The Study Of Single Visual Lane Detection In ADAS

Lane intelligent identification is a very important part of ADAS research.Through the real-time lane video recorded by the OBD front-facing camera,the intelligent lane recognition technology based on monocular machine vision carries out real-time and rapid intelligent recognition of the admitted lane in video through the analysis of algorithm and program and reminds the driver of the deviation from the lane by visual or auditory methods to avoid danger.In this paper,enough tests and researches have been conducted on the detection of key lanes in ADAS and other relevant investigations.The goal is to develop a method that can be applied to the detection of ADAS lanes on highways.Finally,this method can eliminate the useless part of video,but it will not mistakenly identify the part of lane video that needs to be identified and processed.Firstly,this paper introduces the types and parameter information of cameras used in recording the actual real-time lane video,and then lists the gray-scale processing that has been widely used in the video processing of monocular camera as its preprocessing method and makes improvements according to the actual application scenarios.Then USES the median filter occurring in the process of image smoothing method to remove noise,and then compare common edge detection algorithm in video processing Roberts operator,Sobel operator,etc.,in many aspects of comparative method in this paper,after selecting the Canny operator as a method of image edge reinforcement to the end of image binarization processing,this article selects the between-cluster variance method,since then all recorded video preprocessing process is complete.In terms of lane line edge detection,after rough detection by Hough transform,the point set is simulated to be a straight line by least square method.At the same time,innovative minimum characteristic root straight line detection is integrated to form an improved Hough transform.Traffic sign detection,with the improvement of road signs scanning methods of road sign recovery,and connecting with the lane mark recovery criteria to get a complete traffic signs to prepare for the next recognition,given at the same time to work with the lane line detection so choose lane line candidate marker detection method as a way to recognition of traffic signs in this paper.The improved method in this paper,through the verification of practical operation,can better deal with the difficulties encountered by lane detection in real life,and has a good performance in response to the processing time and anti-interference.