Infrared moving target detection and tracking based on tensor locality preserving projection

This paper presents an algorithm for detecting and tracking dim moving target in IR image sequence with low SNR. The algorithm, which is designed based on the locality preserving projection, accepts tensors as inputs. The justification for the algorithm comes from the role of the generalized eigenvalue problem in providing an optimal embedding for the manifold. Not only does the proposed method inherit the attractive characteristics of the locality preserving projections in terms of exploiting the intrinsic manifold structure, it is also appealing in terms of significant reduction in both space complexity and time complexity. Experimental results on two IR image sequences demonstrate the effectiveness of the proposed algorithm.