The proposed method is based on deep learning and graph neural network, which analyzes the player's action images of volleyball teaching and training matches based on deep learning and graph neural network, annotates the technical action of tennis player's serve, extracts the contour of tennis player's serve action based on deep learning and graph neural network, extracts the features of player's serve action, constructs the player action image fusion model, designs the player action image data fusion process, and constructs the player action image fusion model, designs the player action image data fusion process, and constructs the player action image fusion model. Model, design player action image data fusion process, realize player action image fusion, test results show that the fusion accuracy coefficient of this method is much higher than the level of existing methods.