The article describes the teaching methods used by physical education teachers in the classroom setting and explores how observational learning, contextual learning, and flipped-classroom teaching methods interact in university physical education classes. It develops a multidimensional model of the application of various teaching methods in the field of university physical education. The MonoLoco method is utilized to attain the fuzzy localization process, and the information fusion method is utilized to improve human body posture data. Further, the HRNet model is introduced to include de-redundancy design and multi-resolution supervision to develop a keypoint detection algorithm for human skeletons based on the DHRNet network. By combining the monocular human-positioning algorithms with the DHRNet model to detect skeletal keypoints, the integrated physical fitness evaluation model for students is generated. The model is used to assess the performance of physical education teaching methods from a multidisciplinary perspective. The experimental results of integrating several teaching methods showed that the experimental group had improved their comprehensive evaluation index, including 50 meters, 1000 meters, standing long jumps, and shot put. The mean values of the 50 meters time reduced by 0.222 s, 1000 meters increased by 0.274 points, standing long jumps increased by 0.295 m, and shot put increased by 0.386 m. The above results indicate that the effectiveness of the physical education teaching with multiple teaching methods is considerable and applicable in university-level physical education.
