Moving Objects Detection And Tracking Using Fisheye Camera

Moving objects detection and tracking are the hot and important topics of intelligent vehicle and video surveillance. Traditional objects detection methods based on features can process when the camera is both moving and static, but the processing speed and robustness for different kinds of objects are not very ideal; traditional background subtraction can process fast, but it doesn't work when the camera is moving; traditional optical flow method can process when the camera is both moving and static, but the speed of processing is slow because of the big computation of optical flow, it isn't suitable for real-time system. For that reason, researching a method which processing fast and robustly is necessary.In this thesis, we research a objects detection and tracking method based on ego-motion compensation to process the videos shot by fisheye camera. Main of works in this thesis are generating the motion area and tracking objects.Generating the motion area:first we get the mapping between the two consecutive frame images by study a static point images in two camera coordinates; second we use Harris corner detection and LK feature points tracking method to get a group of points which are the same points image in two image planes; third we estimate the parameters of the mapping; forth we get the image, which is compensated from the first frame image, by the mapping all of the points in the first frame image; fifth we got the different image between the second frame image and the compensated image.Objects tracking:we use particle filter to track the objects based on the different image.