Application Of Neural Networks In Radar Target Recognition

Since the radar was invented, scientists have been studying the technology of radar targetrecognition constantly. Radar target recognition was from artificial recognition to intelligentalgorithm recognition which represents the development direction of the radar targetrecognition. Artificial neural network is an important field of machine learning. And it hasmany advantages: strong fault tolerance; a strong ability to adapt; can be massively parallelprocessing. It has great importance, that artificial neural network can be used in radar targetrecognition, and new algorithms can be proposed.Here in, Support Vector Machines and the Self-Organizing Feature Map Neural Net areapplied in radar target recognition.Firstly, this paper introduces the problem of radar target recognition in detail. We canfind the accurate position through selecting threshold, removing the isolated points and theunreasonable clustering.Secondly, this paper recommends the principle of Support Vector Machines (SVM) andthe derivation process. Then it analysis the parameters optimization methods of SVM: CrossValidation (CV) and Particle Swarm Optimization (PSO). We also defines four test index.Finally, Experiments show that the SVM can be achieved very good results in radar targetrecognition.Finally, the Self-Organizing Feature Map (SOM) Neural Net is used to classify differentradar targets, which is also tested by real data. The results demonstrate the effectiveness ofSOM Neural Net. Then, SVM and SOM Neural Net are combined, which can solvethe SVM shortcomings by the target, but also reduce the computing time of SOM Neural Net.