Development of an automated coin grader

This paper introduces a computer-operated, rare coin grading system. In the numismatic community today the official grading of coins is accomplished by human inspection. This, by its very nature, lends itself to inconsistent results that can misrepresent the actual value of a coin by thousands of dollars, and the human element can further misrepresent for personal gain. Therefore, an automated computer grading system should be a welcome unbiased alternative. To demonstrate its potential utility in this application we are developing a system that grades, appraises, and authenticates rare coins by comparing them to pre-appraised coin values stored in a database. As a first step our Web interface is structured as an educational quiz environment. Digitally scanned coin color images are represented by three histograms (hue, saturation and brightness) and one method (edge detection). Using histogram class methods, the histograms are constructed, manipulated and visualized. The outcome is a highly accurate and consistent dataset that converts the measurements of the coin features extracted into the form of a grade. Although these four coin features alone may not suffice to optimally determine a final grade, they are an important first step towards a potentially accurate and valuable system