Image recognition based on Neural Network (NN), merging the two technologies Image Recognition and NN, is a new kind of technology, which is developed with the theory of computer science, image processing, artificial intelligence and pattern recognition.The theory of image recognition and NN were introduced in the beginning of this article, with emphasis on the introduction of feed-forward NN model and Back Propagation (BP) algorithm which are commonly used. After the discussion of both advantage and disadvantage of standard BP algorithm, some improved BP algorithms were introduced. The method of image recognition using BP NN without feature extraction used in this paper overcomes the difficulties in feature extraction under blurred backgrounds, whose feasibility and good robust were validated by the results of simulation experiments.With theoretical analysis and experiments, this image recognition method was introduced into an image tracking system. After a new image matching function was designed and with the help of traditional technologies usually used in image tacking system, we fulfilled the image tracking method based on a BP NN and received some useful results. |