No description available.

No description available.

A topic of computer vision that has been recently studied by a substantial number of scientists is the recognition of objects in digitized gray scale images. The primary goal of model-based object recognition research is the efficient and precise matching of features extracted from sensory data with the corresponding features in an object model database. A source of difficulty during the feature extraction is the determination and representation of pertinent attributes from the sensory data of the objects in the image. In addition, features which are visible from a single vantage point are not usually adequate for the unique identification of an object and its orientation. This paper will describe a regimen that can be used to address these problems. Image preprocessing such as edge detection, image thinning, thresholding, etc., will first be addressed. This will be followed by an in depth discussion that will center upon the extraction of local geometric feature vectors and the hypothesis-verification model used for two dimensional object recognition. / Department of Computer Science

In this study, fluorescein angiograms of the ocular fundus have been digitised to enable them to be processed and analysed by computer. A fully automated technique for counting microaneurysms (MA) in these images was developed with a view to producing an objective, accurate and highly repeatable way of quantifying these lesions. Prior to any other image processing, a number of pre-processing stages were applied in order to compensate for non-uniformaties and to remove the background fluorescence component present in all the images. Matched filters modelled on two-dimensional Gaussian distributions were employed to detect MA in the 'shade-corrected' images. A binary image representation of the vascular network was constructed. This 'vessel mask', used in conjunction with the original match-filtered images, enabled MA to be detected by grey-level thresholding the filtered images. The resulting binary objects could then be counted by the computer as MA. The automated technique was assessed by comparing the computer's results for six fluorescein angiograms with MA counts obtained by ophthalmologists analysing both analogue and digital images. The performance of both man and machine were judged with respect to 'gold standards' compiled from prints of the original negatives. The best results were obtained by the clinicians analysing the analogue prints, although they differed greatly in their ability to detect microaneurysms. The computer performed better than the clinicians when they were counting MA in the digital images and produced highly repeatable results. To improve the performance of the automated technique, images were captured at approximately four times the previous spatial resolution and a smaller area of each image was analysed. Additionally, more complex image-processing techniques were employed to increase the accuracy of the computer analysis. Although the performance of the automated technique was improved, the computer results only matched those of the clinicians' analogue analyses for two of the images.

Electric power systems are characterized by their immense complexity. The assessment of their security on-line has always been a challenging task. Many possibilities were investigated in the past in an attempt to characterize the secure operating region of a power system. Pattern recognition is thus far the only tool that can take various degrees of network complexity into consideration. / In the present study, an efficient algorithm which learns adaptively the secure operating region is proposed. At each iteration, training operating points are generated sequentially on a piecewise linearly approximated separation surface computed by the one-nearest-neighbor (1-NN) rule. The separation surface so estimated approaches the true one as the number of training points increases. The algorithm not only provides a consistent technique in learning an unknown region, it generates a highly efficient training set. It is found to be effective in reducing the size of the training set without adverse effect to the classifier. / Once the secure region of a power system is available, the task of on-line security monitoring reduces to one of determining whether the current operating point resides in the secure region. As demonstrated in the thesis, both the security status and the security margin of the operating point can be assessed very efficiently. By using the piecewise linearly approximated secure region, the thesis proceeds to give efficient ways of moving an insecure operating point into the secure region. This comprises the problem of security enhancement. / The regionwise methodology via the Voronoi diagram developed in the thesis is also applied to a wide range of problems, such as network planning, coordinating tuning of machine parameters and automatic contingency selection. The major merit is that the dynamics and the nonlinearity of the system no longer present a limitation to solving these problems.

The Fourier-Mellin transform was implemented on a digital computer and applied towards the recognition and differentiation of images of plant leaves regardless of translation, rotation or scale. Translated, rotated and scaled leaf images from seven species of plants were compared: avocado ( Persea americana), trembling aspen (Populus tremuloides), lamb's-quarter (Chenopodium album), linden (Tilla americana), silver maple (Acer saccharinum), plantain (Plantago major) and sumac leaflets (Rhus typhina ). The rate of recognition was high among translated and rotated leaf images for all plant species. The rates of recognition and differentiation were poor, however, among scaled leaf images and between leaves of different species. Improvements to increase the effectiveness of the algorithm are suggested.

Traditional Automatic Target Recognition (ATR) Systems often fail when faced with complex recognition tasks involving noise, clutter, and complexity. This work is concerned with implementing a real time, vision based ATR system using an Artificial Neural Network (ANN) to overcome some of the shortcomings of traditional ATR systems. The key issues of this work are vision, pattern recognition and artificial neural networks. The ANN presented in this thesis is inspired by Prof. Stephen Grossberg's work in Adaptive Resonance Theory (ART) and neurophysiological data on the primate brain. An ANN known as Selective Attention Adaptive Resonance Theory (SAART) (Lozo, 1995, 1997) forms the basis of this work. SAART, which is based on Grossberg's ART, models the higher levels of visual processing in the primate brain to provide an ATR system capable of learning and recognising targets in cluttered and complex backgrounds. This thesis contributes an extension to the SAART model to allow a degree of tolerance to imperfections including distortion, changes in size, orientation, or position. In addition to this extension, it is also demonstrated how modulated neural layers can be used for image filtering. A possible extension of the architecture for multi-sensory environments is proposed as a foundation for future research. / Thesis (MEng)--University of South Australia, 2000

Includes bibliographical references (leaves 225-241). / xxviii, 241 leaves : ill. ; 30 cm. / Title page, contents and abstract only. The complete thesis in print form is available from the University Library. / Proposes an architectural framework based on neural networks for visual scene analysis with attentional mechanisms. / Thesis (Ph.D.)--Adelaide University, Dept. of Electrical and Electronic Engineering, 2001

Traditional Automatic Target Recognition (ATR) Systems often fail when faced with complex recognition tasks involving noise, clutter, and complexity. This work is concerned with implementing a real time, vision based ATR system using an Artificial Neural Network (ANN) to overcome some of the shortcomings of traditional ATR systems. The key issues of this work are vision, pattern recognition and artificial neural networks. The ANN presented in this thesis is inspired by Prof. Stephen Grossberg's work in Adaptive Resonance Theory (ART) and neurophysiological data on the primate brain. An ANN known as Selective Attention Adaptive Resonance Theory (SAART) (Lozo, 1995, 1997) forms the basis of this work. SAART, which is based on Grossberg's ART, models the higher levels of visual processing in the primate brain to provide an ATR system capable of learning and recognising targets in cluttered and complex backgrounds. This thesis contributes an extension to the SAART model to allow a degree of tolerance to imperfections including distortion, changes in size, orientation, or position. In addition to this extension, it is also demonstrated how modulated neural layers can be used for image filtering. A possible extension of the architecture for multi-sensory environments is proposed as a foundation for future research. / Thesis (MEng)--University of South Australia, 2000

Thesis (Ph. D.)--University of Texas at Austin, 2007. / Vita. Includes bibliographical references.