Soccer Video Retrieval Technology Research And Realization

This dissertation using the soccer video from the television relay as the object of study, conducts an thorough investigation in correlative technology in soccer video retrieval and proposes a project of detecting shoot event in soccer video. The major content and contribution of the dissertation are as follows:(1) Shot division is the first step of video structurization, which is also premise of the following modules of analysis and processing. Combined with traditional division algorithms, an improved shot division algorithm applying to soccer video is proposed, based on the auto-adapted threshold.(2) Realizing shot classification with the k-average value cluster algorithm. Firstly the shot classification algorithm selects the key frame from the video shot which is obtained from the video shot division, to accomplish field segmentation, and then divides the video shots simply into non-field shots or field shots, using the main color ratio of field. On this basis, through analyzing the image edge information the algorithm divides the shots containing little grass color field into near shots or outside-field shots. In succession, based on the k-average value cluster algorithm, this algorithm divides the shots containing an abundance of field area into far shots or medium shots with synthesizing the image characteristics, which include the major color ratio of the image ,the number of players in the competition field, and the average spatial area occupied by the player and so on. Thus complete the shot classification in soccer video.(3) Detecting shoot event with the mode of multiple characteristics. In the soccer competition, shoot event usually occurs nearby the forbidden area, accompanied by the audience soaring cheers simultaneously. Therefore, by detecting the forbidden area in soccer competition and the audio information separately, the thesis proposes a detecting scheme, which taking the temporal sequence characteristic as the master line with combining the image space characteristic, the audio characteristic and the time characteristic.Summarizing the content above, on the developing platform of VC++ 6.0 and MATLAB 6.5. this paper realizes a detecting system related to shoot event in soccer video. The experiment indicates that the project of detecting shoot event in soccer video, proposed by this paper, acquires satisfying effects.