Obstacle detection method and system

Abstract

The invention discloses an obstacle detection method and system. The obstacle detection method comprises the following steps: acquiring laser radar point cloud data, wherein the laser radar point cloud data is detected by a single-line laser radar installed on a train; clustering the laser radar point cloud data according to the sequence number of the radar points in combination with a set maximumsearch range to obtain a plurality of clustering sets; extracting feature parameters of a target corresponding to each clustering set of the current frame according to the clustering sets, wherein the characteristic parameters comprise the number of radar points, an average azimuth angle, an average distance, a centroid position, average reflection intensity and reflection intensity variance; matching all the targets of the current frame according to the feature parameters of the targets corresponding to each clustering set of the previous frame; and when the matching succeeds, determining the target of the current frame as an obstacle target. According to the invention, the obstacle detection accuracy can be improved.

Description

technical field [0001] The invention relates to the field of track obstacle detection, in particular to an obstacle detection method and system. Background technique [0002] During normal driving on urban rails, it is necessary to detect obstacle targets in real time. At present, lidar is usually used to perceive medium-distance environmental objects and provide accurate distance information of environmental objects. The traditional lidar target detection mainly relies on the classic clustering algorithm. First, the clustering radius M is determined, and then a random seed point p is selected in the lidar point cloud image. i , calculate p i and all other radar points p i ,p i+1 …p n The distance between other radar points and p i Points whose distance is less than the clustering radius are added to the set V 1 {p i …p m}, and then in the set V 1 Select the next seed point in the point cloud image, which is not in V 1 The points in the distance are calculated, if...

AI Technical Summary

Problems solved by technology

[0003] In the prior art, due to the high speed of the vehicle, the background cannot be modeled. Therefore, it is not easy to distinguish the obstacle from the background by using the classical clustering algorithm to detect obstacles, so the detection accuracy is not high.

Images