Traffic Congestion and Road Anomalies Detection Using CCTVs Images Processing, Challenges and Opportunities

Abstract

Traffic congestion and road accidents are some of the most common problems that urban residents suffer from. Because they have a direct impact on their security, physical and psychological health, and even on the economic side etc. this is due to several reasons including the huge number of vehicles on roads, irresponsibility of drivers, and also the state and the architecture of the infrastructures. In order to solve these problems or at least reduce them, several studies have addressed the issue and proposed different solutions from different sides. In this paper, we would like to study the possibility to use Computer Vision techniques and Big Data Analytics subfields, such as machine learning and deep learning algorithms, to build reliable systems which will allow us to detect, identify, and track vehicles and pedestrians in real-time. This will enable us to detect traffic congestion or any other anomaly. These systems will be installed in the network of the Closed-Circuit Televisions (CCTVs) fixed on roads and make them connected. Moreover, this study provides comparative analysis progress of the methods used to detect objects, especially vehicles and pedestrians to choose the most accurate model capable of tracking road users and analysing their behaviours, detecting frauds and saving useful numerical datasets.