The Research And Application Of Artificial Neural Network In Robot

On the base of robot vision recognition system, the thesis discusses neural network in robot research. According to the specific identification task that robot vision recognition system will accomplish, the thesis builds a neural network based visual recognition system.Through the research that this system will complete function, combining techniques of neural networks, image processing, image recognition and machine Vision and according to modular thinking, Neural Network Based visual recognition System is divided two parts, one is input information processing system, another is neural network recognition system. The thesis has studied and designed and realized these two parts.Details are as follows:First, the gray-scale of the two-dimensional image is expounded, the technology of image enhancement including smoothing, sharpening and filter is researched in-depth. The thesis gets thresholding image through adopting a best threshold value of image segmentation method. On base of image processing, the thesis defines a number of image features and studies and realizes the method for calculating the characteristics. The thesis has studied and realized the method for calculating volume of feature and has given experimental results.Second, Neural network recognizer's design. Through neural network research, the thesis chooses a kind of neural network model and analysis principle of neural network related parameter and neural network BP learning algorithm. According to shortcoming of BP algorithm, proposing the method that improves BP algorithm and realizes it.The thesis realizes the system with VC++6.0, Samples of image recognition shows that neural network based visual recognition system can distinguish the training sample.