The evolution of smart cities has highlighted the critical need for intelligent crowd monitoring in modern sports facilities. This paper presents a comprehensive IoT-based crowd monitoring system designed specifically for gymnasium environments. Traditional monitoring approaches often struggle with real-time accuracy and system responsiveness, particularly during high-occupancy events. To address these challenges, we developed an integrated system incorporating multi-sensor fusion, edge computing, and cloud analytics. The system architecture employs a novel dual-stage attention network for sensor fusion, achieving a 27% reduction in data conflicts while maintaining real-time processing capabilities. Our implementation includes strategically positioned 4K cameras, infrared sensors, and UWB positioning devices, supported by an optimized MobileNetV3 edge computing framework. Through extensive testing in a standard gymnasium environment, the system demonstrated exceptional performance with 97.8% detection accuracy under normal conditions and 95.2% accuracy during peak loads. The hierarchical alert mechanism, combining LSTM networks and gradient boosting classifiers, achieved a remarkably low false alarm rate of 0.1%. The system successfully handled 10,000 concurrent connections while maintaining five-nines availability. Real-world deployment validated significant improvements in crowd management capabilities, including enhanced emergency response efficiency and reduced congestion. These results establish a robust foundation for next-generation crowd monitoring systems, offering practical solutions for smart facility management challenges.