|Proposed Statement of Research
William D. Braswell
Texas A&M University-Kingsville
Purpose of this Document
This document describes the proposed research to be undrtaken by myself, William D Braswell, while attending Texas A&M University - Kingsville as a candidate for the degree of Master of Science in Electrical Engineering and a NASA/Texas Space Grat Consortium Fellowhip recipient. The proposed research will be conducted under the supervision of Dr. Rajab Challoo of the Department of Electrical Engineering and Computer Science and Dr. Robert McLauchlan of the Department of Electrical Engineering and Industrial Engineering.
Description of Proposed Research
The proposed research will be part of a project already in progress in the Intelligent Control Systems Laboratory at Texas A&M University-Kingsville. The objective of the research project is to develop an artificially intelligent machine vision system using multiple neural networks to solve the problems of object recognition and localiztion. The premise is that by studying the physiological structure of human/mammilian visual system and emulating the aschitechture of, and process performed by, the biological neural structures, the proposed research wil1 yield a machine vision system capable of performing at least some of the complex functions of its biological counterpart. Note that it is not the objective of this project to duplicate precisely the form and function of the mammalian viausal system; rather, it is intended that the mammalian visual system be used as a model from which a viable machine vision system may be developed.
The system under development consists of a pair of CCD cameras and associated hardware, image acquisition hardware and a 486 PC-compatible computer modidifed for image processnig applications. The CCD cameras, analogous to biological retinas, will require a stereo pan of digitized images. These images wil1 be processed by a software simulation of the neural architecture and processing found in mammalian visual systems. The neural architecture will consist of multiple neural networks, implemented in software, to perfom the various processing functions of the retina ganglion cells, lateral geniculate nuclei, and the several layers of primary visual cortex common to most mammals. The specific archtechtures and methods of adapting the networks have yet to be determined; however, the mammalian visual system is largely a product of self-organization, and therefore much of the proposed research will probably involve self-organizing artificial neural networks.
Relevance of Proposed Research to Space Science and Engineering
In its original conception, the proposed research project was intended to serve as a means by which a robotic arm could visually sense potential obstacles and avoid them, as well as visually target objects with which it was working for the purpose of visually servoing the end-effector to the desired location. With further developemnt and modifications, such a system could potentially serve as the primary means of navigation for space-based autonomous mobile vehicles operating in a non-Earth environment, such as the Viking Mars lander, and could simultaneously provide autonomous vision-based navigation as well as autonomously collecting visual data concerning terrain features, visible atmospheric conditions, etc.
The proposed system could also be of direct benefit to human beings. A secondary motivation for developing a system based on the mammalian visual system is the potential for its application as a biomedical unplant that can be directly intefaced with the human neurovisual system in order to restore lost sight, correct abnormal vision problems, and possibly to enhance or extend normal visual capabilities, such as providing vision in the normally invisible infrared or ultraviolet portions of the electromagnetic spectrum.
Wednesday, 26-Mar-2003 21:49:22 CST
CSR/TSGC Team Web