Hybrid-FireID: Fire Identification using Hybrid Features Extraction for Combustible and Fluid Fire Segmentation

Abstract

Fire detection systems are implemented and intended to detect fires early so it can help the people on a building or house for safe evacuation and immediately notify the firemen. After the firemen put out the fire, that will be the time that they can conduct an investigation in determining the source or cause of fire which they often experience some difficulties. Therefore, this study proposed an algorithm for identification of combustible and fluid fire with hybrid feature extraction techniques. The algorithm use RGB model, applying HSV conversion and Canny edge detection for the growth of fire. Then combine the results of HSV and Canny edge detection and used image segmentation of color space for combustible and fluid fire. The algorithm got an accuracy of 94% for 50 fire images demonstrated usefulness and effectiveness.