Alif Xavier, Rizqon (2025) Sistem Pendeteksi Kecelakaan Lalu Lintas Menggunakan Algoritma Yolov8. Other thesis, Universitas Islam Riau.

Text 203510661.pdf - Submitted Version Restricted to Registered users only Download (10MB) | Request a copy

Abstract

raffic accidents are one of the main problems that often occur in Indonesia, with the number of incidents continuing to increase every year. This study aims to develop a video-based traffic accident detection system using the YOLOv8 algorithm, which is equipped with an automatic notification feature via Telegram BOT. The research data was obtained from the YouTube and Twitter platforms, then processed into 2,949 frames from 8 videos obtained after going through the preprocessing stage, such as cutting the video into frames, cleaning data to remove bad data, and labeling objects manually using Roboflow. Labeling includes seven classes: people, car, motorcycle, truck, bus, bajaj, and accident. The system was trained using two types of optimizers, namely AdamW and SGD, with evaluations showing that the AdamW optimizer gave the best performance, with mAP50 reaching 98.3% for all classes. The resulting model was able to accurately detect accidents in test videos. In addition, the system automatically sends a notification to Telegram, which contains an image of the accident frame along with information on the location of the incident. This system is expected to accelerate the authorities' response to accidents, reduce the impacts caused, and support the development of smart cities by providing more accurate and real-time accident data. This research makes a significant contribution to the application of artificial intelligence technology to improve traffic safety.

Actions (login required)

View Item