cosine Domain

Abstract: The problem of recognition of objects in images is investigated from the simultaneous viewpoints of image bandwidth compression and automatic target recognition. A scenario is suggested in which recognition is implemented on features in the block cosine transform domain which is useful for data compression as well. While most image frames would be processed by the automatic recognition algorithms in the compressed domain without need for image reconstruction, this still allows for visual image classification of targets with poor recognition rates (by human viewing at the receiving terminal). It has been found that the Mandala sorting of the block cosine domain results in a more effective domain for selecting target identification parameters. Useful features from this Mandala/cosine domain are developed based upon correlation parameters and homogeneity measures which appear to successfully discriminate between natural and man-made objects. The Bhattacharyya feature discriminator is used to provide a 10:1 compression of the feature space for implementation of simple statistical decision surfaces (Gaussian and minimum distance classification). Imagery sensed in the visible spectra with a resolution of approximately 5-10 ft is used to illustrate the success of the technique on targets such as ships to be separated from clouds. A data set of 38 images is used for experimental verification with typical classification results ranging from the high 80's to low 90 percentile regions depending on the options choosen.

Robot Vision For Machine Part Recognition

Abstract: Most industrial applications of computer vision can be categorized into two groups. They are (1) visual inspection and (2) machine parts recognition. There are several review articles for automatic visual inspection [1,30,31]. This paper gives a brief review of robot vision system for machine part recognition. A robot vision system for machine parts recognition contains four sub-systems: (1) sensing, (2) segmentation, (3) description, and (4) recognition. A block diagram of such a system is shown in Fig. 1.© (1983) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

A parallel algorithm for muscle tissue images classification

Abstract: We report on the development and implementation of a parallel classification scheme for muscle tissue images. A sample image consists largely of fibers of three gray level values - black, gray and white. Global structure as well as local structure of the three different types of fibers are stressed as features in the classification scheme. We adopt a base-2 pyramid data structure by which an image is partitioned into 2/sup p/P x 2/sup p/ windows, p = 0, ---, n. At each level p, different information about the distribution of fibers can be obtained. Local processing is carried out concurrently in each window at the lowest level p = n. Number of black and white fibers is counted in each window at level p = n. Information at higher levels is expressed in terms of statistics at-the lowest level p = n. Five features were extracted and used for classification. Sample data was classified into three classes -- normal, intermediate and pathological, with a correct classification rate of 88%. This compared to an optimistic classification rate of 70% in a prior work[S]. The classification scheme obtained better performance than prior works, both in terms of speed and accuracy.

Performance Of Synthetic Discriminant Functions For Infrared Ship Classification

Abstract: Our unified synthetic discriminant function (SDF) filter synthesis technique using the correlation matrix of the image training set is reviewed Four different synthetic discriminant functions for intra-class recognition, inter-class discrimination and both intra and inter-class pattern recognition are considered All techniques proposed are appropriate for object identification, location and classification in the presence of 3-D geometrical distortions in the input object Initial results obtained on a set of four different classes of infrared ship imagery are presented Excellent performance (over 90% correct classification) was achieved© (1983) COPYRIGHT SPIE--The International Society for Optical Engineering Downloading of the abstract is permitted for personal use only

Linear Algebra Techniques For Pattern Recognition: Feature Extraction Case Studies

Abstract: Many linear algebra operations, matrix inversions, etc. are required in pattern recogni­ tion as well as in signal processing. In this paper, we concentrate on feature extraction pattern recognition techniques (specifically a chord distribution and a moment feature space). For these two case studies, we note the various linear algebra operations required in distortion- invariant pattern recognition. Systolic processors can easily perform all required linear algebra functions. 1. INTRODUCTION Linear algebra operations are required in many signal processing applications and these have been extensively discussed elsewhere in this volume. In this paper, I note that similar operations are also required in many pattern recognition and object identification applica­ tions. In this paper, specific attention is given to feature extraction or feature space based pattern recognition problems and to viable applications such as achieving object rec­ ognition in the face of geometrical distortions in the input image. The two feature extrac­ tion case studies considered are the use of a chord distribution and a moment feature space. Each of these results in considerably different linear algebra operations required on the object features to achieve the desired object identification. Sections 2 and 3 address the chord distribution feature space and Section 4 addresses the moment feature space case study.

Image Registration Through The Exploitation Of Perspective Invariant Graphs

Abstract: This paper describes two new techniques of image registration as applied to scenes consisting of natural terrain. The first technique is a syntactic pattern recognition approach which combines the spatial relationships of a point pattern with point classifications to accurately perform image registration. In this approach, a preprocessor analyzes each image in order to identify points of interest and to classify these points based on statistical features. A classified graph possessing perspective invariant properties is created and is converted into a classification-based grammar string. A local match analysis is performed and the best global match is con-structed. A probability-of-match metric is computed in order to evaluate match confidence. The second technique described is an isomorphic graph matching approach called Mean Neighbors (MN). A MN graph is constructed from a given point pattern taking into account the elliptical projections of real world scenes onto a two dimensional surface. This approach exploits the spatial relationships of the given points of interest but neglects the point classifications used in syntactic processing. A projective, perspective invariant graph is constructed for both the reference and sensed images and a mapping of the coincidence edges occurs. A probability of match metric is used to evaluate the confidence of the best mapping.

Statistical Approach For Forward Looking Infrared (FLIR) Target Classification

Abstract: A statistical approach for forward looking infrared (FLIR) target classification is presented. The implemented functions include enhancement, segmentation, feature extraction and classification. A 5 x 5 median filter is used for image smoothing. Segmentation involves an adaptive thresholding technique. This technique is capable of automatic selection of local thresholds of individual targets based on the local property in terms of minimal change in target area. Features extracted from segmented target candidates characterize grade shade, texture and geometry properties of these regions. Among several evaluated classifiers the Bayes decision rule is used for its performance, flexibility and future modifications. The presented approach has been applied to 92 FLIR images from three different data sets. Five types of target candidates examined in this study are tanks, APC's, jeeps, burning hulks, and noise regions. Among 281 targets of interest, 260 belong to these five categories. The Bayes classifier has achieved 87.69% detection and 76.92% classification with a FAR of 0.07 per image.© (1983) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Constructive Invariant Coupling Method Applications For Statistical Data Analysis And Effective Decision Rule Synthesis In Digital Image Processing Problems

Abstract: Some digital image processing problem types, such as sorting, reference comparison, classification, remote sensing, monitoring and quickest adaptive detection of image "disorders" are dealt with. Principal fundamental problems to be solved here are: 1) selection of informative features, and 2) as full as possible extraction of useful information out of data and its effective use. A solution approach is suggested for the above problems, based on the concept of invariant "coupling" of unknown parameters by some data functions, and integration or summation for these invariant "couplings". Here, on the one hand, it appears possible to derive a synthesis procedure for effective statistical decision rules which are not strictly dominated by any other decision rules with respect to specified loss functions, and, on the other hand, the decision rules per se are readily implementable on digital computers. Some new results have been obtained, showing the highest effectivity to be achieved when handling small data volumes. Several examples are presented.

The Perturbing And Iterating Classifier Algorithm

Abstract: A classification algorithm called the Perturbing and Iterating Classifier (PIC) is presented. This algorithm is a heuristic classifier that determines the classification of a segment by examining the number of self-consistent perturbations that are necessary for a segment's descriptor vector to become very close to a model descriptor vector. Unlike many other classifiers, this algorithm does not rely on the initial closeness or similarity of descriptor vectors. The theory of PIC is initially explained, an application of PIC in two dimensional shape matching is given, and then the physical interpretation of the algorithm is presented. An example of how PIC can discriminate shape over a wide range is also presented.