Smart cities traffic congestion monitoring and control system

Abstract

The traffic monitoring projects responsible for the current traffic monitoring infrastructure utilized by companies and government agencies tend to be very expensive and require difficult and extensive implementation. The challenge and goal of this paper is to create a smaller scale, low cost method of analyzing, controlling, and predicting traffic conditions. Traffic data including car count, frequency, and direction, is gathered from a USB camera and sent to a microcontroller to be interpreted using computer vision libraries. The traffic data is then transferred and stored onto the cloud to be further analyzed. This paper focuses on two aspects of managing traffic. The first aspect involves the optimization of traffic cycles at an intersection using incoming car counts to minimize the wait time between traffic light cycles. The second aspect involves predicting future traffic flow by training a deep neural network utilizing collected traffic data and machine learning techniques.