Roadside laser radar target detection method and device

Abstract

The invention provides a roadside laser radar target detection method, which comprises the steps: selecting multiple frames of background point cloud data in different time periods from data acquiredby a roadside laser radar as background data, performing rasterization processing on the background data, performing statistics on grid features, and performing calculation to obtain grid average statistical features as a background grid statistical table; for the actually measured original point cloud data, carrying out rasterization processing with the same grid size as the background data, performing statistics to obtain grid statistical features corresponding to the original point cloud data, and carrying out background filtering in combination with the background grid statistical table to obtain non-background point cloud data; and inputting the non-background point cloud data into a constructed multi-scale voxel three-dimensional detection network, and outputting a detection resulttensor of the target, the result tensor comprising category information and bounding box information of the target. According to the method, a large number of invalid points are filtered through background filtering, the training and reasoning time of the network is remarkably shortened, interference of a large number of background points is avoided, and the precision of a detection result is improved.

Description

technical field [0001] The invention relates to the technical field of machine vision and intelligent driving vehicles, in particular to a roadside laser radar target detection method and device. Background technique [0002] Existing target detection technologies for lidar point clouds are mainly divided into traditional machine learning target detection methods and deep learning-based target detection methods. [0003] The target detection method based on traditional machine learning is mainly divided into four steps: 1) Perform ground segmentation or background filtering on the original point cloud to filter out a large number of background points or ground points. 2) Multiple clustering methods are used to cluster the filtered point cloud to obtain clusters formed by point clouds belonging to the same target. 3) Manually extract features from the obtained clusters. Common features include features such as density, height difference, normal vectors, and statistical hist...

AI Technical Summary

Problems solved by technology

However, the deep learning method requires a large amount of labeled data as input, and at the same time relies on the diversity of data to avoid over-fitting, the amount of calculation is large, and the real-time performance is difficult to guarantee.

In the roadside environment, the lidar is installed on a fixed roadside base station, and it is impossible to obtain a large amount of diverse point cloud data. The terrain background part of the obtained point cloud data is often highly consistent, resulting in overfitting of the trained model.

Images