Camera deflection detection and image correction for applications in ADAS algorithms

Cameras that are mounted on almost every modern vehicle can be deflected due to the different reasons such as mechanical or environmental. This deviation of the camera from reference position can affect performance of the camera based Advanced Driving Assistant Systems (ADAS) such as lane detection. In this paper one possible approach for camera deflection detection and image correction is given. The proposed approach is based on estimation of homography matrix that is used for image transformation after which nearest-neighbor interpolation is performed. The presented algorithm for image correction is implemented in TDA2xx System-On-Chip on ADAS Alpha development board. The approach is evaluated for different deflections of the automotive camera from the reference position.