SISTEM DETEKSI MULTI-ROBOT DAN API MENGGUNAKAN IMAGE PROCESSING BERBASIS ALGORITMA YOLO

Abstract

Accidents due to undetected fire have caused huge losses in various sectors in the world, such as office buildings, residential areas, and forest areas. This causes the need for an efficient fire detection system to increase. Image processing based on object detection system is considered capable to overcome this problem. One of the methods that are used to detect objects, and that is now developing, is the deep learning method. YOLO algorithm is a part of deep learning method. Therefore, this research will build a fire detection system with robots and fire as the objects, using YOLO-based image processing in real time. The most suitable YOLO model for the system is the Tiny YOLO VOC model, which is built with the darkflow framework. The system can detect the fire and robots with 100% accuration and a training loss value of 0.816988. The confidence value obtained to detect Robot_1 objects is 84%, Robot_2 is 92% and Fire is 88%. Thus, this research proves that the image processing system with YOLO algorithm to detect fire is a success, and can be implemented on a fire extinguisher system.