Study On The Key Technologies Of Target Recognition Based On UWB Signal

Ultra Wide-band (UWB) wireless communication technology has the ability to penetrate the foliage and detect and recognize target obscured by foliage, it has a great potential applications. How to extract the features of target obscured by foliage from UWB signals and recognize the types of the targets are important key technologies need to be research. The topic of this thesis originates from the items such as NSFC (Natural Science Foundation of China) and Important National Science&Technology Specific Projects, and has important theoretical and practical significance.This paper mainly investigated the target recognition techniques based on UWB signal, and several innovative contributions are completed:In order to effectively extract the feature of the target, a target feature extraction method based on sparse representation is proposed. Firstly, a redundant dictionary should be built, the basis functions of the redundant dictionary can be selected from the sine function, wavelet function, etc., also can be obtained by learning from different target echo signals. Then, the sparse representations of the measurement UWB signals can be solved with the redundant dictionary, subsequently, the sparse features of targets are extracted from the processed the sparse coefficients. In the experimental verification with the foliage-concealed target echo signal waveforms, the extracted features based on sparse representation exhibit good discriminability, and the extracted features are able to improve the performance of target recognition effectively.In target recognition, the performance of support vector machine (SVM) is highly dependent upon the selection of parameters, and the traditional methods of SVM parameter selection are easy to fall into local minima. In this paper, two improved particle swarm optimization (PSO) algorithms are proposed to optimize the parameters of SVM. In the improved algorithms, the control parameters of the algorithms can adjust adaptively with the evolution of the algorithms, and chaotic search and differential evolution based local search are respectively integrated in the two algorithms to further improve the convergence speed and the abilities of searching for the global optima. Then, foliage-concealed target detection and recognition method based on improved PSO-optimized SVM is proposed. The experiment results show that the proposed algorithms have good globe search capability and fast convergence, can improve the performance of target recognition.In target recognition, the performance of SVM is highly dependent upon the selection of kernel function and its parameters. Wavelet SVM based on wavelet kernel function is constructed, and a hybrid quantum particle swarm optimization (HQPSO) algorithm is developed to select the parameters of WSVM. In HQPSO, the positions of particles are encoded by quantum bits; the quantum gate rotating is then explored to update particle positions in an efficiently parallel way. As a hybridization strategy, quantum-rotation-based local search is integrated in the lifetime learning to further refine individuals’ performance and accelerate their convergence toward the global optimality. Based on the excellent search capabilities of HQPSO, the performance of WSVM for target recognition is improved.For recognition of multi-target under multi-scenario (MTMS), a method for recognition of MTMS based sparse representation is proposed. This method is divided into two stages:firstly, two overcomplete dictionaries one for scenarios and one for targets are learned from measured real target echo signal waveforms via dictionary learning technique. Then the sparse representations of the monitored UWB signals are solved using the two redundant dictionaries, and the information of the target and scenario are obtained from the two sparse coefficient vectors. Compared with the basic sparse representation classification, the proposed method can effectively improve the performance of target recognition and significantly improve the efficiency of target recognition. Finally, the research works of the whole dissertation is summarized, and several valuable research directions of target detection and recognition based on UWB signal are discussed.