Crosswalk Localization from Low Resolution Satellite Images to Assist Visually Impaired People

In this article, we propose a model for crosswalk detection and localization by using satellite images captured from Google Maps, for the purpose of assisting visually impaired people. The detection is performed by an SVM classifier, which is combined with Google Road Map to speed up computation time and to eliminate some possible false alarms. We assume that a visually impaired person holds a smartphone with an embedded GPS, which is used to initialize the extraction of images from Google Maps, as well as to assist its user by providing audio feedback of the nearest detected crosswalk. This issue brings forward significant interest and it is also very challenging, mainly due to illumination changes, occlusion, image noise and resolution, besides the quality of crosswalks that sometimes are badly painted in many developing countries. Experimental results indicate that the proposed model works well in low resolution images, effectively detecting and localizing crosswalks in simulated scenarios.