Visual localization aid for the blind

Abstract

This project describes a computer-based location determination system for helping blind or visual impaired persons to navigate easily in an indoor environment. Unlike the existing systems, the proposed system requires only a light weight personal computer and a small camera fixed on their walking cane. The camera captures images from the place where the blind exists and sends it to the computer where it will be analyzed. As a response, a voice message is played to the person explaining where she/he is. The method adopted here is based on a Content-Based Image Retrieval (CBIR) approach which involves retrieving images from a large image database that are similar to a query image based on features taken from the content of the image. Image-based features are extracted using the well known SIFT algorithm. These extracted features are used as the basis for the similarity check between images. This approach is combined with Markov localization to enhance the localization rate. Experimental results in a public indoor environment show that the proposed approach can lead to a high localization rate