Small-scale Complex Environment Perception And Terrain Modeling For Patrol And Exploration Missions

The moon is rich in mineral resources,energy reserves and environmental resources,providing stable support for the sustainable development of human society,and its unique location in space is the cornerstone of human exploration of distant stars.and even the vast universe.Lunar exploration has always been an important strategic goal of the major spaceflight countries.At present,the mainstream means of lunar surface exploration is still to carry out sampling missions with unmanned patrol detectors,and in the process of lunar surface patrol and exploration,the perception of unknown lunar surface environment is the prerequisite and key technology for obstacle avoidance and path planning.In this paper,based on the month table environment,the study on the small-scale complex terrain environment perception technology and 3d scene modeling technology is carried out.The main research contents are as follows:Firstly,the paper analyzes the main obstacles in the operation environment of lunar surface patrol and exploration mission,such as rocks,meteorite craters and steep slopes.This chapter introduces the principle of data acquisition based on stereoscopic vision,the collection of data of point cloud and moon table,and the pre-processing operations such as filtering,de-noising and sampling,etc.,so as to provide good topographic point and cloud data of month table for subsequent obstacle detection.Then the terrain perception method based on 3d point cloud data is studied to identify the moon obstacles.The main process is as follows: the random sampling consistency algorithm(RANSAC)is used to extract the topographic datum plane,and the remaining data is obtained by the density based clustering algorithm(DBSCAN).The results show that the accuracy of base-level estimation is very important for obstacle detection.Combined with the lunar surface environment and the analysis of algorithm principle,it can be seen that the application of traditional RANSAC algorithm will affect the applicability of multiple scenes,and the high abundance obstacles will cause the deviation of plane estimation.In view of these two problems,adaptive threshold and morphological algorithm are improved to extract relatively accurate datum and improve the accuracy of obstacle detection.Next,this paper classifies and summarizes the clustering points,analyzes the types of obstacles,and judges the position,range,height and depth of obstacles.Using the plane model parametric equation and the simplified obstacle data of convex hull algorithm to reconstruct the scene not only improves the reconstruction efficiency but also preserves the terrain features,so as to guarantee the accuracy and real-time control of lunar rover’s detection and obstacle avoidance.Finally,the experimental platform of complex terrain perception and scene modeling on the small scale of patrol and exploration mission is built,and the obstacle perception and scene reconstruction experiments based on 3d point cloud are carried out by using simulated terrain data and Lunar terrain data.By comparing and analyzing the experimental results,the correctness and feasibility of the proposed algorithm can be verified.In small-scale complex terrain patrol and detection tasks,the method in this paper can accurately obtain the position coordinates of obstacles and quickly conduct scene modeling by using point cloud characteristics,which greatly saves time cost.