Research On Mimo-sar Array Design And Three-dimensional Image Recognition Method

MIMO-SAR is a new type of synthetic aperture radar with three dimension resolution capability. MIMO-SAR can obtain much more than the actual number of equivalent antenna array element with using multiple input and multiple output(MIMO) technology. Therefore, the cost and the platform weight can be greatly reduced. Compared with the 3D imaging of linear array MIMO-SAR, the planar array MIMO-SAR can get the 3D image of the target in several pulse repetition time and also can realize the forward looking work mode. MIMO-SAR antenna array layout directly affects the system cost and imaging quality. Therefore, this paper studies the MIMO-SAR planar array antenna optimization technology. In the aspect of recognition, 3D SAR image has more abundant target information. It provides a potential way to reduce the cost and improve the recognition. So this article studies the 3D SAR image recognition technique. The main contents and innovation are as follows:1、This article introduces the mathematical model of linear array and planar array antenna array, and analyzes the effect of non-uniform array different array element spacing uniform array direction map properties; This paper reviews the basic principles of SAR target recognition briefly, and focus on the deep learning of the target recognition method.2、This article puts forward an improved genetic algorithm for array antenna array optimization method based on MIMO-SAR. Firstly, in the constraint conditions of antenna array range, minimum element spacing and array element number fixed, using the phase center approximation principle, establish a receiving and transmission array joint optimization model of MIMO-SAR array optimization. Secondly, the genetic algorithm encoding method was improved, so as to meet the constraints of minimum element spacing. The optimization object is the peak side lobe antenna ratio and the width of the main lobe. Based on the adaptive crossover rate and mutation rate genetic algorithm, the planar array position optimization method is proposed; The experimental results show that the improved genetic algorithm can restrain the premature phenomenon and achieve the global optimal solution of the optimization problem. This optimization method can achieve lower side lobe and main lobe than general plane array optimization method.3、Study on MIMO-SAR 3D image recognition method. The backward projection imaging algorithm is introduced in this article, and the 3D SAR images of several kinds of targets are simulated. A 3D SAR image recognition method based on adaptive parameters for deep confidence network is proposed. On the basis of two-dimensional image recognition technology, this paper develops a 3D SAR image recognition method. First, a small amount of 3D SAR image is mapped into a large number of 2D SAR images. The error rate of recognition of each class of 3D SAR images is obtained indirectly by the recognition of 2D SAR images. Then we can determine whether the class of 3D SAR images can be correctly identified by setting the threshold to judge the error rate. Based on the idea of 3D image recognition, we can use the existing recognition method and experience of two-dimensional image. In the recognition method, this paper improves the original depth confidence network and proposes an adaptive method for setting depth confidence network parameters. The method uses cross-validation to complete depth confidence network parameters optimization. It can solve the problem that the previous parameter could not get the best network parameters. By comparing the experimental simulation analysis, the improved deep belief network has a certain advantage in the recognition rate and time.