Study On Multi-Information Integration Based Guiding System For Blind

This thesis proposed a multi-information integration based guiding system for blind, which is able to provide reliable out-door environment sensing. The fundamental functions of the guiding system are walkable region extraction and obstacle detection. A homography based method is used for walkable region extraction. This method requires a binocular stereo system as input and utilizes the homography relation between two images of plane road surface captured by two cameras to extract road region. This method is wildly used in road extraction area.This thesis innovatively proposed a modified homography matrix estimation method which is essential part of road extraction method, and an additional depth camera is equipped as an auxiliary device for homography matrix estimation. The traditional method of homography matrix estimation takes advantage of a set of matched features in stereo images. Features may be un-sufficient or lie in close region in some cases, which may cause failure to road extraction. The modified method makes use of road plane parameters to estimate homography matrix, which avoids the disadvantages of traditional method and provides better stability.Moreover, this thesis proposed a touch-based feedback device which is able to demonstrate environment sensing results. The front panel of the device is composed of2D dot matrix, each dot has2states:up or down. Environment sensing results are transferred into dot-map and demonstrated by the feedback device.