No description available.

No description available.

This thesis presents some substantial theoretical analyses and optimal treatments of Kohonen's self-organising map (SOM) algorithm, and explores the practical application potential of the algorithm for vector quantisation, pattern classification, and image processing. It consists of two major parts. In the first part, the SOM algorithm is investigated and analysed from a statistical viewpoint. The proof of its universal convergence for any dimensionality is obtained using a novel and extended form of the Central Limit Theorem. Its feature space is shown to be an approximate multivariate Gaussian process, which will eventually converge and form a mapping, which minimises the mean-square distortion between the feature and input spaces. The diminishing effect of the initial states and implicit effects of the learning rate and neighbourhood function on its convergence and ordering are analysed and discussed. Distinct and meaningful definitions, and associated measures, of its ordering are presented in relation to map's fault-tolerance. The SOM algorithm is further enhanced by incorporating a proposed constraint, or Bayesian modification, in order to achieve optimal vector quantisation or pattern classification. The second part of this thesis addresses the task of unsupervised texture-image segmentation by means of SOM networks and model-based descriptions. A brief review of texture analysis in terms of definitions, perceptions, and approaches is given. Markov random field model-based approaches are discussed in detail. Arising from this a hierarchical self-organised segmentation structure, which consists of a local MRF parameter estimator, a SOM network, and a simple voting layer, is proposed and is shown, by theoretical analysis and practical experiment, to achieve a maximum likelihood or maximum a posteriori segmentation. A fast, simple, but efficient boundary relaxation algorithm is proposed as a post-processor to further refine the resulting segmentation. The class number validation problem in a fully unsupervised segmentation is approached by a classical, simple, and on-line minimum mean-square-error method. Experimental results indicate that this method is very efficient for texture segmentation problems. The thesis concludes with some suggestions for further work on SOM neural networks.

No description available.

This thesis begins by providing a review of techniques for interpreting the thermal response at the earth's surface acquired using remote sensing technology. Historic limitations in the precision with which imagery acquired from airborne platforms can be geometrically corrected and co-registered has meant that relatively little work has been carried out examining the diurnal variation of surface temperature over wide regions. Although emerging remote sensing systems provide the potential to register temporal image data within satisfactory levels of accuracy, this technology is still not widely available and does not address the issue of historic data sets which cannot be rectified using conventional parametric approaches. In overcoming these problems, the second part of this thesis describes the development of an alternative approach for rectifying airborne line-scanned imagery. The underlying assumption that scan lines within the imagery are straight greatly reduces the number of ground control points required to describe the image geometry. Furthermore, the use of pattern matching procedures to identify geometric disparities between raw line-scanned imagery and corresponding aerial photography enables the correction procedure to be almost fully automated. By reconstructing the raw image data on a truly line-by-line basis, it is possible to register the airborne line-scanned imagery to the aerial photography with an average accuracy of better than one pixel. Providing corresponding aerial photography is available, this approach can be applied in the absence of platform altitude information allowing multi-temporal data sets to be corrected and registered.

A number of researchers have investigated the application of neural networks to visual recognition, with much of the emphasis placed on exploiting the network's ability to generalise. However, despite the benefits of such an approach it is not at all obvious how networks can be developed which are capable of recognising objects subject to changes in rotation, translation and viewpoint. In this study, we suggest that a possible solution to this problem can be found by studying aspects of visual psychology and in particular, perceptual organisation. For example, it appears that grouping together lines based upon perceptually significant features can facilitate viewpoint independent recognition. The work presented here identifies simple grouping measures based on parallelism and connectivity and shows how it is possible to train multi-layer perceptrons (MLPs) to detect and determine the perceptual significance of any group presented. In this way, it is shown how MLPs which are trained via backpropagation to perform individual grouping tasks, can be brought together into a novel, large scale network capable of determining the perceptual significance of the whole input pattern. Finally the applicability of such significance values for recognition is investigated and results indicate that both the NILP and the Kohonen Feature Map can be trained to recognise simple shapes described in terms of perceptual significances. This study has also provided an opportunity to investigate aspects of the backpropagation algorithm, particularly the ability to generalise. In this study we report the results of various generalisation tests. In applying the backpropagation algorithm to certain problems, we found that there was a deficiency in performance with the standard learning algorithm. An improvement in performance could however, be obtained when suitable modifications were made to the algorithm. The modifications and consequent results are reported here.

Urban regions present some of the most challenging areas for the remote sensing community. Many different types of land cover have similar spectral responses, making them difficult to distinguish from one another. Traditional per-pixel classification techniques suffer particularly badly because they only use these spectral properties to determine a class, and no other properties of the image, such as context. This project presents the results of the classification of a deeply urban area of Dudley, West Midlands, using 4 methods: Supervised Maximum Likelihood, SMAP, ECHO and Unsupervised Maximum Likelihood. An accuracy assessment method is then developed to allow a fair representation of each procedure and a direct comparison between them. Subsequently, a classification procedure is developed that makes use of the context in the image, though a per-polygon classification. The imagery is broken up into a series of polygons extracted from the Marr-Hildreth zero-crossing edge detector. These polygons are then refined using a region-growing algorithm, and then classified according to the mean class of the fine polygons. The imagery produced by this technique is shown to be of better quality and of a higher accuracy than that of other conventional methods. Further refinements are suggested and examined to improve the aesthetic appearance of the imagery. Finally a comparison with the results produced from a previous study of the James Bridge catchment, in Darleston, West Midlands, is made, showing that the Polygon classified ATM imagery performs significantly better than the Maximum Likelihood classified videography used in the initial study, despite the presence of geometric correction errors.

No description available.

No description available.

This research is concerned with the application of neural network techniques to the problems of classifying images in a manner that is invariant to changes in position and scale. In addition to the goal of invariant classification, the network has to classify the objects in a hierarchical manner, in which complex features are constructed from simpler features, and use unsupervised learning. The resultant hierarchical structure should be able to classify the image by having an internal representation that models the structure of the image. After finding existing neural network techniques unsuitable, a new type of neural network was developed that differed from the conventional multi-layer perceptron type of architecture. This network was constructed from neurons that were grouped into feature detectors. These neurons were taught in an unsupervised manner that used a technique based on Kohonen learning. A number of novel techniques were developed to improve the learning and classification performance of the network. The network was able to retain the spatial relationship of the classified features; this inherent property resulted in the capability for position and scale invariant classification. As a consequence, an additional invariance filter was not required. In addition to achieving the invariance property, the developed techniques enabled multiple objects in an image to be classified. When the network had learned the spatial relationships between the lower level features, names could be assigned to the identified features. As part of the classification process, the system was able to identify the positions of the classified features in all layers of the network. A software model of an artificial retina was used to test the grey scale classification performance of the network and to assess the response of the retina to changes in brightness. Like the Neocognitron, the resulting network was developed solely for image classification. Although the Neocognitron is not designed for scale or position invariance, it was chosen for comparison purposes because it has structural similarities and the ability to accommodates light changes in the image. This type of network could be used as the basis for a 2D-scene analysis neural network, in which the inherent parallelism of the neural network would provide simultaneous classification of the objects in the image.