IMPROVED FAULTY VEHICLE DETECTION AND CLASSIFICATION FOR INTELLIGENT TRAFFIC MANAGEMENT SYSTEM USING MACHINE LEARNING APPROACH.

Abstract

One of the most important aspects of a smart city is an intelligent traffic light control system. The number of people who own cars has dramatically increased, which causes traffic jams in urban areas, especially in Nigeria. One of the most difficult issues individuals deal with on a daily basis is traffic congestion, which reduces productivity at work, increases the risk of accidents, and more. The efficiency of security personnel's traffic control has not been demonstrated. This issue was not resolved by the traffic light's static or constant time under all conditions (low and high vehicle volumes). Additionally, studies that take into account counting, classifying, and detecting the number of vehicles at intersections lack sufficient accuracy in doing so because they are unable to identify and categorize defective vehicles among the vehicles, which could result in delays, waste of time, or even an accident at the intersection. In order to increase efficiency and decrease delays or time wasting, this study suggested creating an enhanced vehicle classifier with an effective timer that could identify and categorize defective vehicles among those on the side of the crossing as well as allocate the proper amount of time. Additionally, the proposed system's scalability was tested on a large dataset using a camera and machine learning approach; vehicles at intersections were detected and classified using a Faster R-CNN based detection model; and a timer was developed using the DevC++ Environment to help with appropriate timing, which in turn reduces fatalities, wastes time, and increases productivity. The suggested system's efficiency was assessed using performance measures, and it was found to have an F1 score of 95.08% and an overall accuracy of 97.52%.

DOI: https://doi.org/10.5281/zenodo.21196132