Real-Time Object Detection Using YOLOv8 for Transport System in Case of Tulu Dimtu to Adama Toll Roads

Abstract

Real-Time object classification and detection are crucial tasks in computer vision that address issues in transportation systems and provide a framework to enhance safety and efficiency in transportation monitoring and surveillance. The study focused on road patrolling operation system designed for the Addis Ababa (Tulu Dimtu) to Adamma highway toll road. The proposed system was implemented as the state-of-the-art deep learning algorithm, specifically the YOLOv8 (You Only Look Once version 8 algorithm, to enable efficient vehicle counting, identifying, and classification. The proposed system offers essential real-time information, including object counting, detection, and classification, which can lead to enforcing transportation regulations and optimizing transport management. The system consists of three key stages: object counting, detection, and classification. The experiment result showcases the effectiveness of the proposed system in providing valuable insights to road authorities, ultimately contributing to improved transportation services and road patrolling operations on the expressway. This research insights into the potential of object detection and classification algorithms, and real-time object detection in addressing and visualizing by tracking any events as valuable information for transportation monitoring applications. The results were obtained using highway video datasets and implementing the YOLOv8 algorithm to demonstrate the accuracy of the proposed vehicle and object identification and classification system for the transport system from Addis Ababa to Adama highway. The model has been trained using diverse highway datasets and YOLOv8s pre-trained model produced precision, recall and mAP50 are 95%, 95.5%, and 98.4% respectively. And YOLOv8n produced precision, recall and mAP50 are 94.8%, 95%, and 98% respectively. The result showcases the system's ability to accurately locate objects within video frames. Furthermore, object classification provides valuable information such as object type and class. These results show the algorithm's capabilities to identify and classify objects precisely, used to enhance transportation management and improved road services in the specified events.