Research On Obstacle Recognition For Ground Travel System Of Split Rotor Aircraft

With the rapid growth of the total number of automobiles in China,urban traffic is facing increasingly severe pressure.Flying cars,as a new means of transportation,have gradually attracted people’s attention.The split rotor aircraft has two working modes of separation and carrying,which makes it have a huge advantage in solving urban traffic congestion,and has great value in medical treatment,rescue,urban travel,freight and other fields.As the "eyes" of the ground driving system,the perception part of the ground driving system of the split rotor aircraft is of great significance for its shuttle between cities.In this paper,the Lidar point cloud perception algorithm of the ground driving system is explored according to the working condition requirements of the ground driving system of the split rotorcraft and the characteristics of point cloud data.Firstly,according to the special structure of the ground driving system of the split rotor aircraft and the requirements of the actual operating conditions of the system,the perception system of the ground driving system is built.In order to ensure the safe operation of the ground driving system under the automatic driving mode,the Lidar is selected as the main sensor of the ground driving system perception part,which ensures the perception range and accuracy of the ground driving system by means of multi-sensor cooperation.Moreover,each lidar has its own coordinate system,so it is necessary to convert lidar data into its own coordinate system to obtain a complete set of point cloud data.The complete point cloud contains ground points and obstacle points,so the ground points need to be separated from the point cloud.In this paper,the random sampling consistency algorithm and through filtering are combined to ensure the accuracy of ground point segmentation.The remaining obstacle point cloud can divide the disordered points into each independent obstacle point cluster through clustering.Secondly,aiming at the serious problem of over-segmentation and under-segmentation in traditional Euclidean clustering methods,an adaptive Euclidean clustering algorithm combined with the idea of density clustering is proposed,which can effectively reduce the problem of under-segmentation and over-segmentation in traditional Euclidean clustering.Finally,due to the existence of common obstacles and ground sag obstacles in the clustering of partitioned obstacle points,and the ground sag obstacles cannot directly show their avoidance range through the bounding box,an algorithm for estimating the range of ground sag obstacles is proposed to solve this problem.The algorithm needs to distinguish ground sag obstacles from obstacles,and then estimate the range of ground sag obstacles.Experiments show that the algorithm can estimate the range of common ground depression obstacles and achieve the expected effect of the algorithm.