Research On Lane Detection System Based On Automatic Exposure

With the increase in car ownership,the hazards caused by traffic accidents are also increasing.The issue of traffic safety is gradually being emphasized seriously.In order to improve traffic safety,automotive safety technologies,especially active safety technologies have been continuously developed.The active safety system judges potential threats by detecting the environment around the vehicle in real time and responds or prompts in advance.In the vehicle environment perception system,lane detection is an important link.Therefore,it is of great practical significance to study lane recognition systems.This article used Blackfin534 chip of ADI company as the core chip of the DSP system and Agilent's MT9V022 sensor is used as an image sensor to build a lane detection platform.Based on the above,the following studies were conducted:(1)An object-based auto-exposure algorithm is proposed to solve the problem that a picture obtained by a single exposure parameter cannot clearly display a lane line in a complex lighting environment.First of all,the histogram is used to determine the matching of the dynamic range and the illumination range of the image sensor;then,the dynamic range expansion trend is judged according to the clearness of the lane line;finally,the dynamic range of the imaging is adjusted by the method of multi-slope integration so that the image The lane line is maintained in a high contrast state,which improves the recognition rate of the lane line recognition algorithm.(2)Aiming at the problem of too many feature points in the process of lane line recognition leading to a large amount of fitting operations,an optimization algorithm for lane line recognition is proposed: Uses improved Sobel operator in the binarization process,retains features of gradient ascent and descent,and uses ascending and descending features as criteria for lane line edges.Further,an algorithm for distinguishing between the dashed line and the solid line is proposed: In the narrow region,the continuity difference between the dotted line and the solid line is enlarged,and the variance of the feature points in the region is distinguished between the dotted line and the solid line.Considering the error caused by the vehicle shake in the recognition process,the Kalman filter is used to smooth the distance between the vehicle and the lane line,and give the Kalman model parameters of uniform and uniform acceleration.(3)Transplantion the automatic exposure algorithm and lane recognition algorithm to the hardware platform and optimization the real-time performance of the algorithm from two aspects of hardware and software.The recognition time of each frame is compressed to about 24 ms.The actual vehicle experiment shows that the automatic exposure algorithm can improve the image quality effectively in the case of excessive exposure and insufficient exposure.The recognition system meets the real-time requirement with the optimization processing of hardware performance,with optimization of hardware performance.Through the real vehicle test under different scenes,the system designed in this paper can identify the lane line in a variety of complex lighting environment,which has good accuracy and robustness.