A HYBRID CNN–RANDOM FOREST AND CNN–XGBOOST FRAMEWORK FOR REAL-TIME FOREST FIRE DETECTION

Abstract

This technology aims to detect forest fires immediately. Finding flames is made possible through methods for processing video and photos. Methods for photo recognition constitute the bulk of the effort. Due to the dynamic nature of fire borders, a procedure involving the elimination of the background is necessary. By utilizing a color segmentation approach, the potential zones are subsequently located. Use of a Convolutional Neural Network (CNN) allows for the detection of fires in the selected regions. A real fire may be located with pinpoint accuracy even in the absence of monitoring devices, and the forest service is promptly alerted.