Playfield Detection Algorithm Based on AGMM in Sports Video

Because under the condition of multiple factors, the sports video content analysis of the traditional playfield detection algorithm can cause the non-homogeneity of playfield color, the playfield detection algorithm based on AGMM is proposed in this paper. Firstly, extract the image at random from the sports video and make automatic analysis of the image color and find the approximate values of color distribution. Then, use GMM fitting color distribution, the playfield test results and incremental expectation maximum algorithm to update model parameters so as to get more accurate parameter estimation which will be used to make classification for pixels of playfield and non-playfield in subsequent image. Finally, the test of algorithm performance is carried on and the proposed algorithm in this paper is used to classify sports scene, experiments. The experiment shows that compared with the traditional playfield detection algorithm, a playfield detection algorithm based on AGMM model is greatly improved on the recall ratio and accuracy.