Pocket Change: Results 1.1 2006/12/18 23:06:41.551 US/Central 2006/12/19 01:55:55.488 US/Central Tyler James Barth tbar@rice.edu Aaron David Cottle acottle@rice.edu John P Stallcup john.stallcup@rice.edu Christopher J. Vaucher cvaucher@rice.edu Tyler James Barth tbar@rice.edu Aaron David Cottle acottle@rice.edu John P Stallcup john.stallcup@rice.edu Christopher J. Vaucher cvaucher@rice.edu coin recognition machine vision pocket change signal processing This module describes the results from testing the Pocket Change coin recognition system. We are very pleased with the results of our project. Passing in a composite image where the coins are identical to those in the database but are translated and rotated provides us with a perfect match for every coin. Where the accuracy starts to suffer is when we have to deal with the issues brought on by real world conditions. We ran a series of tests with various numbers of coins to determine an average accuracy rating of the system. Generally, it seems to be approximately 75% accurate.

Pocket Change Results Accuracy of coin recognition across multiple trials. Y-axis values are percentages.

These results are promising. While these do not match the perfect accuracy rates of the composite images, we do approach 90% accuracy in some tests. However, it is important to note that we ran the tests using various amounts of coins to test runtime. The number of coins used should not impact accuracy.