Behavior recognition method based on graph convolution and Transform composite neural network

The invention discloses a behavior recognition method based on graph convolution and a Transform composite neural network. The behavior recognition method comprises the following steps: performing human body posture estimation on a video stream through an open source human body posture estimation framework OpenPose to obtain key node coordinate data of a human body part of each frame in a video; the graph processing module processes the key node coordinate data of the human body part into a graph data structure to obtain a human body skeleton sequence feature matrix as the input of a graph convolution model; modeling features in spatial dimensions through a graph convolution model; the input end adaptation module is used for carrying out the input end adaptation of the Transform module; the method comprises the following steps: modeling features in a time dimension through a Transform model; outputting a prediction result of the model through the classification head; and iteratively training and optimizing to