Research On Target Simulation And Discrimination Tecnology Of Wideband Radar

Synthetic Aperture Radar(SAR)can obtain SAR image with high resolution,which contains important information such as target structure and size.Automatic Target Recognition(ATR)technology can extract useful information of targets in SAR images,so it is very important in battlefield perception and is a hot topic in domestic and foreign research.The target SAR image simulation technology is not only the basis for the realization of the model-based SAR ATR system,but also can provide a lot of cheap data support for the research of the target recognition algorithm,which has a wide application prospect in the SAR image interpretation field.Based on the goal of the SAR image simulation technology and its application in SAR target identification as the main research content,the first systematic study a target SAR image simulation method,on this basis,based on the experiment of the measured data and simulation data,carry out a variety of methods of feature extraction and target identification method of research.The specific work of this paper is as follows:1.In view of the low efficiency of current mainstream SAR image simulation methods,which makes it difficult to generate a large number of SAR simulation data efficiently,this paper studies an algorithm for SAR image simulation.Firstly,the calculation method of scattering energy in the traditional Ray-Tracing algorithm is improved,and the calculation is carried out strictly according to the scattering mechanism of the electromagnetic wave by the target,so as to reduce the error of energy calculation,and the conversion from 3D model to 2D SAR image is directly completed according to the SAR imaging mechanism.Secondly,aiming at the computational redundancy in traditional tracking methods,a scientific partitioning mechanism is introduced,and the scene space is divided adaptively by SAH-BVH algorithm,which reduces the computational redundancy and optimizes the tracing process.Experimental results show that the proposed method can improve the efficiency of imaging simulation significantly on the premise of ensuring image quality.2.Traditional SAR image identification features often require manual design,which is not only dependent on the professional ability of practitioners but also time-consuming and laborious.To solve this problem,target discrimination algorithms based on Deep Belief Networks(DBN)and Convolution Neural Networks(CNN)are studied in this paper,andthree Convolution Neural Networks with different structures are designed.The discrimination effect of different Network models is analyzed through experiments.Since the deep network can automatically learn features from the input data without human intervention,the method achieves better identification performance.3.Target samples are the basis of SAR target identification.Both feature extraction and algorithm research cannot be separated from the support of a large number of target samples.In order to improve the accuracy and generalization performance of the identification method,this paper studies the method of using simulated SAR data in training sample sets.Experimental results demonstrate the effectiveness of this method and provide a possible way to improve the identification performance under limited measured samples.