High-resolution One-dimensional Distance Image Recognition Technology Research

The radar target recognition, which plays an important role in modern radar system, has been one of the key components of the present and future defense weapon system. Techniques for automated moving target recognition are required in many military and civilian applications. The scattering information collected by different radar systems is different. The range profiles of targets, containing more information for recognition, can be obtained more easily by high-range resolution radar than two dimensions profile. Many methods for feature extraction and classification in radar target recognition by range profiles are studied intensively and extensively in this dissertation. The main work of this dissertation is listed as follows:1. The scatter-center model is discussed. Three kinds of simulated point targets are designed and their range profiles at aspect angles are computed.2. The utility of phase information of complex high resolution radar is investigated on both theoretical analysis and experimental examinations. The experimental results show the complex signatures demonstrate much better classification performance than the magnitude-only ones by the appropriate feature extraction and BP neural networks, because it utilizes the information about targets of the phase and reduces the sensitivity of the range profiles.3. The dynamic aspect warping requires the warping path to start and finish in diagonally opposite corner cells of the warping matrix, which is constructed by test range profiles and the template sequence. The dynamic aspect warping based on the relaxation of the boundary condition is presented to cancel the limitation and provide better recognition of radar targets.4. The radar targets identification based on denoising with wavelet transformation and BP neural network is studied in the chapter. Because HRRP is stationary, and wavelet transformation posses favorable time-frequency localize property, so we use wavelet thresholding reducing noise. The results show coif wavelet base for range profile gain better denoising effect and coif-BP neural network improves the recognition performance.