Find in Library

The indication of moving targets with low velocities, which is an extremely serious problem for traditional ground moving targets indication (GMTI), can be achieved relatively easily by shadow detection in VideoSAR imagery. Numerous image and video processing technologies have been applied to shadow detection in VideoSAR to improve the performance for GMTI. Among these processing technologies, background modeling is the key technology, which has received lots of research based on multi-frame imagery but little on single-frame imagery. This paper introduces the formation of the moving target shadows and adds the concepts of occlusion, umbra, and penumbra to prior work. In addition to this, the paper proposed a local feature analysis method based on single-frame imagery, which can accurately detect moving target shadows. Based on the result of this method, the background model can be reconstructed using single-frame imagery, which avoids the ”ghost” phenomenon in moving target detection algorithms based on multi-frame imagery. Finally, we replace the initial frame in the visual background extractor (ViBe) algorithm by the reconstructed background model we get, and the result shows that the ghosts are removed effectively.