Computer vision for traffic and object detection

• Autor(a) principal: Mendes, Marcos André Lopes
• Data de Publicação: 2023
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/41041
• topic: Accident detection; Vehicular collision; Centroid based object tracking; Collision avoidance; Edge computing; Smart city; Vulnerable road users

On average, in the European Union, there were 44 road deaths per million inhabitants in 2021, representing an increase of 5% compared to 2020, but a decrease of 13% compared to 2019, before the pandemic. In Portugal this trend represents the occurrence of 52 road deaths per million inhabitants in 2021. This work aims to improve safety at crosswalks and predict collisions in cities. This work aims to have a significant impact in terms of road safety and protection of users, by improving safety at crosswalks, preventing collisions, reducing accidents and injuries, increasing safety in a public environment, with the versatility to be implemented in other cities. This thesis also aims to demonstrate the benefits of computer vision for the development, control, optimization and safety of a city. It proposed mechanisms to retrieve information from video cameras to analyse mobility and danger situations. This way, we study danger relations in crosswalks, in order to understand how the danger varies in this type of zone, considering cars and pedestrians. The goal will also be to do this study for different zones and verify data sets that can lead us to draw conclusions about a given occurrence and thus activate the necessary means to solve them. The goal of the thesis was the development of algorithms for accident detection and prediction, which allow the city to become more interconnected and more intelligent from a technological point of view.