Key point matching method based on graph matching algorithm

The invention discloses a key point matching method based on the graph matching algorithm. The method comprises the following steps: (1) inputting and preprocessing an image; (2) carrying out key point feature extraction by using a CNN convolutional neural network; (3) obtaining key point features through the linear difference value; (4) constructing a key point graph by adopting a triangle planing algorithm; (5) constructing a global structure virtual key point; (6) carrying out information exchange among the key points by utilizing a GNN graph neural network; and (7) calculating the similarity of key points of the two graphs to obtain a matching result. According to the method, semantic information with richer key points can be obtained through a CNN convolutional neural network, then graph formed by to-be-matched key points of two images input by a triangulation algorithm is utilized, a global structure virtual key point is introduced, and the global structure virtual key point is input into a GNN graph ne