Smart Fire Detection System with Early Notifications Using Machine Learning

House fire is one of the major concerns for designers, builders, and residents of property. In the case of detecting fire, individual sensors have been used for a long time, but they cannot detect the level of fire and notify the emergency response units. To solve this problem, this study attempts to propose an intelligent early fire detection system that would not only detect the fire by using integrated sensors but also notify the appropriate authorities including fire department, ambulance services, and local police station simultaneously to protect valuable lives and properties. Signals from the integrated detectors e.g., heat, smoke, and flame go through the machine learning algorithms to check the potentiality of the fire as well as broadcast the predicted result to various parties using a GSM modem. To consolidate the predicted output, structured forest for fast edge detection has also been applied. The final outcome of this development also minimized false alarms, thus making this system more reliable. © 2017 World Scientific Publishing Europe Ltd.