Motion recognition method and system based on graph comparison model

The invention provides an action recognition method and system based on a graph comparison model, and relates to the field of action recognition. Comprising the following steps: acquiring and preprocessing a contrast video and a to-be-classified video to obtain multiple frames of contrast images and multiple frames of to-be-classified images; inputting the plurality of frames of contrast images and the plurality of frames of images to be classified into an image contrast network, extracting human body joint skeleton points, and generating an image structure; performing single-graph feature extraction on the graph structure to obtain single-graph features; inputting the single image features into an image neural network formed by the relation of the joint skeleton points corresponding to the image structure, and carrying out cross-image feature extraction; aggregating distance features of all points of the cross-graph features, inputting the distance features into a multi-layer perceptron, and outputting a pre