Image fire detection module for automatic fire extinguishing system with unmanned ground vehicles

Especially in responding to large fires, the use of unmanned vehicles can reduce the risk of people getting hurt or encountering situations where they can get hurt. At the same time, the use of unmanned vehicles can increase the efficiency of the intervention. In this direction, one of the most important modules for the unmanned ground vehicles to be used to achieve the desired results is the fire detection module, which will detect the fire and report it to the necessary systems for intervention. In this study, certain deep learning networks were examined for fire detection. These networks are Faster-RCNN, Mask-RCNN, SSD and YOLO. After these networks were trained with the same data sets, they were compared with FPS and mAP data. As a result, it was seen that the YOLO algorithm gave a more positive result than other deep learning networks in terms of both detection and output speed. As a result, YOLO was selected and used as the deep learning network to be used for fire detection.