Deep learning based highway vehicles detection and counting system using computer vision

Abstract

In the field of highway management, intelligent vehicle detection and counting are becoming increasingly significant. City planners have been fighting the issue of traffic analysis for years. To analyze traffic and streamline the process, intelligent methods are being created. The type of vehicles and the number of vehicles in an area over some arbitrary time period may be taken into account. Over the years, humans have created a variety of mechanisms, because the vehicles are come in different sizes, it is still difficult to identify them, which can directly impact the accuracy of vehicles count. In the field of vision research, computer vision is one of the most often used applications for traffic scenarios. A variety of computer vision applications, including object detection and categorization, have proven that deep learning techniques like Convolutional Neural Networks (CNNs) greatly outperform conventional approaches. Hence in this work, deep learning-based highway vehicles detection and count system suing computer vision is presented. Vehicles on the highway can be efficiently detected by this technique. YOLOv3 (You Look Only Once Version 3) is a real-time object detection technique which is used to detect vehicles in real-time. Precision, recall and real-time rate are measured to evaluate the performance of this system.