The Computer Vision System for Tennis Video Abstraction and Player Movement Trajectory Analysis
| Autor: | CIOU, CI-DUAN, 邱啟端 |
|---|---|
| Rok vydání: | 2017 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 105 Sports tournament programs, especially live broadcasting, is a popular entertainment for television audiences. However, these programs are always lengthy, in which the transition scenes become rather time-consuming for the viewer. In addition to players detecting and tracking are not simple in sports videos. Because the movement of camera, the size of target or the texture of the tennis court for different reasons, so the player is not easy to be detected successfully. Therefore, the study will be used digital image processing technology for tennis broadcast video to splice abstract. And then the study is how to implement the players detecting and tracking for the splice abstract videos. First, through the histogram to calculate the RGB value and then according to its relevance comparison to be segmented tennis video from frame to frame. And then through the method “perspective transformation”, “edge detection” and “Hough transform” for determining whether the field by characteristics of the tennis court. Finally, the “Motion History Image(MHI)” method is used to detect the players. And the image pixel value from frame to frame that subtract to each other and take the absolute value. Then according to the size of the area to determine the target and detecting and tracking it. Finally, the experimental results that the average recognition rate of 84%, the average video removal rate of 88.72%, the average player detection of 89.71%. To prove that method proposed in this study has been successfully achieved abstract splicing of video and player detection and tracking. And also be evaluated for the effectiveness of this method. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|