Diagramas veiculares espaço-tempo em vias urbanas utilizando a visão computacional

• Autor(a) principal: Santos, Dênnys Araújo
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da Universidade Federal do Ceará (UFC)
• Texto Completo: http://www.repositorio.ufc.br/handle/riufc/50322
• topic: Transportes; Visão computacional; Engenharia de tráfego; Urban microscopic flow; Space-time diagram; Ffeature-based tracking

The Brazilian management of urban vehicular traffic lacks robust strategies for disaggregated variables estimation with a high level of accuracy. The space-time diagram is a classic tool for traffic analysis and still relevant nowadays. In turn, the Computer Vision has been widely used successfully in traffic studies. Amongst the main methods used in these studies, the Feature-based tracking has proved to be the most suitable for the urban environment for dealing adequately with situations of partial occlusion of vehicles, interruptions of vehicle movements among others. Thus, this work aims to use Computer Vision to produce space-time diagrams and to estimate urban microscopic flow variables. Therefore, this dissertation proposes a method to assist disaggregated studies of urban vehicular traffic through space-time diagrams. The algorithm developed in this work had a success rate of 91,52% for vehicle counting. Then, a case study was conducted to estimate variables of characterization of urban traffic of interrupted flow through the space-time diagrams created by using the vehicular information extracted by the Computer Vision algorithm. The following parameters were estimated: vehicle length, distance between stopped vehicles, queue length, number of stopped, number of lane changes, average speed, average delay and headway. By the end, the robustness of the Computer Vision algorithm as well as the fidelity of the space-time diagrams were verified. Therefore, the method proposed in this work proved to be viable and flexible to be used in several disaggregated studies of urban microscopic vehicular traffic.