Camera-Lidar Extrinsic Calibration via Traffic Signs

To address the problem of how to carry out camera and 2D Lidar calibration in general natural environment, this paper proposes a method that can be calibrated based on various traffic signs on the road. Firstly, this method utilizes the features of the traffic signs extracted from the image plane. In order to obtain accurate traffic signs edge features, the guided filtering of images and Otsu algorithm are used to extract the traffic signs area. Then this method utilizes Sobel operator and Hough transform to extract the straight features, utilizes Canny operator and arc-support line segments to extract the curve features. Secondly, aiming at the problem of how to find the corresponding relationship between laser points and the points on the camera image when the scanning laser of 2D Lidar are not visible, the angle information collected by 2D Lidar and the linear fitting of the point cloud collected by laser is used to find the intersection point between the laser scanning plane and the edge of traffic signs. Finally, the projective matrix of the camera and Lidar is solved by using the Point-to-Line geometric constraints for the intersection points and the edge features of traffic signs, and this method utilizes the internal reference matrix of the camera and the projective matrix to obtain the rigid transformation matrix from the Lidar coordinate system to the camera coordinate system. The experimental results are analyzed and it is concluded that this method is more convenient to realize the external parameter calibration of Lidar and camera under the premise of maintaining a certain accuracy.