Moving Vehicles Detection in Airborne Video

A method to detect moving vehicles in airborne video is proposed in this paper. In order to deal with moving object detection in non-stationary background, the global motion of background is estimated frame by frame according to the correspondence control points extracted by Harris feature point extraction algorithm. The relative position of the feature points is used to remove the inadequate point pairs. With an affine transformation model, the background compensation is processed. Neighboring frames are aligned to maintain a stationary background. A statistical background modeling is applied to estimate background image and background subtraction is used to detect moving objects. Besides, a motion blobs mask is obtained from frame difference and detection results to update only areas of non moving objects in the background model. Experimental results show that the approach is able to obtain a robust detection result in airborne video.