Robust homography optimization in soccer scenes

Abstract

Soccer content-based analysis is still a challenging process because of lack of information about the soccer scenes. An accurate homography is one of the most important information that is used to map the frame coordinate to the world coordinate, and vice versa. In this paper, a novel approach is presented to extract the frame line mask and tune the initial homography. The proposed method utilizes lines and arcs in the frame to find the best solution. The proposed method can also detect the invalid solutions. The obtained results are compared to the state-of-the-art method that uses the Levenberg-Marquardt technique in its optimization stage. The experimental results show that the proposed method is robust enough against strong shadows, lighting conditions, noise, and motion blurring.