Method for detecting retaining wall and obstacle behind vehicle in unloading region of mining region

Abstract

The invention discloses a method for detecting a retaining wall and an obstacle behind a vehicle in an unloading region of a mining region. The method comprises the following steps of: carrying out noise point removal based on a connected domain on a three-dimensional point cloud of an environment behind the vehicle, which is acquired by a laser radar sensor; carrying out height interval division on the three-dimensional point cloud after the noise points are removed, and distinguishing the condition that only the retaining wall point cloud exists based on the dispersion of the points in the height direction; based on the height difference and the slope angle between the maximum point and the minimum point of the three-dimensional point cloud in the x direction, judging whether the three-dimensional point cloud is ground point cloud or ground and retaining wall point cloud; further distinguishing the ground point cloud and the retaining wall point cloud based on the change of the slope aiming at the condition of adding the retaining wall point cloud on the ground; and for the ground point clouds obtained in the step 3 and the step 4, carrying out polar diagram-based ground re-filtering to further distinguish pure ground or obstacles. According to the method disclosed by the invention, the obstacle detection accuracy degree can be greatly improved, and the missing detection and false detection conditions are reduced.

Description

Technical field[0001]The present invention relates to the field of intelligent transportation, and in particular, to a method of detecting a retaining wall and an obstacle in the rear of the mine unloading area.Background technique[0002]With the comprehensive development of new infrastructure, combined with key technologies such as big data, 5G and artificial intelligence, the mining also ushered in a comprehensive intelligent transformation and upgrading opportunity. As a key ring of intelligence, a variety of sensors are configured to achieve accurate knowledge of mine environments, and provide important information for behavioral decision-making to ensure safe operation of mine cars. Among them, the unloading area is one of the most important regions of the mining area. During the reversing unloading process in the region, the mine cars not only need to accurately perceive the ground obstacle to avoid avoidance, but also need to determine the mine based on the end marker-retainin...

AI Technical Summary

Problems solved by technology

In addition, the time and economic cost of deep learning algorithms in data labeling and model training are relatively high. Relatively speaking, it is very uneconomical to apply in scenarios such as mine unloading areas.

Common rule algorithms for point cloud obstacle detection are grid-based and pole-map-based. The grid-based method often distinguishes possible obstacles from ground points through a simple fixed height difference threshold, but unloading If the ground height of the area is inconsistent, it is easy to have false detection and missed detection. However, the method based on the polar figure relies on the relationship between the fitted distance and the expected height of the ground, and distinguishes it through the obtained expected ground height, which cannot be used for floating noise. exclude

In addition, the shape of the point cloud behind the mine car in the unloading area is also quite diverse, including only the retaining wall, only the ground, and the retaining wall plus the ground. The selection of ground candidate points and straight line fitting using this method will cause the retaining wall to be mistaken for the ground and be filtered

Images