Computer Vision For Sight Computer Vision Techniques To Assist Visually Impaired People To Navigate In An Indoor Environment

This chapter focuses on computer vision techniques to assist visually impaired people to navigate in an indoor environment. First, the problem is defined in terms of tasks, sensors, devices, and the performance requirements (real-time, accuracy, and robustness). Then, a recommended paradigm is proposed to build these systems for real-world applications, which include three important components: environment modeling, localization algorithms, and user interfaces. A broad review of the recent research achievements is provided in two categories: omnidirectional image-based and three-dimensional (3D) model-based approaches. As an example, an omnidirectional-vision-based indoor localization solution is described with algorithms and corresponding implementations in maximizing the use of the visual information surrounding a user. The system includes multiple components: floor plan parsing and path planning for scene modeling, deep learning for place recognition, image indexing for initial localization, 3D vision for position refinement, and portable user interfaces. Finally, we summarize the work and present some discussions for future work.