Pendeteksian Pelanggaran Jalur Bus Transjakarta Menggunakan Algoritma YOLOv3

Abstract

Police of Metro Jaya fined about 42,286 Motorcycles that crossed lane of Transjakarta in a period of eleven months in 2014. To reduce the number of bus lane violation, it’s very important to take some continues action from the authorities. The purpose of this study is to detect vehicles that break through the bus lanes with the application of computer vision until get the notification of bus lane violation. In this study used training data as much 470 pictures data with three classes of vehicles like motorcycles, bus, and cars. This study used You Only Look Once v3 algorithm with training data that was trained as many as 6000 Iterations which resulted in an map of 99.96% and was succesfully that used to detect vehicles with help of sort method in kalman filter to add counting function to detect vehicles that cross in bus lane with the dividing lane, and creating some new file in .mp4 format for insertion into the GUI.