The Target Recognition of Synthetic Aperture Radar Based on Locality-constrained Sparse Representation using Pairwise Constraints

Sparse representation technology has a wide range of applications in the field of SAR image target recognition. However, when the number of target images that can be acquired is small and only the target dual a priori information is known, the recognition performance obtained by this technique is degraded. To solve this problem, this paper proposes a SAR image target recognition method based on the dual-recognition feature local constraint coding sparse representation. Firstly, the paper constructs a new objective function, which introduces local constraint terms between samples and discriminates sparse coding error terms while considering image sparse reconstruction. Secondly, iterative optimization strategy is used to complete sparse coding. On the basis of the use of the gradient descent method to achieve online learning of the dictionary. The algorithm is verified by the MSTAR data measured in the United States. The simulation results show that the proposed method can obtain better recognition results and is a feasible method.