Research on Intelligent Algorithm for Regional Detection and Statistics of Sports Videos
Xiao Sheng Cheng, You Li · 2021
Combining with the training needs of athletes, the region detection and characteristic statistics and analysis of video images are performed, and the whole idea of designing automatic key posture extraction is put forward in this paper. Aiming at the video of weightlifting training, an approach of target detection based on inter-frame difference accumulation is proposed. Firstly, the candidate foreground area is extracted from the shooter as an example to provide effective preprocessing for the subsequent motion statistical analysis, to reduce background interference and improve the running speed. Then, a weighted histogram based on amplitude intensity is defined to describe the degree of direction disorder of optical flow characteristics in the region. The entropy of the histogram is calculated and applied to motion video analysis. Finally, the statistic characteristics of the optical flow tracking results are analyzed, and the close relationship between the extreme point of the channel strength and the key attitude is proposed. The experimental results show that the intelligent algorithm has high accuracy and real-time, which also effectively solves the problem of frame similarity.