An approach to automate the scorecard in cricket with computer vision and machine learning

Cricket is beyond the shadow of a doubt one of the most popular forms of sports in the southern region of Asia. This form of sports is widely played in more than 125 countries recognized by the International Cricket Council. With the flourishing of cricket, various aspects of the game are being automated with the advent of technology. Use of computer vision in assisting third umpire decision is indubitably the well-liked one. One of the most challenging issues that first initiates the discussion on its prosperity is the duration of the game. The on-field umpire has to authorize decisions almost after each delivery following which modifications are carried out in the scorecard which is a very tedious process. The traditional approaches that have been considered so far involve wearing a specialized hand glove, which collides with the beauty of the originality in the field. In this paper, an approach is proposed and a prototype is implemented to automate the umpires decision by interpreting his hand gesture. The region of interest is selected using a Haar-cascade-classifier and then the particular gesture is recognized using logistic regression. This process would eliminate the manual updating of scorecards and thereby reduce the game duration notably. In addition, it excludes the prerequisite of wearing special gloves involving sensors. The efficiency of the algorithm is then cross-checked with the training and test data. This proved to be a very simple but efficient algorithm for umpires gesture detection.