Fuzzy clustering in vision recognition applied in NAVI

Abstract

In this paper a system for navigation assistance for visually impaired (NAVI) is presented. The system includes a single board processing system (UPS), vision sensor mounted headgear and stereo earphones. The image of environment is captured by the vision sensor. The image is then processed by a novel real time image processing methodology using the fuzzy clustering algorithm. The proposed methodology incorporates certain human vision properties for clear representation of the environment. The image processed is mapped to a stereo acoustic pattern and transferred to the user's earphones. The sound produced varies with the gray value and orientation of the obstacle in front. Blind individuals are trained with NAVI and tested for obstacle identification. Suggestions from the blind volunteers regarding pleasantness and discrimination of sound pattern were also incorporated in the prototype. Certain improvements for faster convergence of clustering are also considered.