Computer Vision in Sports: Real-Time No-Ball Detection in Cricket using Motion Rythms and Thresholding Based Approach

Abstract

Computer Vision has been a vital part in sports decisions, analysis and broadcasting. The motion properties, gestures and other activities of the players or the sports objects can help in detection, tracking and thus, decision making and analysis in sports. Cricket involves computer vision in many decisions like LBW(Leg-Before-Wicket), decision reviews and other broadcaster analysis, Tennis makes use of ball tracking technique for making line-calls. No balls in Cricket are the illegal deliveries that cannot claim a wicket and award an extra run to the opponent team. We propose a computer vision based approach which can automatically detect the front-foot no-balls using the techniques like object detection, edge detection and thresholding based on motion rythms. This work tries to manage the workload of the umpires with some assistance from the machines. The proposed model would take the side-angle video feed, detect the first point of contact between the foot and the ground(Point of Landing) implicitly and determine in real-time whether the delivery is fair or it is a no ball with the help of Computer Vision techniques, Object Detection, Edge Detection and Image Thresholding based on pixels. This model would have lesser manual intervention than the techniques tested so far. The accuracy obtained on testing the approach on newly prepared dataset of 40 side-angle videos of bowling shot in LBS ground, IIT Roorkee was 82%.