Date of Award

5-2013

Degree Type

Report

Degree Name

Master of Science (MS)

Department

Computer Science

Committee Chair(s)

Vladimir A. Kulyukin

Committee

Vladimir A. Kulyukin

Committee

Daniel W. Watson

Committee

Nicholas Flann

Abstract

Many Visually Impaired individuals are managing their daily activities with the help of smartphones. While there are many vision-based mobile applications to identify products, there is a relative dearth of applications for extracting useful nutrition information. In this report, we study the performance of existing OCR systems available for the Android platform, and choose the best to extract the nutrition facts information from U.S grocery store packages. We then provide approaches to improve the results of text strings produced by the Tesseract OCR engine on image segments of nutrition tables automatically extracted by an Android 2.3.6 smartphone application using real-time video streams of grocery products. We also present an algorithm, called Skip Trie Matching (STM), for real-time OCR output error correction on smartphones. The algorithm’s performance is compared with Apache Lucene’s spell checker. Our evaluation indicates that the average run time of the STM algorithm is lower than Lucene’s. (68 pages)

Share

COinS