International Journal of Scientific Research and Engineering Development

Fire detection technology has evolved significantly from basic mechanical systems to sophisticated intelligent platforms integrating artificial intelligence and multi-sensor fusion. This comprehensive review examines the state-of-the-art in fire detection systems, with particular emphasis on Convolutional Neural Network (CNN) approaches, multi-sensor data fusion methodologies, and Internet of Things (IoT) integration. Traditional single-sensor fire alarm systems suffer from high false alarm rates (up to 95% in some jurisdictions), detection latency issues, and limited discriminative capability between genuine fire events and benign environmental conditions. Recent advances in deep learning, particularly CNN architectures such as You Only Look Once (YOLO), Residual Networks (ResNet), MobileNet, and specialized fire detection models, demonstrate substantial improvements in detection accuracy while maintaining real-time performance capabilities. Multi-sensor data fusion techniques combining temperature, smoke, flame, and visual sensors through decision-level, feature-level, and signal-level integration strategies offer enhanced robustness and reduced false alarm rates compared to single-sensor systems. Kalman filtering and Bayesian fusion methods provide mathematical frameworks for optimal sensor integration under uncertainty. However, significant challenges remain in deploying sophisticated machine learning models on resource-constrained embedded platforms, necessitating model optimization through quantization, pruning, and efficient architecture design. This review synthesizes current research across fire detection methodologies, critically analyses their strengths and limitations, identifies persistent research gaps including standardized evaluation frameworks and real-world validation needs, and proposes directions for future investigation. The integration of embedded machine learning, IoT connectivity, and intelligent sensor fusion represents a promising pathway toward next-generation fire safety infrastructure capable of reliable, autonomous operation across diverse building environments.