Moving Target Recognition Based on Transfer Learning and Three-Dimensional Over-Complete Dictionary

In radar target recognition using high-resolution range profile, moving target recognition is a challenging issue, due to the target-aspect angle variation. To address the problem, two key issues need to be solved. First, we need to reflect the target moving status. Next, we need to find the common knowledge among different target-aspect angles. Accordingly, a novel moving target recognition based on three distribution over-complete dictionary in conjunction with transfer learning is proposed. Specifically, we propose a three distribution over-complete dictionary to represent the target and extract its moving status by dictionary learning. Moreover, we structure the feature set with generation among target-aspect angles by using a transfer learning method. This framework can be trained by using a small number of samples from limited target-aspect angles to recognize the targets of other target-aspect angles. Another advantage of this method is that it is robust against signal noise rate variation. Simulation results are presented to demonstrate the effectiveness of the proposed scheme.