Search

This study analyzed the characteristics of pitched baseballs from TV broadcast videos to understand the effects of the Magnus force on a pitched-baseball trajectory using aerodynamic theory. Furthermore, an automatic measurement and analysis system for pitched-baseball trajectories, ball speeds, and spin rates was established, capturing the trajectory of the baseball thrown by the pitcher before the catcher catches it and analyzing its related dynamic parameters. The system consists of two parts: (1) capturing and detecting the pitched baseball in all frames of the video using the YOLOv3-tiny deep learning algorithm and automatically recording the coordinates of each detected baseball position; (2) automatically calculating the average speed and spin rate of the pitched baseball using aerodynamic theory. As the baseball thrown by the pitcher is fast, and live-action TV videos like sports and concerts are typically at least 24 fps or more, this study used YOLOv3-tiny algorithm to speed up the calculation. Finally, the system automatically presented pitching data on the screen, and the pitching information in the baseball game was easily obtained and recorded for further discussion. The system was tested using 30 videos of pitched baseballs and could effectively capture the baseball trajectories, throw points, catch points, and vertical displacements. Compared with the values from the TV broadcast, the average errors on the calculated ball speed and spin rate were 1.88% and 7.51%, respectively. Using the ratio of the spin rate and ball speed as a parameter to analyze the pitching state of the pitcher’s four-seam fastball in the Nippon Professional Baseball and Major League Baseball matches, it was observed that when this ratio increased, the Magnus displacement of the ball increased, thereby decreasing its late break. Therefore, the developed system provides scientific pitching data to improve the performance of baseball pitchers.