Slow Ground Moving Target Detection And Classification Based On FMCW Radar

Compared with pulse radar, the continuous-wave(CW) radar does not suffer from the range blind zone problem. The CW radar has been used widely for ground surveillance thanks to its outstanding performance anti-stealth, background clutter cancellation, as well as anti-jamming. In most scenarios, the targets with slow velocity have a great impact on the whole battlefield situation, and thus this kind of targets deserve to be observed. How to detect these targets is a hot research topic. What’s more, with the development of the modern technology and military tactical requirements, the radar are not only utilized for detection and range measurement. The recognition and classification would provide additional useful information for condition analysis, and thus it is of great necessity that the modern radar could have the capability of targets classification. In this thesis, a systematic study of the detection and classification problems in the CW radar are presented, and the contributions can be summarized as follows:First, the principle and signal preprocessing techniques of the CW radar are introduced. The signal model of multiple cycles and the frequency response of the beat signal are derived. Based on the difference of the targets characteristics and platform types, several methods for suppressing the clutter are proposed, such as the moving target indication(MTI), CLEAN and clutter map. These methods are effective for distinguish the clutter and targets, and provide the premise foundation for successful target detection for slow moving targets.Second, the range-velocity coupling in target detection is analyzed. For frequency modulated saw-tooth wave, by making use of the parametric estimation and ambiguous velocity estimation methods, the range and velocity of the targets can be calculated accurately. For triangle wave, based on the symmetry of the target’s spectrum, not only the conventional spectrum matching methods, but also the matching methods based on moving target detection in frequency domain and in time-frequency domain are introduced. Therefore, the targets can be matched accurately and the real range and velocity can be obtained.Third, the micro-Doppler spectrum of the target motion is analyzed. The motion of the slow moving pedestrians and vehicles are modeled in an analytical mathematical way. And the validation of the model is tested using the real-measured data. Further, based on the simulated and real-measured data, the differences of the micro-Doppler spectrum between pedestrians and vehicles are analyzed, and the feature vector for classification is determined.Finally, the classification scheme of the pedestrians and vehicles is presented. The energy, entropy, contrast, and correlation of the gray level co-occurrence matrix which extracted from spectrogram are used for composing of the feature vector. Then the principal component analysis(PCA) is applied to reduce the dimension of the feature vectors and get the eigenvalue of its. At last, the Eigen vectors is fed into the support vector machine(SVM) classifier for training and testing. The classification results indicates the effectiveness of the proposed scheme in my thesis.