Basketball game tactics decision modeling using video analytics for action recognition

Abstract

Basketball players typically use the jump shot to score points while playing. Because of this, the jump shot is considered the most important aspect of basketball talent and calls for extraordinary ability. Coaches can make better decisions and significantly improve their athletes’ competitiveness by utilizing body domain network technology in sports training and posture recognition. This manuscript proposes the decision modeling of basketball game tactics based on video analytics (TM-BPJA-SBTT-DVAN). Initially, images were collected from eight male testers performing various basketball-related actions. The input image is fed into the pre-processing stage using Interaction-Aware Labeled Multi-Bernoulli Filter (IALMBF) to remove noise and improve image quality. Then, the pre-processed images are provided for feature extraction using the Signed Cumulative Distribution Transform (SCDT) to extract geometrical features such as area, perimeter, centroid, solidity, and slope. These extracted features are then passed to the Directional Variance Attention Network (DVAN) to recognize the shooting actions of basketball players based on sports biomechanics. However, DVAN does not adopt adapting optimization strategy to determine optimal variables to identify the shooting actions of basketball players. To address this, the Black-winged Kite Optimization Algorithm (BKOA) is applied to improve the weight parameters of DVAN, enhancing its ability to recognize shooting actions. The proposed TM-BPJA-SBTT-DVAN is then implemented in Python, and performance metrics such as Precision, Recall, Accuracy, F1-Score, Specificity, ROC, and Computational Time are analyzed.