Camera Auto-Calibration Based on Motion Detection for Airborne Traffic Surveillance

Camera calibration is considered as a fundamental issue in computer vision which has been widely used in visual surveillance. Unfortunately, calibration information is rarely available in airborne traffic surveillance systems because of camera motions, wide-field scenes and dynamic scene structures. In this paper, we propose an automatic method to calibrate camera intrinsic and extrinsic parameters from airborne traffic surveillance videos. First, moving objects in airborne videos are detected by means of global motion estimation and pyramidal optical flow. Next, camera parameters are recovered linearly from vanishing points intersected by orthogonal lines which are estimated by optical flow and appearance of moving objects. Experimental results in different aerial videos indicate the feasibility and the effectiveness of our method which also considerably improves the accuracy of the calibration in airborne traffic surveillance.