Sports video feature mark detection and scene recognition technology based on convolutional neural network

This paper draws on the convolutional neural network framework widely used in the field of image recognition. Aiming at the specific characteristics of sports images, this paper studies sports image target detection and scene recognition methods based on convolutional neural networks, and detects and detects athletes in sports videos. Tracking is conducive to higher-level analysis of the video, such as automatic game commentary, important event detection and tactical analysis. The application of image-based data mining technology can extract information such as the athlete's movement trajectory from the game video, which can help coaches analyze the player's behavior, study the opponent's strategies and weaknesses, and then use the Bootstrapping algorithm to train the convolutional neural network classification Device. For the input detection image frame, multiple candidate athlete positions are detected through the convolutional neural network, and then the candidate athlete positions are merged to determine the final athlete position. Experiments are carried out on some sports game videos. Compared with the AdaBoost algorithm, the scheme in this paper has obtained superior performance in both detection rate and false alarm rate, and the detection speed is faster.