A PROPOSED FIRE DETECTION SURVELLIANCE THROUGH MACHINE LEARNING BASED ON HAAR CASCADE CLASSIFIER

Fire is an unwanted event that could carry a high loss of social wealth and human life. To stop such losses, different alarm systems such as smoke detectors, and temperature sensor-based systems have been developed. Our proposed system is aimed to design and develop a fire detection system that detects fire without any heat or temperature sensor. The primary objective of the fire detection system is to detect a fire early and warn authorities when a fire takes place. The Machine Learning Algorithm has been used to detect the accurate image of fire because it has a prior pattern of fire images already fed into it. On occurring the fire, the camera will send this pattern to Raspberry pi which has already predefined patterns of fire written in the form of an algorithm and afterward will compare it with the new existing fire pattern. When both the pattern matches system will do processing based on the image processing technique. Finally, the system generates a warning message which will be sent on the LCD screen for display, and thereafter the buzzer starts working. The key benefit of this system is it will decrease the risk of losses which occurs mainly due to failure in controlling the fire. The experimental results showed that the designed system can efficiently extract and keep trace of fire pixels in the form of patterns and is worthwhile in providing better output results. This system has compact circuitry and functionality like it is easily implanted in public and commercial places for security and surveillance