Research on Missing Value Compensation of Skiing Data Based on Neural Networks in Optical 3D Motion Capture

It is necessary to adopt an human action capture system to capture and analyze the movement posture of athletes. However, it cannot cover all experimental areas during the collection process as the camera's perspective, resulting in missing marker points in the collected data, which is inconvenient for later analysis of sports data. This research focuses on the missing value completion of skiing data based on the neural networks in optical 3D motion capture. Specifically, the Qualisys motion capture system is utilized to collect the skiing data of athlete by wearing 17 key points. In the case of a missing key point in skiing data, all the collected key point data is randomly masked with one key point, and part of it is used as the training set and the reset part as the validation set. Then, the fully connected neural network and ResNet are combined to train and compensate the collected data. Finally, the compensation results are verified by real data, which is of great significance for analyzing the sports posture.