Detection of vehicles and analysis of traffic volume by real-time video-graphic technique using python algorithm

• Published in: Discover Civil Engineering Vol. 2; no. 1; pp. 146 - 19
• Main Authors: Ramadan, Osman Kantunsung, Kumar, R. Srinivasa, Reddy, A. Narender
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 14.08.2025; Springer Nature B.V; Springer
• Subjects: Algorithms; Automobiles; Civil Engineering; Classification; Computer vision; Data acquisition; Engineering; Identification; Intelligent transport systems; Python algorithm; Real-time video; Roads & highways; Statistical analysis; Traffic analysis; Traffic control; Traffic flow; Traffic volume; Transportation planning; Vehicles; YOLO; YOLO; Intelligent transport systems; Real-time video; Traffic volume; Python algorithm
• ISSN: 2948-1546; 2948-1546
• DOI: 10.1007/s44290-025-00288-8

Transport management is burdened by traffic congestion, including ineffective vehicle detection systems. The research proposes a real-time traffic count detection through video analysis, which combines computer vision methods. Vehicle detection and classification, alongside counting on a real-time basis, are enabled through the proposed system running YOLO (You Only Look Once) on a Python programming framework. First, followed by data acquisition of traffic videos, is the methodology implemented through five consecutive procedural steps, including algorithm development, system testing, and data exportation before manual vehicle count validation. YOLO performs object detection more effectively due to preprocessing functions, including negative image filtering and image rotation augmentation with enhanced classifier reliability. Each detected vehicle received a numbered box for identification through which tracking and consistent counting occurred between entry points and exits. The traffic evaluation component analyzed quantities including average vehicle speed, detection type, and size measurements. The detection accuracy achieved through testing three 5-min traffic videos in different conditions reached 92%, 90.84%, and 92.16%. The Mean Absolute Percentage Error performance evaluation produced three separate results totaling 2.5%, 3.1%, and 2.8%. GEH statistical analyses confirmed the model reliability at minimal error rates through three metrics that read 4.2, 3.8, and 4.0. These statistics showed acceptable reliability levels. The detection data helped PTV Vissim conduct traffic simulations for existing and proposed traffic conditions. The updated signal timing schedule demonstrated decreased delays, including reduced queue formations for better service levels. The study shows that video analysis in real-time provides precise traffic counting, which opens opportunities for Intelligent Transportation Systems (ITS) systems to utilize.