A Study of Computer Vision Techniques for Currency Recognition on Mobile Phone for the Visually Impaired

Currency recognition is a challenging task in computer vision. Various Image processing methods and artificial intelligence techniques have been employed to handle this. We have studied various currency recognition applications employing different keypoint detection, keypoint extraction and keypoint matching approaches. In this paper we also present a study of some recent computer vision techniques like BRIEF, ORB, FAST, AGAST, BRISK, FREAK and earlier techniques like SURF, SIFT. The mobile based currency recognition applications discussed here are meant to support visually impaired and blind users. Mobile devices are ubiquitous and come with a built in camera having a fair resolution. Currency note images captured by blind users can have several failings: images of the notes may not be ideally aligned/oriented; there can be scale changes due to variation of distance from camera; illumination changes can also lead to differences in images. There is also the possibility of a cluttered background in the image, or the note being partially occluded, folded, worn and/or wrinkled, etc. This paper illustrates computer vision techniques employed till now and studies new improved techniques which can be used in their place. These new methods offer several advantages for efficient currency note recognition in mobile applications. We have also studied blind users’ needs and expectations from such applications and how researchers have developed variety of touch screen utilities like Braille keyboard, food product identification, and indoor navigation applications. Also, our in person visit to an NGO for blind helped us to understand their special needs from touch screen mobile devices. This visit has really motivated and has enriched our work in correct way. Keywords— Currency note recognition, Computer Vision, Visually impaired blind users, Assistive mobile phone applications-FREAK, BRISK, SURF, SIFT