Real-time Detection And Tracking Of Mine Road Boundary Lines

Due to the special nature of the mining environment,the application of unmanned mines needs to be implemented.Automated driving mining trucks not only free the mining truck drivers,but also ensure the efficiency and stability of the work.Environmental awareness technology is the focus and difficulty of implementing autonomous driving.Road boundary detection is an important part of the autonomous driving perception system,which provides a safe driving range for autonomous driving and guarantees the safety of vehicles in the mine.Since mining roads have no shoulders,and the boundaries of mining roads are very blurred,it is an important research topic to distinguish between exerciseable areas within road boundaries and non-driving areas outside the boundaries.This topic collects point cloud data by installing a Velodyne HDL-32E 3D lidar sensor in front of a mining truck.First,study the effective segmentation algorithm of mine roads to obtain the elevated point information excluding the ground point cloud.Then,study the effective extraction and tracking of road boundary points on elevated points.The beam model is divided on the elevated points,the road boundary candidate points are extracted in each beam area,and the road boundary points are filtered out using the candidate points.Finally,the road boundary points are tracked and the curve is fitted to the extracted road boundary points,and the final result is fed back to the unmanned mining truck.The main contributions of this thesis are as follows:(1)According to the characteristics of uneven and multi-slope of mine roads,this thesis proposes a Double Meshing algorithm to effectively detect the surface of mine roads.(2)Because there are no clear road boundary points on the mine road,this paper designs a method to effectively detect road boundary points among multiple road boundary candidate points,which can effectively filter false points outside the road.(3)This thesis establishes a dataset for mines.The data of the Lidar is collected on different types of roads in the mine,and the road boundary points and targets are marked in the dataset.(4)This thesis proposes a road boundary point tracking method based on Kalman filtering,which can improve the stability and accuracy of the detection results.Experiments with a large amount of data verify the performance of the proposed method,and the experimental results prove the accuracy and real-time performance of the method.In most literatures,the research objects of perceptual technology are urban scenes or rural scenes.The perceptual technology algorithms proposed in this thesis are aimed at mine scenes,and solve the problem of complex and difficult to detect mine roads.This thesis implements the detection and tracking of the mine ground and the boundaries of mine roads.Road boundary tracking can predict and update the boundary point information in real time to prevent false detections and missed detections.Experimental verification data of mine roads show the accuracy and robustness of the algorithm.The algorithm is not limited to roads in mine scenes,but can also be applied to unstructured roads.