Study On Computer Vision Based Navigation System For Blind

For decades evolving technology have helped researchers build electronic devices for the blind's navigation. Of all, Electronic Travel Aids(ETAs) is of the categories that trans-form information about the environment that would normally belayed through vision into a form that can be conveyed through another sensory modality. Mostly, information can be gathered through laser scanners, sonar, or cameras, the blind are informed through the auditory or tactile sense.In this thesis, we analysis the aims and meanings of studies on electronic device for navigation, briefly describe advantages and disadvantages of the existing navigation de-vices, then present our novel electronic technological solution to computer vision based navigation system for the blind. The novel navigation solution is analyzed from system, hardware, and software perspectives. Later we provide a maturity analysis for the main operational features of the navigation system, introduce a robust method to dynamically estimate the geometrical homography relation and planar regions. The validity of the pro-posed method is shown by a number of sequential images in general environments.This thesis merits two aspects. Firstly, we proposed a novel design framework for ma-chine vision based navigation system for the blind. Most of existing blindness navigation system is for route planning and obstacle detection using laser, sonar sensor devices. The navigation system we proposed in this thesis, which aims to detect safe walking areas, is affordable, small and convenient. Secondly, we present a novel road region extraction algo-rithm that dynamically estimation the geometrical homography relations and planar areas in the view. As extraction of safe walking area is the fundamental technological solution for mobility problem blind people face, testing results in general environments demonstrate the algorithm's effective and feasible.