Braille Block Recognition by Analyzing Autocorrelation Patterns

Abstract

We propose a method for detecting Braille blocks with dot-pattern from images, aimed at assisting the mobility of visually impaired individuals. Braille blocks, installed on sidewalks and pathways, consist of circular protrusions (dots) arranged periodically. By detecting the dots in an image and assessing whether they are regularly spaced, it seems possible to selectively detect dot-pattern blocks. However, accurately detecting each dot from an image is challenging due to the diversity in the material, size, and three-dimensional shape of the dots. In this study, we utilize these characteristics to detect dots on the autocorrelation image and assess whether they are arranged in a grid pattern to determine if they are dot-pattern blocks. Experimental results using real images confirmed that the proposed method is effective in recognizing various types of dotted blocks.