基于KLT的动态目标检测与状态估计

Motion detection and state estimation is an important field of computer vision. It is also widely used in mobile robots, unmanned aerial vehicles and other mobile platforms. This thesis proposes a practical motion detection and state estimation algorithm which can be used in moving cameras.Firstly,we transform images based on KLT algorithm, then combine the frame difference algorithm, and create a motion detection algorithm that can be used for moving cameras.Secondly,we use particle filter to estimate the target's state. We model the target and choose the difference between two images as an analysis object, to develop a target's state estimation algorithm which make use of previous motion detection results.Experiments indicate that our method can detect the target and estimate its state in image sequences effectively even in occlusion situations