Sistem Cerdas Pendeteksi Dan Penghitung Jumlah Korban Bencana Alam Menggunakan Algoritma Deep Learning

Abstract

Abstrack – The speed of searching for victims of natural disasters is an important factor that affects the chances of survivors. Most of the locations affected by natural disasters will be difficult to access, and currently, the rescue team is still using heavy equipment to open the access. This requires a long time to go to the location and search for victims. In this research, we propose a smart device to help the National Search and Rescue Agency (BASARNAS) use a drone equipped with cameras to search for victims of natural disasters in real time. By using this smart device, the search will be more effective, because it can speed up the rescue team in searching for victims at the disaster site. The process of detecting and calculating the number of victims is carried out on the camera using the Convolutional Neural Network algorithm and the You Only Look Once version 4 (YOLOv4) architecture. The results of this study are that the Convolutional Neural Network algorithm has a fairly high accuracy of 92.29% in detecting the victim's condition. Besides that, the height of the camera and the position of the camera also affect the accuracy obtained, whereas the results obtained with a static camera have optimal accuracy, which is 95% and the higher the camera, the accuracy will decrease, which is 67%.