A new methodology applied to dynamic object detection and tracking systems for visually impaired people

Autor(es)

Cordeiro, Natal Henrique ^{[1, 2]} ; Pedrino, Emerson Carlos ^[1]

Afiliação do(s) autor(es)

^[1] Univ Fed Sao Carlos, Dept Comp Sci, Sao Carlos, SP - Brazil

^[2] Fed Inst Sao Paulo, Dept Comp, Votuporanga, SP - Brazil

• Tipo de documento: Artigo Científico
• Fonte: COMPUTERS & ELECTRICAL ENGINEERING; v. 77, p. 61-71, JUL 2019.
• Citações Web of Science: 1

Resumo

The presence of dynamic objects (DO) affect the mobility of the visually impaired people (VIP) and so it is important to indicate the safest areas to move in. This article presents a new methodology for detecting and tracking DO to reproduce their paths. This methodology consists of: a requirements analysis specific to the context of the VIP; the development of four architectures adapted to the VIP; the generation of a 3D map composed of paths made by DO; and a comparative analysis between the architectures in order to obtain the best solution from the union of the best characteristics detected. These architectures, the main adaptations being related to the factors of illumination and movement undergone by the acquisition sensors that can interfere with the process of DO segmentation. Experiments were performed in different environments with the absence and presence of movement of the acquisition sensors. (C) 2019 Elsevier Ltd. All rights reserved. (AU)

• Processo FAPESP: 17/26421-3 - Investigação do uso de Sistemas Inteligentes para Mapeamento Eficiente de Aplicações em Arquiteturas Many-Core
• Beneficiário: Emerson Carlos Pedrino
• Modalidade de apoio: Bolsas no Exterior - Pesquisa