Accessible Area Detection Based On Sparse Point Clouds In Outdoor Environment

Environmental perception,as the "eyes" of mobile robots,is the basis of its navigation,positioning and planning control.Therefore,real-time and effective passable area detection plays a vital role.Multi-line lidar has an irreplaceable position in the field of environmental perception by virtue of its excellent geometric characteristics,accurate measurement characteristics and stable working characteristics.Aiming at the problem of insufficient detection performance of existing methods in complex outdoor environments,based on the sparse point clouds data of VLP-16,this paper studies a passable area detection method for complex outdoor environments.It aims to overcome the influence of complex environmental factors and provide mobile robots with stable,accurate and fast environmental perception.The research content mainly includes the following aspects:(1)In order to break through the limitations of typical methods and improve the accuracy and quality of detection methods in outdoor environments,this paper proposes a road boundary detection method based on line feature extraction optimization.First,the local optimization extraction of feature points and feature lines is realized according to the constraints.Secondly,through the combination of K-Means clustering,horizontal and vertical integration and RANSAC screening,the straight line segment describing the real road boundary is quickly and accurately obtained.Finally,make full use of the excellent characteristics of the cubic B-spline interpolation model to draw a complete and delicate road boundary.Experiments prove that the algorithm has excellent accuracy and robustness in a variety of outdoor test environments.(2)Obstacle detection is a key task of environmental perception.This paper combines the grid method of point clouds data and geometric features to detect obstacles,which improves the calculation efficiency and avoids the loss of accuracy.The method introduces the dynamic threshold method to extract obstacle points in the process of radial feature analysis,which enhances the robustness and accuracy of the algorithm,and effectively reduces the occurrence of false detections by judging lateral features.Finally,taking the nearest neighbor obstacle points as constraints,the road boundaries are merged to form a passable area for the robot.Comparative experiments in multiple scenarios prove that the algorithm can overcome the influence of complex outdoor environments such as curves,ramps,multiple obstacles,and negative obstacles,and has good real-time,accuracy and robustness.(3)Information such as the category and outline of obstacles has higher application value and research significance.In this paper,vehicles and pedestrians are taken as the main obstacle models,and their classification tasks are studied through the neighborhood growth clustering of the grid.The method uses prior knowledge to construct a barrier grid map,and performs regional clustering and segmentation of barrier grids based on the neighborhood growth method.Then extract the rectangular outline according to the distribution characteristics of the cluster,and accurately mark the category,scale and location of the obstacle.Finally,combining the height information and geometric characteristics of the clustering,a simple classification of vehicles and pedestrians is realized.The experimental results verify the correctness and effectiveness of the algorithm.