Due to the proliferation of fake money these days, detection of counterfeit coins with high accuracy is in strong demand, yet not much research has been conducted in this field. The objective of this thesis is to introduce modern computer vision techniques and machine intelligence to differentiate real coins and fake ones with high precision, based on visual aspects.
To that end, a high-resolution scanning device – IBIX Trax is deployed to sample the coin images. On top of that, three visual aspects are thoroughly inspected, namely lettering, images and texture.
Six features are extracted from letterings, i.e. stroke width, contour smoothness, lettering height, lettering width, relative angle, and relative distance. As for classification, a hierarchical clustering – max spacing K-clustering—is adopted. Our experimental results show that the fake coins and real ones are totally separable based on these features. 
As for images, we propose a novel shape feature— angle-distance. After images are segmented, a vector of size 360*1 is deployed to represent each shape. For classification, a dissimilarity measurement is used to quantize the difference between two shapes. The results show it can recognize the fake coins successfully.
As for texture, a cutting-edge feature maximum stable extremal region is adopted to automatically detect the holes and indents on the coin surface. Parameters associated with this feature are adjusted in the experiments. The detection results show this feature can be used as an indicator for assessing the qualities of coins.