Abstract:Since there are differences between students' understanding level, it is easy to produce wrong actions during the process of sports teaching. In order to improve the video detection effect, the video fidelity and compression rate, a remote video detection algorithm for sports wrong actions under wireless network is proposed. By extracting key frames in video clips, a real-time video transmission model is established based on wireless network, which transmits key frame extraction results real time with the support of real-time transmission protocol and forward error correction mechanism. The deterministic constrained nonlinear optimization method is used to estimate joint and overall motion parameters. Based on each parameter, the aggregated channel feature method is used to identify the target in the video sequence, and the support vector machine is used as a classifier of sports actions to recognize wrong actions. Then, the multi-scale space of moving images is constructed, the multi-layer box filter is used to simulate Gaussian convolution, and the simulation results are corresponded to the target areas in the remote video of sports wrong actions, thereby realizing the remote video detection of sports wrong actions. The experimental results show that the fidelity rate of the proposed algorithm is always above 90.3%, the compression rate is more than 80%, and only one key point is omitted. This proves that the video fidelity rate and compression rate of this method are high, and the detected key points are the same as the actual key points, which verifies the effectiveness of the proposed algorithm.

Remote video detection algorithm of sports wrong actions under wireless network