Laser radar point cloud reflection intensity completion method and system

Abstract

The invention discloses a laser radar point cloud reflection intensity completion method and system, and the method comprises the steps: obtaining a gray image and an original point cloud of the sameroad surface through employing a calibrated vehicle-mounted camera and a laser radar; extracting edge information of the grayscale image by using a preset edge extraction strategy to obtain an edge image of the grayscale image; preprocessing the original point cloud to obtain an original point cloud reflection intensity projection image and an interpolation completion point cloud reflection intensity projection image; and inputting the grayscale image, the edge image of the grayscale image, the original point cloud reflection intensity projection image and the interpolation completion point cloud reflection intensity projection image into a pre-trained point cloud reflection intensity completion model, and outputting a completion point cloud reflection intensity projection image. Accordingto the point cloud reflection intensity completion method provided by the invention, the potential association between the point cloud and the image can be fully utilized, so that the reflection intensity image of the laser radar point cloud can be effectively and accurately completed.

Description

technical field [0001] The invention belongs to the technical field of intelligent driving, and in particular relates to a method and system for complementing reflection intensity of laser radar point clouds. Background technique [0002] Intelligent driving vehicles need to sense and measure the surrounding environment through lidar to ensure safe driving, including lane line detection, object detection and tracking, etc. However, the number of scanning lines of the lidar limits the density of the recorded point cloud, and the sparse point cloud is difficult to use for high-precision calculations, so it is necessary to complete the sparse point cloud into a dense point cloud. The main difficulties of point cloud completion include: 1) the original point cloud is too sparse; 2) outdoor scenes lack annotations for neural network supervised learning. [0003] The existing methods are mainly aimed at the completion of point cloud depth information, but the completion of point ...

AI Technical Summary

Problems solved by technology

[0003] The existing methods are mainly aimed at the completion of point cloud depth information, but the completion of point cloud reflection intensity has some difficulties that depth estimation does not have: 1) The reflection intensity is related to the material of the object, and has little to do with the geometric information. The reflection intensity of different parts may be different; 2) The accuracy of the graphic segmentation label is not enough, and it is difficult to directly use it as a supervision label

Images