The object segmentation technology research on tennis match video based on particle filter and a priori probability modelAuthor(s): Bo Zhang, Weigang Zou
Segment complete video results for a moving object can reduce the search amount of moving target and have great help for effectively enhance the tracking speed and precision. The purpose of video segmentation is to extract a moving target in video sequences from the background and to achieve segmentation of the foreground and background. The traditional method for extracting is the pixel value in the background is in accordance with the Gaussianmodel and conduct image segmentation according to 3ó rule. Although the rules can preferably extract the background, there will still be the case where foreground can be divided as the background by mistake. Therefore, this paper uses particle filter and a priori probability model to predict the moving target in the next frame, and then obtain a adaptive segmentation threshold to achieve video object segmentation according to the prediction value. This algorithm reduces the ratio that foreground points are divided into background points by mistake in the segmentation results, and it has improved for image segmentation compared with 3ó rule. Through the analysis and implementation of the algorithms, it provides a theoretical basis for the tennis tournament video object segmentation and better analyticalmethods for the technique improvement of all kinds of sports.