Uma Nova Arquitetura para Classificação de Tráfego de Véıculos Baseado em Processamento de Imagens

• Andre Luis Cristiani
• Renan Guilherme Nespolo
• Luis F. C. Maschi
• Luis Nakamura
• Jo Ueyama
• Rodolfo I. Meneguette

Resumo

With the increase of vehicles in the big urban centers we can observe an increase in the number of congestion and accidents in the public ways caused not only by the lack of attention of the driver but also by the lack of a traffic infrastructure that assists the driver in the decision making before a event or condition of the route. One of the services provided to the driver is the monitoring of the roads to detect the flow of vehicles that are traveling in a road. In order to increase the efficiency of the traffic flow and optimize the process of propagating this information to the user, we propose an architecture in which to detect and classify the flow of vehicles on the highway. For this, the proposed architecture has a mechanism in which the concepts of computational vision are used to extract the maximum information from the roads in order to inform drivers about the traffic conditions of vehicles on Brazilian highways. The efficiency of the proposed method was evaluated. The results showed that the developed method obtained a lower false positive around 3%, a lower false negative (9%), consequently a better classification of vehicles. In addition, the proposed solution obtained a flow classification score around 90