Content-based Auto Exposure Control for On-board CMOS Camera

With the development of the Active Safety Technology on Automobile, more and more applications need the on-board camera to get the information of the road. In this paper we propose a new approach to auto exposure control bases on the content of the scene for on-board CMOS cameras. Traditional mean value based auto exposure algorithms are widely used for better visual quality. However, in automobile application concerned active safety technology, better image quality does not always mean better visual quality. Such applications include lane detection, vehicle detection and pedestrian detection. Thus we use a triple-slope CMOS camera to realize the content-based algorithm aiming at such a problem. Our approach finds out the lane markings in a pre-defined window of interest in the image, and then estimates the variables to feature the triple-slope response of the camera. Experiments show that the algorithm has a robust performance, and is able to adapt to the rapidly changing luminance of the environment and achieve an acceptable image in one or two frames.