Vehicle Counting Method Based on Convolutional Neural Network with Global Density Feature

Abstract

Despite efforts to minimize and reduce it, traffic congestion has been one of the main problems that most metropolises are experiencing. One of the biggest issues facing engineers, planners, and policy-makers in metropolitan settings has been traffic congestion. The proposed approach for managing traffic is based on machine learning methods. To improve traffic control and management, a robust and trustworthy traffic monitoring system is essential. The detection of vehicle traffic a crucial component of the surveillance system. The traffic flow aids in management and control, particularly when there is a traffic jam, by displaying the traffic situation at regular intervals. A traffic surveillance system for vehicle counting is proposed by the proposed system. The suggested technique processes the image and classifies the vehicle using an SVM algorithm, whereas the YOLO-5 system uses convolution neural networks (CNN) to recognise objects in real time. The results of the experiments demonstrate that the suggested system is capable of offering accurate information for traffic surveillance in real time. The use of machine learning techniques helps to lessen traffic and congestion. Additionally, this method lessens the need for human labor when continuous monitoring is used.