A Hierarchical Sensing IoT Framework for Fire Detection Integrating Low-Power Sensors with On-Demand AI-Driven Visual Verification

Fire detectors play a vital role in terms of safeguarding the residential, industrial, and commercial settings. Nevertheless, traditional single sensor detectors tend to give false alarms because of non-hazardous factors like steam, dust or cooking smoke. On the one hand, the vision-based systems although more precise consume too much power when constantly in use. This paper suggests a multinode two-layer hierarchical sensing architecture that integrates the scalability of sensor-based systems with accuracy of on-demand object detection. The system connects a set of ESP32 nodes, which have temperature, flame and smoke sensors and are connected to a central Raspberry Pi via NRF24L01 modules. Upon detecting the abnormal readings, the Raspberry Pi triggers an ESP32-CAM to check the visual confirmation of an abnormal reading with the help of a trained flame detection model. The prototype had a 2-3 second detection latency, 97.8% communication reliability, 96% visual accuracy and 60% lower power consumption than continuous vision systems. The proposed solution provides a scalable and low power with high accuracy solution in smart IoT-based fire monitoring.