Research On Target Detection Of High Frequency Over-the-horizon Radar In Complex Clutter Environments

Over-The-Horizon Radar(OTHR)utilizes the high-frequency electromagnetic wave diffraction or reflection to achieve over-the-horizon detection for targets in long distance,providing ocean monitoring in a wide range.Since OTHR works in a complex electromagnetic environment,in the Azimuth-Range-Doppler(ARD)map,which is obtained from the converted the received signals of OTHR by the signal processing,there are not only targets and atmospheric noise,but also the nonhomogeneous and widely distributed clutter signals,such as ionosphere clutter,sea clutter.The effective target detection is a challenging task in such a complex environment of OTHR.The target detection of OTHR is usually performed in the RD map.Detection threshold is calculated by mainly utilizing the statistical characteristic of the amplitude of reference cells around cell-under-test(CUT)in conventional detection methods.However,through the analysis of real RD maps in logarithmic scale,there are other information of targets and clutter besides the statistical characteristic of their amplitudes.In different clutter regions,the adjacent background cells probably belong to the same type of clutter,and there is spatial correlation in category attributes.In a RD map,the average power of adjacent background cells in the homogeneous region is similar,while is varied slowly along the distance or Doppler dimension in a large-scale non-homogeneous clutter region.Owing to that the average power can be calculated from the statistical distribution parameters,the distribution parameters of adjacent background cells should also vary slowly in the clutter region with slowly varied mean power,and there is spatial correlation among them too.In addition,compared with the number of background cells,the number of targets is of an absolute minority and with sparse characteristics.Due to the front-end signal processing and other reasons,the target usually occupies multiple adjacent CUTs and shows a local expansion characteristic.Based on the analysis of the characteristics of RD maps,aiming at the problems of target detection in the actual RD map of OTHR,this dissertation proposes some methods to estimate the statistical distribution parameters of background cells by using the information of target sparsity and expansion characteristics,spatial correlation of background cells and spatial correlation of their distribution parameters,so as to achieve the purpose of improving target detection in the complex and nonhomogenous environment.The specific research works are as follows:Firstly,in a multi-target background,an algorithm of estimating the background distribution parameters using the target sparsity is proposed to solve the problem of target/outlier interference in target detection.In the proposed algorithm,the interfering target/outlier is identified adaptively by regularizing with sparsity on the targets.Then,all the homogenous cells are used to estimate the distribution parameter to improve the estimate accuracy.In the background of multiple distributed parameters,compared with the traditional target CFAR detection method,the proposed algorithm can identify the targets adaptively without the prior information,such as the number of targets and distribution parameters,to achieve the CFAR detection of targets.Simulation and experiment verify the feasibility,effectiveness and robustness of this algorithm with the sparsity of targets.Meanwhile,it also verifies the positive role of target sparsity in target recognition,which can provide support for the treatment of interfering target / outlier in the complex background.Secondly,in a background of clutter edge,aiming at the problem of nonhomogeneity of reference cells caused by the clutter edge in target detection,an algorithm is proposed to estimate the distribution parameter by using the spatial correlation of background cells.In the process of estimating the distribution parameter,the proposed algorithm identifies the background cells and locates the clutter edge by limiting the clutter category attribute of adjacent background cells,and then using all background cells in the same homogenous clutter area to estimate the distribution parameter,which improves the estimation accuracy.Finally,the recognition of clutter cells is improved.Compared with the traditional detection method,the proposed method can utilize the spatial correlation of background cells in the estimation process,and be not limited by the number of clutter edges contained in reference window and the number of reference cells.The performances of the proposed algorithm in clutter edge location,distribution parameter estimation and target detection are evaluated by simulations.Meanwhile,the experiments based on the actual RD maps of OTHR with embodied targets also verify the improvement of the proposed algorithm in target detection performance.Thirdly,in the clutter border background with the slowly varying average power of cells,aiming at the problem that the distribution parameters of adjacent cells are different each other in target detection,an algorithm is proposed to estimate the distribution parameters of cells by using the spatial correlation of the distribution parameters of cells.This algorithm uses the spatial correlation to associate the slowly varying non-homogeneous adjacent cells and determine the spatial relationship of their distribution parameters,so as to realize the estimation of the distribution parameters of the non-homogeneous cells.Compared with the traditional methods,the proposed method directly establishes the spatial relationship of the distribution parameters between adjacent background cells,which makes the estimated distribution parameters have the maximum likelihood significance.The results of simulations and experiments on actual data show that CFAR detection is of feasibility and effectiveness by using the spatial correlation of the distribution parameters of cells.Simultaneously,it also verifies the positive role of spatial correlation in improving the estimation accuracy of the distribution parameters in the area of nonhomogeneous clutter,and provides a strong support for estimating the distribution parameters of the cell in the complex non-homogeneous background.Fourthly,in a mixed clutter background,aiming at the problem of the nonhomogeneous reference cells composed of the mixed clutter in target detection,an algorithm is proposed to estimate the distribution parameters by adaptively using the spatial correlation of the distribution parameters of cells and utilizing the expansion characteristic of the sparsity point-like target.This algorithm can adapt to the complex variation of the spatial correlation of cells' distribution parameters in the background,associate the adjacent non-homogeneous background cells,determine their spatial relationship in distribution parameters,and estimate the distribution parameters of all cells.In the proposed algorithm,the broadened feature of sparsity point-like target is used to improve the recognition performance and reduce its adverse effect on the distribution parameter estimation.Compared with traditional algorithms,the proposed adaptively establishes the relationship between the amplitudes of adjacent background cells and the spatial information of their distribution parameters,which has a meaning of the maximum likelihood.According to results of simulations and verification by real data,the proposed CFAR detection is feasible and effective by adaptively using the spatial correlation of distribution parameters of cells and expansion of the point-like target.