A method for identifying obstacles in complex road conditions for intelligent connected vehicles

Abstract

The present invention discloses a method for identifying obstacles in complex road conditions for intelligent connected vehicles, falling within the fields of intelligent driving and environmental perception. The invention uses a multi-row, vehicle-mounted lidar to scan the environment and acquire lidar point cloud data. Based on this point cloud data, ground points are distinguished from obstacle points, the latter being subjected to cluster analysis. Using the clustering results, the projected volume dimensions and point cloud density are calculated to derive the physical size information of objects. Subsequently, their dynamic properties are calculated to determine the object's state of motion. Simultaneously, each obstacle is assigned a label category, thus completing the identification of obstacles under complex road conditions.This invention effectively avoids the misclassification of small obstacles caused by conventional fixed-height thresholds. It overcomes the problems of under- or over-segmentation that arise from fixed parameters when point clouds are sparse at a distance or dense at close range. Furthermore, it avoids the instability of static feature classification of individual images, thus significantly improving driving safety in complex urban traffic environments.