SAS-DV: Uma arquitetura de predição de riscos de colisão para pessoas com deficiência visual

• Ano de defesa: 2020
• Autor(a) principal: Cordeiro, Natal Henrique
• Orientador(a): Pedrino, Emerson Carlos
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus São Carlos
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação - PPGCC
• Palavras-chave em Português: Fusão de informação de alto nível; Predição de risco de colisão; Deficientes visuais; Consciência da situação
• Palavras-chave em Inglês: High level information fusion; Collision risk prediction; Visually impaired people; Situation awareness
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::SISTEMAS DE COMPUTACAO
• Link de acesso: https://repositorio.ufscar.br/handle/20.500.14289/14135

The technologies developed so far to help visually impaired people (VIP) navigate meet only some of their everyday needs. This project allows visually impaired people to improve the situation awareness (SAW) by generating a risk map following an analysis of the position, distance, size and motion of the objects present in their environment. This comprehension is refined by data fusion steps applied to High Level Information Fusion to predict possible impacts in the near future. A risk map is formed after a set of inferences has been executed. Two specific datasets for projecting the risks of collision in different directions are proposed in this work for the execution of these inferences. These datasets are composed of information regarding the position and size of static objects, free passages, dynamic objects, and paths executed by dynamic objects. All this information was mapped in a 3D plane. Thus, to demonstrate the advantages of implementing High Level Information Fusion in the navigation system dedicated to VIP, experiments were performed with the proposed architecture and with three other navigation systems implemented with different approaches. The results demonstrated it was possible to validate and compare the navigation systems. For this comparative analysis, different situations were tested by the navigation systems so that the VIP could be indicated, namely a direction to travel which includes fewer collision risks. In addition to providing a risk map giving possible collisions, this study provided greater reliability for navigation, especially when obstacles were very close and moving objects were detected. Different techniques for detecting and tracking dynamic objects were also implemented in this study for defining which of them is best suited to the VIP context. This study can provide new contributions such as adaptations to already renowned techniques used for dynamic object analysis and, most notably, with the use of a methodology for comparing the efficiency and computational cost of each technique to analyze such objects.