3D Human Posture Capture Algorithm Based on Artificial Intelligence

3D human pose estimation is a research hotspot. The research on this technology can promote the development of other advanced artificial intelligence technologies based on computer vision, such as sports teaching such as music and dance, 3D stereoscopic movies, and human motion pattern recognition, all of which cannot do without the development of this technology. Currently, there are still many researchers in this research field who are researching and developing it. On the basis of current research by researchers, this article explored a convolutional neural network model with semi supervised learning to obtain 3D human body estimation including spatial positions. The proposed method can estimate the corresponding 3D pose skeleton model from the daily motion videos. After experimental testing, the results can confirm that the method proposed in this article is effective and can more accurately estimate the 3D pose of the corresponding video frame (the FPS of the model in this article was 1130 frames/second, which was higher than the model based on the LSTM (Long Short Term Memory) algorithm by 950 frames/second, and the MPJPE value was 43.7mm, indicating that the error between the predicted joint position and the actual joint position based on the algorithm model in this article is minimal).