Automatic Extraction Of Urban Lane Line Based On Measurable Real Scene

As the main scene of the future car,intelligent driving requires the vehicle to automatically make a series of actions,such as braking,steering,acceleration and so on according to the actual road conditions,all of which are inseparable from the correct perception and determination of the surrounding environment of the lane boundary.Lane detection is one of the hotspots and key technologies in the advanced driving assistance system(ADAS)and intelligent driving research.The fast,stable and reliable Lane extraction algorithm is also a frontier problem to be solved in machine vision.In this paper,according to the data collected by Mobile Mapping System(MMS)under the condition of complex urban road,the extraction method of lane line is studied and practiced.The paper is mainly divided into three parts: lane line image preprocessing,lane detection and extraction,experiment and analysis.The purpose of image preprocessing is to eliminate the interference factors of lane line recognition(vehicle,tree,shadow,etc.)as well as to improve the data quality of the image to be identified,including the setting of ROI in the region of interest,image grayscale,image smoothing and image enhancement,and so on.Lane detection and extraction,including image segmentation and Hough transform to extract lane lines,is also the core content of this paper.The details include: exploring the way of separating the lane line from the background by combining the threshold segmentation and the morphology,and using the constraints of multiple frames to screen the lane lines.Finally,the lane line is identified by the improved Hough transform algorithm.In the end,16 groups of real scene data under the condition of complex urban road are used to test the accuracy and speed of the automatic extraction algorithm of the above lane line in two aspects,and the availability of the algorithm is tested.The experimental results show that the lane extraction algorithm in this paper can basically adapt to scenes of complex urban environment,and has good accuracy and robustness.The correct recognition rate of lane extraction is up to 80% under the specific city scene,but there are still serious leakage and false detection in the scenes with serious occlusion and too complex situation,and the correct recognition rate is low.Besides,the algorithm needs further improvement and improvement in real-time.