Research Of Mobile Robot's Visition System Based On Neural Network

Nowadays, more and more robots are used in automatic transmission of ware and material, automatic assembly, automatic operation in danger occasion , in serve trade and so on. To accomplish these tasks, it requires the robot's vision system to have better function. So the mobile robot need advanced vision system in road recognition and road following, and the automatic assembly robot and other robots need advanced vision system too. The automatic assembly robot need to recognize not only simple detail, but also involved unit. All these require the robot vision system with better recognition ability. In this thesis, the robot's vision system based on neural network are studied. The main work include:1. Introduced and analysed the research status of the robot's vision technology, the pattern recognition technology and the neural network at present.2. To design vision system for robots is the main task of robot's vision technology. The robot's vision system makes the robot to react to the surroundings with the intelligent and flexible method. The robot's vision technology is processing technology based on the image, the original images usually have to be processed such as eliminating noise and image segmentation, and then can be recognized by the computer. The image processing technology is studied and introduced in this thesis.3. Usually, object features include gray-scale (density, colour), fexture, geometry shape, and so on. Measure some feature of one image and take it to classifier, this is an important step of pattern recognition. At same time, neural network has some requirement for the inclusion of the feature parameter. The feature detection methods are studied , and the simple and stable invariant quadarture is decided and abstracted as the feature of image.4. To solve the pattern recognition problems, the neural network can achieve the involved demarcate of earmark space, and it cut out for high acceleration and parallel systems to achieve. The neural network technology and feature abstraction method is combine to recognition the picture in this thesis, and the ART-2 neural network is adopted. Did a amount of simple and involved experiments on picture recognition. The result of the experiments proved that the ART-2 neural network pattern is an ideal pattern of pattern recognition. The system based on theART-2 neural network can recognize the object's shape very well, and it will be adopted more and more in future.5. The program of the image processing, feature detection and pattern recognition are programmed with Visual C++ 6.0.