The Algorithm Research Of Targets Image Recognition And Realization In DSP System

With the developing of image processing and DSP technology, the digital image processing combined with DSP technology is widely used in every department of national economy. The requirement of science and technology of the modern war is higher and higher. The recognition for plane targets is more and more important in the complex and dynamic air war environment. This paper discussed the technology of plane targets image recognition and its application in DSP. The content include as below:Firstly we discussed the edge detection of image. Based on the wavelet theory and the methods of multi-scale wavelet, we researched the modular-angle-separated (MAS) wavelet function which can distinguish edge from noise more effectively. When we analyzed the mathematical model of MAS wavelet and according to experiment, we will figure out that this methods can detect edge of target image effectively and much better than other kind of methods ever before.Secondly we researched the invariable moment of image edge and the edge would be studied in the context. We compared the affine invariable moment with the general invariable moment. According to the mathematical model and experiment results, affine invariable moment has the advantages of low dimension, large variance between classes and easy to distinguish. We can get a nice result in feature extraction of image and lay a solid foundation in design of classifier.Thirdly we put the result of invariable moment into classifier to identify different classes. The main work is design a nice classifier, then put the train sample into classifier for training and use the test sample for testing. We mainly applied the wavelet neural network (WNN) and the learning vector quantization network (LVQ) to design classifier and identify targets. Focus on the different networks we improved the algorithm and analyzed each kind of characteristic and get a content result.And then we briefly introduced the development of DSP and the structure of its software and hardware. Combined with the experiment platform DAM6416P, we discussed the designing methods of DSP program in image processing and its applications. So we can develop a target image algorithm for this platform.Finally we concluded the whole work of this paper and looked forward to the further of our research both in software and hardware.