End-to-end camera and laser radar external parameter calibration method based on corresponding mask

Abstract

The invention discloses an end-to-end camera and laser radar external parameter calibration method based on a corresponding mask. The method comprises the following steps: obtaining point cloud and image sequence data; extracting angular points of a checkerboard in the image, and estimating a mask of the checkerboard in combination with intrinsic parameters of the checkerboard so as to generate Euclidean distance transformation of the checkerboard mask; removing ground points and background points from the point cloud data, performing point cloud segmentation, and extracting checkerboard pointcloud; constructing an energy function of backward projection on a distance transformation field for checkerboard 2D mask distance transformation and checkerboard point cloud; and obtaining an external parameter initial value meeting the requirement of projecting the 3D point cloud to the image plane area through a genetic algorithm, and further optimizing the external parameter initial value through an LM method. According to the invention, the corresponding relation between the mask in the image and the point cluster in the laser point cloud is provided, the whole corresponding relation isclear, but details are fuzzy, the anti-noise capability and robustness of the method are greatly improved, the external parameter calibration process is simplified, and very high precision is achieved.

Description

technical field [0001] The invention relates to a method for calibrating external parameters of a camera and a laser radar, and belongs to the field of multi-sensor fusion perception. In particular, an end-to-end camera and lidar extrinsic calibration method based on corresponding masks. Background technique [0002] Camera sensors have the advantages of low cost, rich texture semantic information, and high frame rate, but they are easily limited by lighting conditions and it is difficult to recover accurate geometric information. LiDAR sensors have objective geometric measurement accuracy and are basically not limited by lighting conditions, but their spatial resolution is low, semantic information is difficult to extract, and the frame rate is low. Combining the two sensors with complementary characteristics has attracted a lot of research attention in the fields of 3D reconstruction, augmented reality and virtual reality, unmanned 3D target detection, multi-level high-pr...

AI Technical Summary

Problems solved by technology

[0005] Closer to the present invention is the method of 3D-2D point pairs based on the checkerboard calibration board, but this method uses the checkerboard to generate a deterministic correspondence. When the environment is complex or the noise becomes larger, it may not be able to extract the correct Corresponding point relationship can not even extract the checkerboard point cloud

A method close to the present invention also includes external reference calibration of natural scenes based on contour registration, but the natural scenes that this method relies on have great uncertainty, and the accuracy of extraction of natural scene images and laser point cloud contours is seriously affected The result of the final external parameter calibration, which will bring uncertainty to the calibration

Images