Research On Lane Recognition And Departure Warning Based On Single-camera Vision

With the development of Intelligence,it has been extended to all aspects of the life.Car,as a tool closely linked to human's life,has entered the stage of intelligence.Intelligent vehicle technology can not only greatly improve the safety and reduce traffic accidents,but also release their own people and enhance comfort when driving,so it has been one direction of advance for the car industry.Visual navigation is a key technology of intelligent vehicle.In this paper,the study of lane recognition and lane departure warning with monocular vision is one research direction of intelligent vehicle technologies.As an important parameter to the control of the vehicle,the lane can be extracted from the road image.It is used to establish the lane model,which shows the relative relationship between the lane,the lane line,and the vehicle.Then,the lane line parameters can be extracted to realize lane recognition and lane departure warning.Visual sensor calibration is the basis of establishment of the lane departure warning model.Therefore,we discuss the monocular camera calibration in the first so that we have an idea of the translation of from image coordinate to world coordinate.The aim of image pre-processing is to extract lane edge.In the process of image pre-processing,gray is used to remove useless information,de-noising smoothing is used to remove noise.In the process of de-noising filtering,the median filter is selected to remove salt noise and pepper noisebecause it is simple and real time after comparing the advantages and disadvantages of various de-noising smoothing algorithms.Then,the image enhancement method is introduced and the lane line enhancement technique based on histogram equalization is put forward,in which the useful sign is outlined.And then,Otsu is selected to realize image binaryzation,which can distinguish lane road region from background region.After pre-processing,the next is lane detection.After comparing all kinds of detection methods,adaptive Canny edge detection algorithm is proposed.based onfit lane line and establish lane line model by improved the Hough transform.In the process of lane recognition,the influence of light and rainy weather are taken full into consideration,the improved Hough transform by using the method of polar Angle pole is come up to overcome it.According to the characteristics of the highway road,lane line recognition algorithm is proposed based on the road model.The next step is to establish the deviation from the early warning model and extract the vehicle centerline and lane line deviation angle parameters.After calculation,we set an appropriate threshold which is the key to work the warning system depending on the relationship between the angle and the threshold,lane departure warning come true at last.The final results show that the proposed lane recognition algorithm improves the robustness and timeliness of the system,and the effect of the lane departure warning model is also in good performance.