摘要:
随着足球视频语义内容分析和理解、自动视频索引、虚拟增强现实表示等研究的不断深入,摄像机定标技术扮演了越来越重要的角色.基于此,提出了一种基于球场模型的广播足球视频摄像机自动定标算法,该算法建立在所定义的球场模型基础之上,首先通过改进的球场线探测算法得到球场平面的特征点,并求解摄像机同形 (Homography)矩阵,然后,利用所求解的同形矩阵来定位球门,得到球门平面的两个特征点,最后根据基于K均值聚类和Hough-like搜索的方法来得到最优的摄像机内外参数.初步的实验结果表明,依据本文算法所求解的摄像机参数最优值相比初始值精确度更高,并且其在连续帧中也更加稳定.
Abstract:
Nowadays, camera calibration plays a crucial role in soccer video processing domain, such as semantic content analysis, video indexing, virtual augmented reality and so on. In this paper, we propose a method for automatic camera calibration of broadcast soccer video. This camera calibration algorithm, which is based on the pre-defined 2D and 3D playfield model, consists of three steps. First of all, the improved line mark detection is applied to detect the feature points in the playfield plane, and the homography matrices are also acquired. Secondly, another two feature points in the goalmouth plane are located according to the homography matrices. Finally, the optimal camera parameters are gained through K-means clustering and Hough-like search. The preliminary experiment demonstrates that the optimal camera parameters are more accurate and stable than the initial one.
下载:
