VISORV: Board reading, getting directions through Marker Detection for partially visually impaired people

Abstract

Augmented reality brings reality into virtual world experience, it's one of the fastest emerging and challenging field experienced. In this paper, we proposed a model VISORV which helps partially visually impaired people to easily navigate through the streets by reading aloud sign (direction) boards usually placed on the streets using a marker detection technique. The proposed algorithm is divided into three phases: Marker Detection Algorithm (MDA), Marker Identification Algorithm (MIA) and Audio formation. The Marker Detection phase involves detecting a square marker by calculating angle for each pixel and applying edge dwindling algorithm (EDA). In the second phase, marker already being detected in the first phase is identified by efficiently matching it against the markers stored in the database. The Matched marker fetches the direction along with their destinations from the database. In the final phase, guidance is being provided to the end user in the form of audio, stored in a database corresponding to marker ID identified in second phase. Various models adopted in this work are: RGB to Gray scale conversion, Shrinking, Expansion, Gray Scale to Binary Conversion, Run length encoding algorithm. Finally the model is simulated for test markers with different patterns stored and validate the proposed design.