Measuring Similarity in CCTV Systems for a Real-time Assessment of Traffic Jams

Abstract

Traffic jams are inevitable on the roads that many of us use every day. Their number and scale are generally increasing, especially in cities where economic activities are flourishing. The causes of these traffic jams are numerous and generally have economic and socio-environmental consequences. Many solutions have been proposed for detecting traffic jams without considering mathematical tools. In this article, we propose to provide solutions based on mathematical tools which make it possible to measure the similarity between two successive images acquired via closed circuit television (CCTV) systems. This similarity measure will allow us to assess the state of traffic jams in a CCTV system in order to prevent them. By analyzing the transmission of images through a variable sliding window, the implementation of the SSIM (Structural Similarity Index Measure) and the cross-correlation metrics which make possible to measure the similarity between two successive images in transmission in standardized Performance Evaluation of Tracking and Surveillance (PETS) datasets. The comparison between these two metrics based on the processing time and the probability distributions reveals that the SSIM metric provides better performance to prevent traffic jams.