On intelligent image processing methodologies applied to navigation assistance for visually impaired

• Auteur principal: G.Sainarayanan
• Format: Thèse
• Langue: anglais; anglais
• Publié: 2002
• Sujets: TK8300-8360 Photoelectronic devices (General)
• Accès en ligne: https://eprints.ums.edu.my/id/eprint/10206/1/24%20PAGES.pdf; https://eprints.ums.edu.my/id/eprint/10206/2/FULLTEXT.pdf

The main objective of this thesis is to develop a computer based Navigation Assistance for Visually Impaired (NAVI) as a vision substitutive system. The hardware of the system includes a Vision sensor mounted on a headgear, a set of Stereo earphones and a Single Board Processing System (SBPS) with batteries, duly placed in a vest. The Vision sensor is a digital video camera. The video camera captures the image of the environment. The captured image is processed, mapped on to specially structured stereo sound patterns and sent to the earphones. A set of image processing requirements for vision substitution is identified and incorporated in the NAVI system. The image processing, developed in this thesis, is designed to work as a model of the human vision system. To model the human vision system in image processing, two properties of human eye, namely lateral inhibition and domination of the object properties rather than background are incorporated. The image processing methodologies applied in NAVI are developed using artificial intelligent techniques. The property of lateral inhibition is incorporated using neural network based Canny edge filter. In vision substitutive system, definition of objects and background is not easy as compared to industrial object recognition system. Therefore, three methods for object enhancement and background suppression are proposed in NAVI using fuzzy logic and neural network. The edge image and the object enhanced image with background suppressed are integrated to produce a resultant image. The resultant image is sonified to produce stereo acoustic patterns. Blind volunteers were trained with the developed NAVI system and they were tested to identify the environment. They were able to understand the logic behind the sound in discriminating the object from background. It was also verified that the discrimination of objects by the blind through the proposed image processing methodologies is effective and easier than that of earlier efforts in this direction.