Research On Road Marking Asset Inspection Based On Hierarchical Image Segmentation Strategy And Dynamic Homography Matrix

In recent years,with the rapid development of China’s highway network,people’s travel has become more convenient,but the integrity of road assets,especially road marking assets,has become worse and worse due to factors such as vehicle wear and severe weather.The quality and integrity of road markings directly have a significant impact on driving safety.In order to ensure traffic safety,road managers must check the road marking assets,that is,regularly check the damage condition of the road markings in accordance with certain standards.The damage detection of road marking assets is the most important part of the largescale inspection research of road marking assets.At present,many domestic and foreign scholars have done a lot of research on road marking asset damage detection.However,due to the huge number of road marking assets,the complex texture of the damaged area,and the limitations of camera shooting,the accuracy of the current road marking asset damage detection methods is generally low.Aiming at the low inefficiency of the road marking asset detection method in the manual field and the low accuracy of traditional road marking damage recognition,this study aims to improve the efficiency and accuracy of the marked asset damage recognition,and the following research work has been carried out:(1)The traditional inspection of road marking assets usually uses manual on-site inspection.In addition,technical methods such as laser scanning and computer vision are also used for inspection.In order to achieve large-scale,rapid and accurate detection of road marking asset damage,this study proposes a road marking asset inspection method based on hierarchical image segmentation strategy,and conducts research and implementation.(2)In order to achieve rapid detection and recognition of road marking damage conditions,this study uses the Faster R-CNN neural network model to detect the marks,and selects the improved U-Net network model to segment the detected target markings.and finally use the image segmentation method to perform damage segmentation on the detected target mark.(3)Because the front-mounted camera on the vehicle will cause the geometric relationship in the image to be distorted and can only effectively capture the detailed texture information of the road markings closest to the camera,in order to reduce the impact of this problem,this study proposes a target marking damage segmentation image integration method,which uses KCF tracking algorithm and dynamic homography matrix to integrate the marking damage information in multiple images,Finally,complete the estimation of the mark damage ratio based on the integrated image.(4)This study uses the collected highway video data for experimental testing.The test results show that this method can complete the inspection of the damage condition of the road marking assets,and the effect is in line with expectations.