Ibraheem, Omar Waleed: Reconfigurable Vision Processing for Player Tracking in Indoor Sports. 2018
Inhalt
- Introduction
- Vision-based Player Tracking in Indoor Sports
- Introduction
- Challenges in Vision-based Player Tracking
- Architectures for Vision Processing
- State of the Art High-Speed Camera Interface Standards
- Related Work of Vision-based Player Tracking Systems
- Broadcast Video Systems
- Dedicated Cameras Systems
- Commercial Solutions
- FPGA Accelerated Object Tracking
- Summary
- Methodologies and Fundamentals
- Video Preprocessing Algorithms
- Morphological Operations
- Image Thresholding
- Object Segmentation using Background Subtraction
- Graph Clustering
- Multiple Object Tracking (MOT)
- RAPTOR-X64 Rapid Prototyping Platform
- Design Flow
- Summary
- The Proposed Reconfigurable Vision System
- System Overview
- Video Acquisition Module
- Video Preprocessing Module
- Player Segmentation Module
- RGB to Grayscale Converstion
- Background Estimation and Subtraction
- AXI4-Stream to NPI Controller
- Morphological Operations
- Masking
- Team Identification & Player Detection Module
- Resource Utilization
- Player Tracking
- Single Camera Player Tracking
- Detections Association to Tracks
- Player Track Transfer Between the Two Cameras
- Summary
- System Evaluation and Results
- System Realization
- Datasets
- Player Detection
- Player Detection in Occlusion Scenarios
- Verification of the Player Detection Implementation
- Player Tracking
- FPGA Architecture
- Multiple GigE Vision Cameras
- Performance and Throughput
- Acceleration Factor and Overall System Performance
- Overall Latency
- Power Consumption
- Comparison with Existing Systems
- Summary
- Conclusions and Future Work
- List of Figures
- List of Tables
- Abbreviations
- References
- Publications