Research On Key Issues Of Video-Based Supporting System For Snooker Sport

Sports video analysis is studied for a long time. Some effective methods have been proposed and have been applied in some sports. Snooker is becoming a popular sport in China recently. There are several international popular tournaments every year, which attracted a lot of audience.In this thesis, we focus on some key issues of video-based supporting system for snooker, such as segmentation of playing surface, transformation between image and table coordinates, resetting and so on. The details are as follows:1. We propose an improved image segmentation method using color and edge. The former method is designed for snooker table covered with green baize and it is not robust to noise. The improved method is designed to extract the color of playing surface on the table automatically, and we filter the noise edge pixels by smoothing and iterated curve fitting. The experimental results show that the proposed method is robust to variation of illumination, color of playing surface, view of camera and balls on the table.2. We design and implement a plane-based camera calibration algorithm. We use the ball spots on the table and intersection points of the closed playing surface as calibration points. Furthermore, there usually are some deformations in the image. In order to resolve the problem, we divide all the calibration points into two groups, from which two coordinate transformation matrixes can be obtained. Calibration parameters can be achieved by resolving linear equations respectively.3. We propose a precise resetting method using image registration. Taking full consideration of the correlation and stability between any two consecutive frames, the balls are reset by the similarity between sub-images in two consecutive frames. The resetting precision is 1 pixel. Coordinate offset is fast computed by a two-step searching algorithm. The true location is marked using transparent color circles. All the above works make the resetting procedure precise, friendly and pleasing.4. We design and implement a demo platform. It can be used to show the result of image segmentation, location and tracking of balls, resetting and so on. Furthermore, it can be used for data analysis.