Date of Award

2015

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical Engineering

First Advisor

Dr. Corey A. Graves

Abstract

Connect-the-Dots puzzles are puzzles which contain labeled dots in a sequence. These puzzles are mostly designed as a way for children to hone in on their counting skills, while having fun. These same puzzles, which are available in abundance online and with modification, can be used to aid students in other areas of education such as spelling. Research shows that the addition of visual imagery provides a significant impact in spelling performance. The objective of this research is to develop an algorithm for processing Connect-the-Dots puzzles to assist in the replacement of the original numbers in the puzzle with characters that will help to facilitate an alternative educational purpose. In particular, the use of Optical Character Recognition (OCR) and image processing algorithms to process pre-existing Connect-the-Dots puzzles is explored. An algorithm was developed to locate and identify the numbers in the puzzles. The system is comprised of five components, namely, an Image Preprocessing component, a Dot Locator component, a Number Locator component, a Number Recognition component, and a Post Processing component. To test the accuracy of the algorithm an experiment was conducted using 20 hand selected puzzles from an online source. The accuracy of the algorithm was evaluated, component by component, as well as overall, by visually capturing the make-up of the puzzles and comparing them to the results generated by the algorithm. Results show that the algorithm performed at an overall accuracy rate of 66%. However, the Dot Locator component performed at a rate of 100%, the Number Locator at a rate of 86%, and the Number Recognition at a rate of 76%. This research will aid in the development of an application that may provide educational benefits to children who are exposed to using technology for learning, at a young age.

Share

COinS