Camera’s Auto-calibration Algorithm in Broadcast Soccer Video

Xunfang LIU

Hunan City University, Yiyang, Hunan, 413000, China

Abstract: In the traditional soccer video, it is difficult for the camera calibration technology to calibrate and provides the object’s three-dimensional height information and other issues, so this paper puts forward camera’s auto-calibration algorithm in soccer video. First of all, the two-dimensional model and threedimensional model of the court are defined and the improved the detection algorithm of court lines obtains the feature points in the court plane to solve camera’s isomorphic matrix, and then the solved isomorphic matrix is used to calibrate the goal and obtain the two feature points in the goal plan. Finally, the optimal internal and external parameters of camera are got from the method based on k-means clustering and Hough-like search. The court line detection and the detection experiment of the 3D – 2D camera are conducted, and the simulation experiments show that the improved court line detection algorithm can effectively improve the site regional classification in soccer video (> 96%), and the accuracy of court line detection (> 92%).

Keywords: Three-dimensional Reconstruction; Mapping; Competition Video; Edge Pixels