Research On Obstacle Detection Technology In Environmental Perception System Of Smart Car

The smart car is a product of the combination of the modern automobile industry and Internet electronic technology,and is of great significance to the future development of road traffic patterns.In order to further ensure the safety of people’s travel,improve the driving efficiency of cars and improve the comfort of driving,driverless car technology came into being.Environmental perception is an important foundation for the realization of intelligent car unmanned driving.The use of advanced sensor perception technology instead of human vision to detect obstacles in the driving road provides a basis for smart car decision-making control,thereby effectively improving the safety of car driving.This paper does systematic research on the obstacle detection link in smart car perception technology,and designs a clustering detection algorithm to effectively extract obstacle information.The thesis research is based on 16-line lidar detection technology,and the working principle,parameter structure and communication protocol of lidar are analyzed in detail.Building a platform for environment perception system by building ROS system and configuring its environment.Before the algorithm processes the lidar point cloud information,the characteristics of the point cloud data are analyzed,and the point cloud is preprocessed.In addition,based on the PCL point cloud library,an algorithm framework for obstacle detection is designed,and the point cloud data is sequentially down-sampled,ground segmented,and clustered.Among them,the application of the voxelized grid scheme completed the downsampling processing of the point cloud,simplifying the information volume of the point cloud data,according to the Ray Round algorithm,the ground point cloud was successfully separated,and the ground point cloud was excluded from the clustering detection of obstacles.Using the KD Tree data structure to reduce the dimensionality of point cloud data to improve the search efficiency of point clouds.The computational efficiency of class detection and the concept of boundary box selection are used to improve the results of obstacle cluster detection.Finally,the detection information is displayed in the Rviz visualization tool.In order to verify the feasibility of the obstacle detection method in this paper,the test was carried out by setting up a real vehicle experimental environment.The experimental results show that the method can segment the point cloud data from the ground better,and caneffectively detect obstacle information,and has certain accuracy and stability.