Efficient coin recognition using a statistical approach

In this paper we propose a coin recognition system using a statistical approach and apply it to the recognition of Jordanian coins. The proposed method depends on two features in the recognition process: the color of the coin, and its area. The recognition process consists of several steps. Firstly, a gray-level image is extracted from the original colored image. The image is then segmented into two regions, coin and background, based on the histogram drawn from the gray-level image. To reduce the noise, the segmented image is then cleaned by opening and closing through several erosion and dilation operations. After that, four parameters are calculated, the area, the average red, blue, and green colors of the coin to be recognized. Based on these parameters, the decision to which category the coin belongs is obtained. The results provided illustrate that the proposed approach is both simple and accurate. Although the proposed recognition approach is applied to Jordanian coins, it can be applied to the recognition of any coins.