A Convolutional Neural Network-Based Segmentation Method for Soccer Players’ Interceptions

Ball interception requires athletes to use reasonable parts of the body to take over or destroy the opponent’s possession of the ball within the scope of the rules. If the player’s body posture is shifted by the center of gravity or the degree of knee bending during the interception, the accuracy of the interception may be reduced due to the small change in the magnitude. To this end, a segmentation method based on convolutional neural network is proposed for soccer player’s interception movement. The human body posture of the soccer player is estimated by the part affinity fields (PAFs), and the skeleton of the human body posture is extracted from the image of the soccer player’s interception action. Based on the skeleton of the posture, the attention mechanism is introduced into the convolutional neural network to construct the spatial-temporal map of the human body skeleton information; centered on the soccer player’s hip node, the spatial-temporal map of the human skeleton information is integrated by the human body-centered projection method, and the subtle changes in the body posture of the soccer player’s interception action are identified, and the key information features of the interception action are extracted; the extracted features are input into a convolutional neural network, and a weight adjustment unit is introduced to adaptively adjust the multilayered convolutional operation and the weights, so as to train the segmentation model of the soccer player’s interception action, and then realize the segmentation of the player’s interception action based on the completed training model. Through experimental verification, this method can extract the key human skeleton map of soccer players, and realize good feature extraction through the center projection of skeleton information spatiotemporal map, and through adaptive adjustment of weights, it can accurately capture the subtle differences in the interception action, and finally realize more accurate segmentation of soccer players’ interception actions, which helps to understand and analyze the technical actions of soccer sports in depth.