Research On Road Lane Detection Method Based On YOLOv3 Algorithm

As a key issue in the field of intelligent driving,road lane detection has been widely used in vehicle assisted driving systems,lane departure warning systems and vehicle anti-collision systems.To improve the speed and accuracy of lane detection,the research on the end-to-end intelligent lane detection method has become a new way in the field of lane detection.Therefore,the study of road lane detection methods based on deep learning is of great significance for improving road traffic safety.The Hough transform method is used to detect the road lane.In order to reduce the noise interference,the road lane image is preprocessed first.Secondly,edge detection is performed on the preprocessed image,and a region of interest is set.Finally,the Hough transform method is used to detect the straight lanes in the image,and the straight lines that conform to the geometric characteristics of the lane are marked.The experimental results show that the method has good detection effect on continuous lanes,and the detection effect of shorter lanes needs to be improved.To enhance the comprehensiveness of the lane detection algorithm and further improve the accuracy of lane detection,the other road lane detection method based on YOLOv3 algorithm is proposed.Firstly,the preprocessed image is divided into multiple grids,and the original K-means algorithm is replaced by the K-means++ clustering algorithm to determine the number of target priori box and the corresponding width and height values.Secondly,according to the clustering result,the network anchor parameter is optimized,so that the training parameters have strong pertinence in lane line detection.Finally,the lane line features are sent to the streamlined network,and the GPU is used for multi-scale training to obtain the optimal weight model.The lane line target in the image is detected,and the bounding box with the highest confidence is selected for marking.Using the image information in the Caltech Lanes database,the comparison results shows that the proposed method has higher accuracy and faster speed than other methods in road lane detection,and road curves can be detected at the same time.Aiming at the problem that the YOLOv3 algorithm needs to improve the detection effect of continuous lanes,a road lane line detection method combined with Hough transform is proposed to realize the comprehensive detection of lanes in images.