**Automatic Coin Classification and Identification**

Reinhold Huber-Mörk1, Michael Nölle1, Michael Rubik1, Michael Hödlmoser2, Martin Kampel2 and Sebastian Zambanini2 *1Department Safety and Security, Austrian Institute of Technology 2Computer Vision Lab, Vienna University of Technology Austria* 

## **1. Introduction**

We investigate object recognition and classification in a setting with a large number of classes as well as recognition and identification of individual objects of high similarity. Real-world data sets were obtained for the classification and identification tasks. The considered classification task is the discrimination of modern coins into several hundreds of different classes. Identification is investigated for hand-made ancient coins. Intra-class variance due to wear and abrasion vs. small inter-class variance makes the classification of modern coins challenging. For ancient coins the intra-class variance makes the identification task possible, as the appearance of individual hand-struck coins is unique. Figure 1 shows sample images for the considered collections of coins.

(a) Modern coins (b) Ancient coins

Fig. 1. Examples of images of modern and ancient coins

Modern coins were acquired by a high-speed machine vision system for coin sorting described in detail by Fürst et al. (2003). For ancient coins the setting is more general, images acquired

This contribution is organized as follows. Section 2 reviews the state of the art in automated coin image analysis. Section 3 describes coin detection and invariant preprocessing and Section 4 discusses matching based on various feature descriptors. Classification, identification and information fusion is described in Section 5. Results are presented in Section

Automatic Coin Classification and Identification 129

For modern coins, i.e. machine struck coins, judging systems using electromechanical devices are wide-spread. Those systems are commonly based on measuring weight, diameter, thickness, permeability and conductivity (Davidsson, 1996), oscillating electromagnetic field characteristics (Neubarth et al., 1998), and photo- and piezoelectric properties (Shah et al., 1986). Typically, such systems are only capable to discriminate a small number of different

Approaches towards classification of modern coins using image processing are described in various papers and patents. A neural network approach capable of discriminating between 500 Won and 500 Yen coins was published by Fukumi et al. (1992). A number of coin authentification methods employing optical means are described in patents, e.g. a system by which both sides of a coin are first imaged by cameras, followed by feature extraction from binarized images, and finally combined with a magnetic sensor measurement is described by Hibari & Arikawa (2001). The so called Dagobert coin recognition system was developed for high volumes of coins and a large number of currencies (Fürst et al., 2003; Nölle et al., 2003). Image binarization followed by area measurement and comparison of coin center and center of gravity was also suggested in a patent (Onodera & M., 2002). Another system based on the analysis of one side of a coin by transformation of its image into polar coordinates and matching of profiles taken along angle direction was described by Tsuji & Takahashi (1997). A special acquisition device for coins employing colored illumination from various angles was suggested by Hoßfeld et al. (2006). Methods based on matching gradient directions (Reisert et al., 2006; 2007) and color, shape and wavelet features (Vassilas & Skourlas, 2006) were suggested. An approach based on multiple Eigenspaces aims at classification for a large number of classes (Huber et al., 2005). This approach initially obtains a translationally and rotationally invariant description and secondly an illumination-invariant Eigenspace is selected from multiple Eigenspaces (Leonardis et al., 2002). Finally probabilities for coin classes are derived for the obverse and reverse sides of

For ancient coins, i.e. hand struck coins, some publications discussing approaches for classification appeared. Early approaches, which achieved a moderate classification performance, were based on matching of contour and texture features (Van Der Maaten & Postma, 2006) or make use of interest point extraction and matching of local features (Zaharieva et al., 2007). More recently, an approach based on interest points and improved feature description and matching was reported (Arandjelovi´c, 2010). The inherent properties of hand struck coins result in individual features of each coin and a large intra-class variance. Therefore, object classification becomes challenging. However, in contrast to object classification, object identification relies on those unique features which distinguish a given object from all other members of the same class. Results on identification of ancient coins were

6 and conclusions are drawn in Section 7.

**2. State of the art in coin image analysis**

each coin and Bayesian fusion is performed.

coin denominations and are mostly limited to a specific currency.

by scanner and camera devices are considered. We will also discuss the use of a 3D acquisition device and 3D models for ancient coins (Zambanini et al., 2009).

The initial step of object recognition will be discussed as the problem of detection. i.e. foreground-background segmentation. Background knowledge on coins, i.e. the circular shape, which holds for most modern coins and is approximately true for ancient coins, is exploited in suggested segmentation methods (Zambanini & Kampel, 2009). Invariance with respect to translation is solved by segmentation, scaling is covered by normalization and rotation is handled in the suggested methods for invariant description, classification and identification.

We will compare two approaches for classification of coins, a method based on matching edge features in polar coordinates representation (Nölle et al., 2003) and a method for matching based on an Eigenspace representation (Huber et al., 2005). Discussion on the influence of dirt and abrasion will be included. Classification of modern coins makes additional use of geometric measurements and information extracted from obverse and reverse side of the coins. Incorporation of geometrical measurements and fusion of coin sides is realized by preselection and Bayesian fusion. In order to limit the number of coin classes to discriminate the concept of multiple Eigenspaces (Leonardis et al., 2002) is applied in the Eigenspace framework. Rejection of unknown coins and a discussion on false classification and false acceptance rates vs. false rejection is included.

For identification of coins we will consider an approach based on shape features describing the edge of an ancient coin (Huber-Mörk et al., 2008; Zaharieva et al., 2007). Features are derived from the Fourier domain representation of the coin contour. Comparison of two coins is done by matching the features derived from contour representations. Bayesian fusion of coin sides is studied. In order to discuss the identification performance a discussion on precision vs. recall is included (Huber-Mörk et al., 2010). Improvement by 3D modeling and analysis is also presented (Hödlmoser et al., 2010).

Results are presented for all considered data sets and methods. The data set for classification of coins consisted of approximately 12 000 coins with images of reverse and obverse sides. The data set contained 932 different coin classes. A derived data set was made publicly available as a benchmark by the EU MUSCLE network of excellence (Nölle & Hanbury, 2006; Nölle et al., 2006). Depending on the acceptable rejection rate correct decisions are taken in more than 92% for the Eigenspace approach. With the edge matching method approximately 86% of the coins are either correctly classified or correctly rejected. Considering only valid coins, i.e. coins in the database of 932 coins, the Eigenspace approach achieved a correct classification rate of 94.58%, whereas the direct edge matching approach scored 84.79%. Correct rejection of invalid coins was obtained at a rate of 78.45% for the Eigenspace approach and 98,29% were achieved in the direct edge matching approach.

The data set for identification of ancient coins was provided by the Fitzwilliam Museum, Cambridge, UK and was made publicly available by the EU COINS project (Kampel et al., 2009). The data set consists of 240 coins of the same class with 1200 images of obverse sides and 1200 images of reverse sides which were acquired by different acquisition devices. Results for identification based on shape matching are on the order of magnitude of 98%.

2 Will-be-set-by-IN-TECH

by scanner and camera devices are considered. We will also discuss the use of a 3D acquisition

The initial step of object recognition will be discussed as the problem of detection. i.e. foreground-background segmentation. Background knowledge on coins, i.e. the circular shape, which holds for most modern coins and is approximately true for ancient coins, is exploited in suggested segmentation methods (Zambanini & Kampel, 2009). Invariance with respect to translation is solved by segmentation, scaling is covered by normalization and rotation is handled in the suggested methods for invariant description, classification and

We will compare two approaches for classification of coins, a method based on matching edge features in polar coordinates representation (Nölle et al., 2003) and a method for matching based on an Eigenspace representation (Huber et al., 2005). Discussion on the influence of dirt and abrasion will be included. Classification of modern coins makes additional use of geometric measurements and information extracted from obverse and reverse side of the coins. Incorporation of geometrical measurements and fusion of coin sides is realized by preselection and Bayesian fusion. In order to limit the number of coin classes to discriminate the concept of multiple Eigenspaces (Leonardis et al., 2002) is applied in the Eigenspace framework. Rejection of unknown coins and a discussion on false classification and false

For identification of coins we will consider an approach based on shape features describing the edge of an ancient coin (Huber-Mörk et al., 2008; Zaharieva et al., 2007). Features are derived from the Fourier domain representation of the coin contour. Comparison of two coins is done by matching the features derived from contour representations. Bayesian fusion of coin sides is studied. In order to discuss the identification performance a discussion on precision vs. recall is included (Huber-Mörk et al., 2010). Improvement by 3D modeling and analysis is

Results are presented for all considered data sets and methods. The data set for classification of coins consisted of approximately 12 000 coins with images of reverse and obverse sides. The data set contained 932 different coin classes. A derived data set was made publicly available as a benchmark by the EU MUSCLE network of excellence (Nölle & Hanbury, 2006; Nölle et al., 2006). Depending on the acceptable rejection rate correct decisions are taken in more than 92% for the Eigenspace approach. With the edge matching method approximately 86% of the coins are either correctly classified or correctly rejected. Considering only valid coins, i.e. coins in the database of 932 coins, the Eigenspace approach achieved a correct classification rate of 94.58%, whereas the direct edge matching approach scored 84.79%. Correct rejection of invalid coins was obtained at a rate of 78.45% for the Eigenspace approach and 98,29% were

The data set for identification of ancient coins was provided by the Fitzwilliam Museum, Cambridge, UK and was made publicly available by the EU COINS project (Kampel et al., 2009). The data set consists of 240 coins of the same class with 1200 images of obverse sides and 1200 images of reverse sides which were acquired by different acquisition devices. Results

for identification based on shape matching are on the order of magnitude of 98%.

device and 3D models for ancient coins (Zambanini et al., 2009).

acceptance rates vs. false rejection is included.

also presented (Hödlmoser et al., 2010).

achieved in the direct edge matching approach.

identification.

This contribution is organized as follows. Section 2 reviews the state of the art in automated coin image analysis. Section 3 describes coin detection and invariant preprocessing and Section 4 discusses matching based on various feature descriptors. Classification, identification and information fusion is described in Section 5. Results are presented in Section 6 and conclusions are drawn in Section 7.
