Contagem de veículos na via através de imagens de trânsito

Abstract

Congestion are increasingly present in our lives. Studies show that congestion is formed when the number of vehicles exceeds about 1/3 of the capacity of the channel. It is difficult to predict traffic departments points slowdowns in cities. We propose estimating the number of vehicles on the road by analyzing images from traffic cameras collected for registration violations. This should be done with a low error rate. To accomplish this task, we employ in this paper algorithms of computer vision and machine learning, inspired primarily in Jain, Sharma and Subramanian (2012) approach, although the characteristics of the cameras used in this paper are clearly distinct. The experiments show that the proposed strategy presents a low error rate, in both day and nighttime videos.