INVys: Indoor Navigation Assistance for Visually Impaired Persons using RGB-D Camera

This research presents an indoor navigation system for visually impaired persons (termed INVys) using RGB-D camera, a camera that provides both colour and depth information. INVys uses depth camera to sense immediate environments for obstacles (micro-navigation) and utilizes colour camera to recognize optical glyph that acts as marker of particular navigational signs or places (macro-navigation). For micro-navigation, INVys proposes Auto-Adaptive Double Thresholding (AADT) method to detect obstacle, calculate its distance, and send feedback to the user to avoid the obstacle. In order to evaluate the algorithm, AADT is compared to baseline method and Auto-Adaptive Thresholding (AAT) method using 4 criteria: accuracy, precision, robustness along depth camera coverage, and execution time. AADT excels in accuracy, precision, and robustness. In regards of macro-navigation, INVys recognizes glyph using Automatic Glyph Binarization method and evaluates it using 2 criteria: accuracy and execution time. Furthermore, this study also provides correlation between sizes of the glyph, distance of the recognized glyph, tilt condition of the recognized glyph and the accuracy of glyph recognition. The correlations defines minimum glyph size that can be practically used in the context of indoor navigation for visually impaired.