Learning to detect small target: A local kernel method

Abstract Small target detection is a critical problem in the Infrared Search And Track (IRST) system. Although it has been studied for years, there are some challenges remained, e.g. cloud edges and horizontal lines are likely to cause false alarms. This paper proposes a novel local learning framework to detect infrared small target in heavy clutter. First, we propose a quadratic cost function to learn the parameters in the weighted local linear model. Second, we introduce the kernel trick to extend the linear model to the nonlinear model. Finally, small targets are detected in the residual image which subtracts the estimation image from original input. Our method could preserve heterogeneous area while removing target region. Experimental results show our method achieves satisfied performance in heavy clutter.