Moving Object Recognition Based On Cellular Neural Network

Since the introducing of Cellular Neural Network, it has been widely studied because of its high speed character of data processing. Now it has been widely used in image and video signal processing, robotic, biological visions, higher brain visions, partial differential equations and other fields. The reason of research on the application of cellular neural network in moving object recognition is the high speed, paralleling operation character of the CNN and the network construction suitable for image processing. Now, the whole operational speed of CNN chip existing abroad has reached T magnitude, which can greatly improve the speed and the real time of sequence images processing. So, it is necessary to study the segment and the tracking of moving object in video image.Choosing and improving the three images subtraction algorithm as the image segmentation method after studying and comparing the usual algorithms, and it is suitable for the architecture of the cellular neural network. The method is simple and its templates can be easily designed. First, the motion area can be extracted by the difference of the first and the second images, then using the third image we can get the area of the moving object in the second image that has the clear contour. The clear contour is very important in the next processing.The recognition of moving object is important and difficult. Now, CNN has been used for image primary processing abroad, and the processing of recognition uses other methods such as DSP algorithm. Through the studying on Hausdorff distance and the nonlinear Hausdorff distance in the foreign papers, have found out the proper algorithm of CNN for tracking the moving object based on nonlinear Hausdorff distance metric, and there is the trigger autowaves method for its implement in CNN architecture . It is simple and the result is very good in sequence images processing.Experimental results show that the proposed approach is a promising method for segment of the moving object in sequence images, and the object can be well matched. But it is only suitable for the static background and it is going to be improved in the future.