Sažetak
This paper presents an overview of recent advancements in technologies for indoor fire prevention, detection and suppression, with a focus on sensor-based systems, vision-based techniques, and robotic solutions. Wireless sensor networks demonstrate significant advantages over traditional wired solutions, enabling early detection, real-time localization, and reliable alarm delivery while simplifying installation and maintenance. Multisensor approaches, integrating flame, gas, and temperature detection, have shown high accuracy for Class-A fires, with innovations such as acoustic-based and NB-IoT-enabled smoke detectors enhancing range and efficiency. Vision-based methods have evolved from hand-crafted features to deep learning architectures, including CNNs, optical flow, and hybrid fusion, improving robustness in challenging environments. In suppression, IoT-enabled smart sprinklers and robotic firefighting robots offer autonomous, responsive, and safe operation in hazardous conditions. Furthermore, FireSense, a proactive and autonomous AIoT system for real-time fire hazard detection is designed to meet the technological and safety requirements of Industry 5.0, currently in the development stage by our team, is also presented. The proposed system introduces a novel multimodal deep learning framework that integrates visual data from RGB cameras with environmental data acquired from distributed gas, temperature, humidity, and pressure sensors. A transformer-based architecture is employed to perform multimodal fusion and temporal analysis, enabling early and reliable detection of anomalous patterns associated with incipient fire events. Collectively, these technologies represent an integrated, intelligent approach to next-generation indoor fire safety.
Ključne riječi
fire detection systems fire extinguishing indoor fire machine learning smoke detection