A method and system for detecting ground obstacles based on laser radar

Abstract

The invention provides a method for detecting a ground obstacle based on a laser radar to solve the problem that in the prior art, obstacle and ground recognition precision is low and algorithm is complex in unmanned driving or obstacle recognition technology. The method comprises the following steps of S1, acquiring point cloud data in a preset scanning range; S2, confirming preset pavement pointcloud data in the acquired point cloud data through a preset recognition algorithm; S3, judging whether the point cloud on a straight line in a preset pavement point cloud normal direction is presetuniform point cloud or not; S4, if yes, determining that no ground obstacle exists in the current preset scanning range; and if not, determining that a ground obstacle exists in the current preset scanning range. According to the method, pavement obstacle information can be determined accurately and rapidly, calculation according to historical data is not needed, and analysis speed is high.

Description

technical field [0001] The invention relates to the technical field of laser radar identification, in particular to a system for detecting ground obstacles based on laser radar. Background technique [0002] Lidar obtains point cloud information of the surrounding environment through scanning and ranging, which has high precision, long distance, fast speed, and is not affected by light. It has broad application prospects in the field of unmanned vehicles and robots. [0003] At present, obstacle detection methods based on point cloud data are mostly used in detection methods based on high-precision UAN maps and detection methods based on machine learning training models. [0004] Among them, the detection method based on the high-precision environmental map is to obtain a high-precision environmental map by splicing the point cloud data collected offline first, and then compare the difference between the online point cloud data and the environmental map to obtain the dynami...

AI Technical Summary

Problems solved by technology

However, this method requires combined calculation of lidar and offline data, and is not suitable for environments with a large number of object movements. The calculation data volume is large and the processing is slow.

[0005] Among them, the detection method based on its learning and training model is to extract the key points and features of the point cloud through calculation, use a large number of training samples, obtain models of various obstacles through various learning methods, and predict and identify the types of obstacles through the model. However, a large number of high-quality samples are required for training, and it is impossible to detect obstacle types that do not appear in the training set

Images