As an important part of Chinese culture, national sports culture not only has profound cultural connotations but also has rich cultural value. This paper proposes a path that integrates national sports culture and martial arts into teaching, focusing on three key aspects: diversification, scientification, and digitization. The neural network operator Involution improves the recognition accuracy of the human pose estimation model after addressing network degradation issues with the deep residual network. We preprocess the collected pose images to extract human wushu pose features from both static and dynamic directions. The feature data is inputted to the encoder to extract martial arts poses. Simulation experiments verify the error correlation between two-dimensional martial arts movements and three-dimensional angles. In the backside horse stance punch movement, the 7th frame of the left knee and right knee, the 3rd frame of the left hip, and the 1st, 2nd, 5th, and 7th frames of the right hip exhibit significant errors, while the remaining keyframes fall within the acceptable range. Through empirical investigation, we analyze the teaching effect of integrating Wushu and national traditional culture. Following the experiment, we observed an average score difference of 8.91 points in the degree of sports participation between the experimental group and the control group, with a P value of less than 0.01. There is a significant difference between the two integrations. The integration of the teaching effect is good.