Sistem Deteksi Kendaraan Menggunakan Metode You Only Look Once (YOLO) pada Jalanan Kota Medan

Abstract

Monitoring traffic conditions is very important with the increasing population, it will affect the number of vehicles in Medan City as an effort to regulate and smooth traffic. Medan City is one of the areas with a fairly high level of vehicle density, especially in the main traffic area. Various types of vehicles pass every day, so technology is needed that can detect and identify vehicles accurately and efficiently. In this study, the author uses the YOLO (You Only Look Once) method as an object detection algorithm to test its ability to recognize various types of vehicles passing through Medan City traffic. This study uses objects in the form of cars, motorcycles, buses, trucks, and pedicabs. This system is built using the You Only Look Once method which groups vehicle types and counts vehicles. From the results of this test based on the YOLO method using 4 test videos with a video duration of over 40 seconds, 180 vehicles have been successfully detected correctly out of a total of 189 vehicles. From the calculation of accuracy and other performance calculations, it was found that the system that implements the You Only Look Once (YOLO) v5 method has an accuracy of 95%.