Research On Classification Of Vehicle And Software Design

With the development of technology, the function of modern radar is more powerful, which makes it more adaptive to the requirements of modern warfare. The radar target recognition technology has also obtained the corresponding development, for example, the successful development and application of new radar system help the target recognition system achieve a higher recognition rate. However, there is a larger quantity of active radar remaining old system, thus how to use them to implement target recognition, rather than through the new radar system, is a meaningful research subject according to the current real-world conditions. Many scholars have studied this subject from several aspects, among which, a kind of effective method is to use micro-Doppler effect. The existence of a component part’s motion adding to the overall movement causes micro-Doppler effect. For components that cause micro-Doppler effect, there geometric structure, movement form, composition material and so on determines the frequency and amplitude of the echo.For data collected in the ideal case, it can achieve ideal recognition rate based on micro-Doppler effect features. However, the real situation cannot guarantee ideal data. The wanted recognition rate cannot be achieved with the same features mentioned while the target is rather far away from radar and the data collected has low Signal-Noise-Ratio(SNR), thus noise robustness is important and noise reduction is needed. The wanted recognition rate cannot be achieved either while the target is obscured, especially when the part which can cause micro-Doppler effect is obscured. Considering the relationship between successive frames of the echo, the method of fusion is adopted in the information processing to achieve a higher recognition rate.The main work consists of two parts in this paper. One part is to do a further study on noise reduction and information fusion in the information processing of the narrow-band radar target recognition on the basis of existing research, and the other is to design and implement The Aided Software of Ground Moving Target Classification Base on Narrow-band Radar.The first part mainly introduces the theoretical basis of this article. First of all it gives the micro-motion models of wheeled and tracked vehicles and analyzes it. Secondly, based on the analysis of measured data, it introduces two methods that can suppress clutter. Thirdly, it introduces two kind of feature extraction methods both based on micro-Doppler effect and compares these two methods. Finally, it briefly introduces the basic theories of Support Vector Machine(SVM), which is mainly used in this paper.The second part mainly studies how to improve the classification performance when it identifies wheeled vehicles and tracked vehicles. Firstly, based on previous studies, it introduces two methods on noise reduction, which are verified by the measured data. And then, it studies the problem of information fusion and mainly implements feature level fusion of the echo. The feature level fusion is between data level fusion and decision level fusion, and has the advantage of reducing the amount of loss information as well as maintaining certain flexibility. This paper focuses on the optimization of weight needed in the implementation of the feature level fusion, while it implements the decision level fusion by voting for the purpose of comparison.The third part mainly introduces The Aided Software of Ground Moving Target Classification Base on Narrow-band Radar. This software is compiled by the mixed language of C++ and MATLAB, with the function of simulating the echo, classifying target and showing the recognition rate’s trend with the change of SNR.