Research On Exposure Method Of On-board Camera For Lane Detection

With the development of automobile active safety technology,the vision-based automobile active safety system gradually become the focus of research.As the main input device in automobile vision system,on-board camera has been more and more applied to the automobile active safety system.The vision based automobile active safety system identify and judge vehicle's own status and driving environment correctly and timely based on the image perception,image processing and pattern recognition technology,then in the early warning and auxiliary driving way to reduce the occurrence of accidents.In order to meet the needs of active safety technology in special application scenarios,the on-board camera should have higher brightness adaptability,higher dynamic range and higher reliability than ordinary cameras.From the important link of lane departure warning technique,the automatic exposure control method of the on-board camera is studied.Based on the high dynamic range CMOS image sensor,an adaptive exposure algorithm for lane mark detection is designed.The algorithm take road lane as content of interest,combined with the response of the camera to achieve real-time control of the triple slope response parameters of the camera,so as to obtain high dynamic range images.The main research work of this paper includes the following aspects:Firstly,the two kinds of image sensor of the on-board camera are introduced,and the principle and characteristics are compared and analyzed.Through analyzing the high dynamic range technology,we completed the selection of multi slope response image sensor.Then we made a comparative analysis of different architectures and cameras with different functions,and completed the selection of on-board camera.Then,the dynamic range,response characteristics and exposure mechanism of the image sensor are studied,and the on-board camera with multi slope response image sensor is developed,and the experimental platform is built,which realizes the programming control of various parameters of the response charact.Secondly,the traditional vehicle camera automatic exposure algorithm—the automatic exposure algorithm based on the image brightness mean value,the automatic exposure algorithm based on image brightness histogram and the automatic exposure algorithm based on multi-exposure fusion are studied and compared,and analyzed the characteristics and limitations.The experimental analysis is carried out aiming at the problem of slow convergence and concussion with the the traditional mean exposure algorithm.The principle analysis of the exposure algorithm based on the brightness histogram is analyzed,and the shortcomings of its application on the vehicle platform are explained.The experiment of multi exposure fusion algorithm is carried out,and the limitations of the algorithm in the application of vehicle vision system are analyzed.Thirdly,A high dynamic range scene image acquisition method based on multiple slope response control is studied,and an automatic exposure control algorithm for lane detection is proposed.Based on the analysis of the multi slope response principle of the image sensor,a mathematical model of the multiple slope response is established.Then,the difference between Lane brightness and road background brightness is maximized as the constraint condition of the mathematical model,so as to solve the parameters of multi slope response.Finally,in the global exposure control,based on the gray mean exposure algorithm,we add PID control method to achieve a fast and accurate automatic exposure control.The method of obtaining high dynamic range image is experimented and analysis.First introduces the concrete realization process of the new auto-exposure algorithm,The new algorithm and the traditional mean algorithm are compared from three aspects: the exposure speed,the image quality captured by the high dynamic scene and the success rate of the lane recognition of the captured image.From the analysis of the experiment,The new exposure algorithm in this paper has higher image quality than the traditional algorithm obtained in the complex illumination scene,and the success rate of lane detection in the later period is better,so the new exposure algorithm is more suitable to be used in vehicle vision system.