Moving Target Detection and Tracking in FLIR Image Sequences Based on Thermal Target Modeling

Abstract

Moving target detection and tracking is an important research field of video processing for its great potential in Military and Civil applications. Motion compensation, motion detection and target tracking are the three fundamental parts of such applications; however, the gravity of each one is purely dependent on the applications being modeled. A novel approach for detecting and tracking a moving target in Forward Looking Infra-Red (FLIR) is proposed. In the proposed approach, a real FLIR imagery from a stabilized moving platform is used to validate the robustness and effectiveness of methodology. In one of the sequence (Seq-1), a stationary target is being detected and tracked from a moving platform. In other sequence (Seq-2), a moving target is being detected and tracked from a maneuvering platform. Single target was tracked in these sequences by processing FLIR stream through a number of stages. Platform motion compensation is not considered being stabilized platform. Moving targets in the sequences are detected by using spatio-temporal Intensity Projection (STIP) techniques. This information is used as seed for finding the location and size of the target in the image space. In the next stage tracking window is opened which fully encompass the target keeping its maximum motion. The search widow is maximized by a 3×3 kernel and segmentation is performed by OTSU method. The segmented regions are considered part of the target based on distance metric and center of gravity of the mass is determined in the search window. The process is performed in a loop to achieve the tracking. The proposed method is implemented in Matlab and experimental results using real FLIR sequence are shown to validate. The results are promising which demonstrate the robustness and good performance of the proposed method.