Conscious GPS: A System to Aid the Visually Impaired to Navigate Public Transportation

Abstract

Using public transportation can be a major challenge for individuals with visual impairments. To navigate a mass transit system independently and safely, one needs to be able to gather information about their surroundings. We propose a Conscious GPS system to assist a user with visual impairment to navigate the public transit system. The system utilizes computer vision to autonomously record the users surroundings by taking various pictures in sequence and leveraging image processing techniques and machine learning to identify and extract information about key objects around the user. Our target object is the circular bus sign on an NYC MTA bus stop. We use Haar Cascade computer vision techniques to detect and locate the sign in an image through feature-based image classification. The proposed system experiences no misclassification after training the Haar Cascade classifier with a minimum hit rate of 0.99 and a maximum false alarm rate of 0.20 when navigating the daily city landscape and can accurately detect a bus stop sign.