Automated Wicket-Taking Delivery Segmentation and Weakness Detection in Cricket Videos Using OCR-Guided YOLOv8 and Trajectory Modeling

TL;DR

The paper tackles the labor-intensive task of cricket video analysis by integrating OCR-based wicket segmentation with YOLOv8-driven pitch and ball detection and a trajectory-modeling framework. The approach delivers a three-component system that first extracts wicket events from scorecards, then accurately detects pitch regions and balls, and finally models ball trajectories to identify weakness zones. Key contributions include an OCR-guided wicket segmentation method, optimized pitch detection achieving $mAP_{50}=0.995$ and ball detection achieving $mAP_{50}=0.9918$ (with transfer learning), and a trajectory-heatmap framework for weakness identification. Experimental results across multiple match videos demonstrate robust wicket-event detection, near-perfect pitch detection, and high ball-detection accuracy, enabling scalable, data-driven coaching and strategic decision-making. This work advances cricket analytics by enabling automated, interpretable insights that can inform training and game strategy, with future work aimed at real-time analysis and broader contextual factors.

Abstract

This paper presents an automated system for cricket video analysis that leverages deep learning techniques to extract wicket-taking deliveries, detect cricket balls, and model ball trajectories. The system employs the YOLOv8 architecture for pitch and ball detection, combined with optical character recognition (OCR) for scorecard extraction to identify wicket-taking moments. Through comprehensive image preprocessing, including grayscale transformation, power transformation, and morphological operations, the system achieves robust text extraction from video frames. The pitch detection model achieved 99.5% mean Average Precision at 50% IoU (mAP50) with a precision of 0.999, while the ball detection model using transfer learning attained 99.18% mAP50 with 0.968 precision and 0.978 recall. The system enables trajectory modeling on detected pitches, providing data-driven insights for identifying batting weaknesses. Experimental results on multiple cricket match videos demonstrate the effectiveness of this approach for automated cricket analytics, offering significant potential for coaching and strategic decision-making.

Figures (6)

• Figure 1: Complete system methodology flowchart showing the three main stages: (1) Wicket segmentation using OCR-based scorecard analysis, (2) YOLOv8-based pitch and ball detection with preprocessing pipeline, and (3) Trajectory modeling for weakness zone identification.
• Figure 2: YOLOv8 architecture used for pitch and ball detection, showing backbone (CSP network), neck (FPN), and detection head components.
• Figure 3: Pitch detection training curves showing loss metrics and performance improvements across epochs for Training 1.
• Figure 4: Ball detection training results using transfer learning, showing superior convergence and performance compared to training from scratch.
• Figure 5: Sample detection results showing successful pitch and ball identification with bounding boxes overlaid on cricket match frames.