Robust unmanned aerial vehicle camera self-calibration for surveillance applications

Unmanned Aerial Vehicles (UAV) become more important in many applications in urban environments and defense applications. Localization and tracking tasks are usually done using a camera mounted in the UAV. Most papers make assumptions about the camera parameters or interest points in the scene. This paper presents a new method to calibrate automatically the camera from a few images. We first present a method to filter outliers from feature points and then use rank 4 factorization to estimate the camera parameters. It does not require any known object or known 3D location point in the scene. Outdoor experimental results show that the feature points have been correctly selected in the scene and the camera can be calibrated correctly with a small reprojection error.