Monocular perception correction method, device and storage medium based on sparse point cloud

Abstract

The invention discloses a monocular perception correction method, device and storage medium based on sparse point cloud. The method includes: collecting original camera data of a monocular camera and original sparse point cloud data of a radar sensor; processing the original camera data , obtain the 3D detection results of multiple targets in the image plane, the 3D detection results include the target depth value and the 2D bounding box; obtain the transformation matrix; based on the transformation matrix, map the original sparse point cloud data to the corresponding position of the image plane, and obtain Point cloud projection depth map, set a point cloud frame for each two-dimensional bounding box in the point cloud projection depth map, the point cloud projection depth map includes multiple projection points corresponding to the original sparse point cloud data, each projection point Including point cloud depth values; based on the point cloud depth values ​​of the projected points contained in all point cloud frames, the target depth values ​​of multiple targets are corrected. The invention improves the accuracy of correcting the target depth value by designing the point cloud frame characteristics.

Description

technical field [0001] The invention relates to the field of automatic driving, in particular to a method, device and storage medium for monocular perception correction based on sparse point cloud. Background technique [0002] Intelligent perception is an important part of autonomous driving and the link between the vehicle and the environment. At present, the mainstream perception sensors include cameras, millimeter-wave radars, and lidars, etc., but multi-beam lidars are very expensive and not suitable for mass production, and the point clouds obtained by low-beam lidars and millimeter-wave radars are very sparse and not suitable for direct use. Do 3D obstacle perception. Compared with binocular cameras, monocular cameras are relatively cheap sensors and have unique advantages in obstacle detection and tracking. However, monocular cameras have limitations in depth information perception. [0003] In the prior art, the scheme of fusion of monocular camera data and sparse...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a monocular perception method based on sparse point cloud, which is used to solve the problems of high algorithm complexity and large target depth estimation deviation when correcting the depth information of the monocular camera in the prior art

Images