Infrared Small Target Detection Based on Prior Constraint Network and Efficient Patch-Tensor Model

Infrared small target detection (ISTD) is a key technology in the field of infrared detection and has been widely used in infrared search and tracking systems. In this paper, a novel ISTD approach based on a prior constraint network (PCN) and an efficient patch-tensor model (EPT) is proposed. Firstly, the PCN trained by numerous synthetic image patches is employed to obtain the preliminary segmentation result of small targets, which is later used as a prior constraint. Then the EPT model deals with the target detection problem by solving an optimization problem of recovering low-rank and sparse tensor. Next, the prior constraint is further applied to the target component of the EPT model as a regularization. Finally, the joint PCN-EPT model can be solved efficiently by the Alternating Direction Multiplier Method, and the targets are obtained by applying a simple adaptive threshold segmentation to the obtained target component from the PCN-EPT. Experimental results on multiple real datasets show that the proposed model outperforms the state-of-the-art.