Real-Time Decision Modeling of Basketball Game Tactics Based on Video Analytics

Abstract

Basketball games, real-time adjustment of defense strategy according to the changes on the court can greatly improve the team's chance of success in defense and thus win the game. In sophisticated team sports, it can be particularly difficult to judge these kinds of talents. This study attempted to develop an accurate and dependable video-based decision-making evaluation for young basketball in order to solve this problem. In this manuscript, Progressive Graph Convolutional Networks based onreal-time decision modeling of basketball game tactics based on video analytics (RDBGT-PGCN-GOA) is proposed. First, the image is taken from the NBA basketball video collection, and the pre-processing section receives the acquired data after that. When preparing, Unsharp Structure Guided Filtering (USGF) is used to remove background noise from the image. Then the preprocessed output is fed to Progressive Graph Convolutional Networks (PGCNs) is successfully used to classify the game tactics such as the Body Postures, Player Positions and Player Actions. Progressive Graph Convolutional Networks (PGCNs) classifiers, in general, do not express adaptive optimisation procedures to find the best parameters to guarantee accurate classification of player positions, player actions, and body postures. Hence, proposed GOOSE Optimization Algorithm (GOA) enhances Progressive Graph Convolutional Networks (PGCNs), accurately classify game tactics such as Body Postures, Player Positions and Player Actions. The weight parameter of the PGCN optimized with GOOSE Optimization Algorithm (GOA) for accurate prediction. The proposed RDBGT-PGCN-GOA proposed is implemented on the Python working platform. The performance of proposed method examined utilizing performance metrics likes Accuracy, Precision, Recall,F1 score, Error rate, and specificity were looked at. The proposed RDBGT-PGCN-GOA approach contains 23.52%, 22.72%  and 24.92%  higher accuracy; 23.52%, 24.72% and 21.92% lower Error rate compared with existing methods, such as Basketball video analysis using deep learning algorithms for technical features (TFBV-DNN), offline reinforcement learning for tactical strategies in professional basketball games (TSPBG-RNN), and real-time defensive strategy optimization using motion tracking and deep learning (RTBDS–CNN) are the three approaches that are being examined.