Automatic Analysis of Basketball Shooting Based on Machine Learning

Abstract

In order to help basketball players to carry out training more efficiently and reasonably, and obtain the technical specifications of their shootings timely in training, this paper constructs a shooting detection algorithm to assist basketball players in training. Firstly, Hough transform algorithm is applied to the shooting video to detect the trajectory of the basketball in the shooting. By analyzing the basketball trajectory, it is found that the best shooting angle should be 49° to 51°. Then, the basic principles of HOG feature and SVM are studied, and the HOG feature of basketball hit and miss is extracted to train the SVM classifier. Finally, the training model was tested, and the classification accuracy of basketball hit or miss was 91.75%. This study will help the basketball players in their training and also could be applied to other ball games.