Moving Target Detection And Motion Analysis From Infrared Image Sequences

Much research has been done on the estimation of optical flow, however, most of the research is tested using the image sequences formed by visible light, and seldom infrared images sequences. In addition, most of the research is focused on the performance of the algorithm, ignoring the feasibility in practice. This paper addresses the problem of testing all kinds of algorithms for estimation of optical flow on infrared image sequences by considering the performance and computational cost. The main purpose is to find an approach suitable for the missiles.Moving target detection and motion analysis from infrared image sequences is discussed in this paper. Optical flow is used for moving target detection. Much simulation is done on infrared images using all kinds of algorithms. The results show that the approach proposed by Horn&Schunck is the better for infrared image sequences. This result is different from the result on image sequences formed by visible light. The reason why other approaches fail is discussed. The computational cost of the approach proposed by Horn&Schunck is estimated and the hardware architecture for implementing it in real time is presented.The second part of this paper deals with motion analysis which is called SFM (structure from motion) problem, that is, recovering the structure and motion of the object, Two types of SFM algorithm are introduced-feature based and optical flow based. Simulations for time to collision are done on image sequences formed by visible light and infrared image sequences. A new method called gate-average method based on the approach supposed by Subbarao for the estimation of time to collision is presented.