Abstract: This paper models each of the video frames as an adaptive Gaussian Mixture Model(GMM), and uses this model to extract the background image and to segment the moving regions. The color information is used to detect playfield based on the background image. In order to enhance the accuracy of the detected player regions, texture similarity measure is employed to remove the shadows, and a morphologic method is adopted to fill the gaps inside the player regions, noise is removed based on the playfield information. This paper improves the CamShift algorighm, which is used to track the players automatically.

Key words: adaptive GMM, playfield detection, player detection, player tracking, CamShift algorithm

摘要： 利用自适应高斯混合模型对视频图像进行建模，从图像序列中获取背景图像并提取运动区域，利用像素的颜色信息从背景图像中提取绿色球场。为提高运动员检测的准确度，利用纹理相似性度量方法消除运动区域中的阴影，用形态学方法消除区域内的裂缝，根据球场信息去除球场外的噪声。改进了CamShift算法，并应用该算法对运动员进行跟踪。

关键词: 自适应高斯混合模型, 球场检测, 运动员检测, 运动员跟踪, CamShift算法

CLC Number: 

• TP391