Symbol Detection in Low-Resolution Images using a Novel Method

Abstract

© 2014 SERSC.The ability to detect taxis, pedestrian walk symbols, zebra crossings, and straight line tactile guide paths and junctions is essential for blind people in any city. This work applies a novel method that detects automatically these five targets via their colors, geometric elements and structural features. To reduce processing time, this method is tailored to low-resolution images that are roughly 80 x 60 pixels. Additionally, another novel method based on the Hough Transform can recognize straight lines, an important geometric element, in low-resolution images. The effectiveness and efficiency of the proposed method are validated experimentally.