Coin recognition based on texture classification on ring and fan areas of the coin image

Abstract

In this manuscript, a coin recognition system based on texture classification of the coin image is proposed. The proposed system consists of the training phase and the testing phase. In the training phase, the proposed system first identify the coin position in the image based on the generalized Hough transform. When the position is identified, the proposed system extract texture features of the coin image and uses these features to generate the dictionaries based on the bag of word (BoW) approach. On extracting coin texture features, the proposed system divides the image into ring and fan structures to take advantages of the rotation invariant property provided by the ring structure, and the distinguish capability on similar textures provided by the fan structure. The textures features extracted from the coin images are further quantized based on the dictionary and a histogram is generated to represent the texture features of the coin. Finally the proposed system use support vector machine to train a model for each class of the coin. In the testing phase, the test coin image is classified based on the models obtained in the training phase. Experimental results show that the proposed method provides higher accuracy of recognition rate when compared with the related studies.