An object finder for the visually impaired

Abstract

Technology is changing rapidly nowadays and is bringing radical changes and opportunities to our society. Computer and image capturing device such as web camera are becoming more common in our daily life. However, there are still some of the population who are not able to fully utilize the technology such as the disadvantage people who are visually impaired. In fact, technology does able to help these disadvantages people in coping with the difficulties that they are facing every day.

In this project, the author aims to develop a Human Computer Interface using computer vision technology to assist the visually impaired to grab objects that are visible within the web camera. The end result is to locate the user’s hand and provide the direction for grabbing the colour object.

This interface is programmed under .NET framework using C++ language together with open source library OPENCV. Several computer vision techniques are used such as background subtraction, template matching, edge detection and feature point detection. To recognize the human finger, multiple approaches have attempted such as enhancing the original background subtraction provided by open source, as well as improving the feature point detection algorithm used by previous researcher.

Results showed that the system could perform remarkably well on hand detection with assumptions that the user is to hold the camera firmly, environment light intensity is good and no other moving objects appear in the scene. Experiment has been done on the colour detection system and the result turned out to be as expectedly good.

At the later part of the report, the author will discuss on the potential and future work of this research work. Some improvements for this project are also suggested.