Euro Coin Recognition Based On Computer Vision

Coins play a very important role in daily life and are widely used in automatic ticket vending machines,vending machines,and other equipments.Coin sorting is an important part of its daily circulation management.In the European Union region,euro coins are widely used in various countries,and euro coins are characterized by a uniform figure of the front,while the back of the coin is designed by each issuing country,which results in the large difference among the back patterns of coins issued by different countries.In the view of this feature,we can identify the country to which the coin belongs through the identification of the back pattern of the coin.At the same time,in the process of coin circulation,because the wear will cause the appearance quality decline,so the coins with serious appearance wear need to be recycled.Through the evaluation of the appearance quality of the coin,it is the usual method to judge whether the coin should be recycled or not.One of the factors for evaluating appearance quality is the issue year of coins.Therefore,if we can detect and recognize the issue year of the coin,which will be conducive to the precise classification of the clearance operation.Because of the time-consuming and labor-intensive task of manually completing the euro coin's country identification and the issue year,to solve above problems,this paper studies the euro coin's country identification and year detection problem based on computer vision inspection system.The paper mainly implements the following works:(1)Aiming at the characteristics of the euro coin belonging to circular image and present central symmetry characteristic,an anti-rotation high-efficiency and high-discrimination binary pattern feature extraction method based on the spatial symmetry position description is proposed,which reconstructs the local coordinate system through radial transformation during feature calculation,and based on it,the local binary pattern extraction with spatially symmetric region with anti-rotation transformation is realized.Meanwhile,the pool operation adopts the annular space division with rotation invariance,thus ensuring the anti-rotation transformation capability of the final feature description.The recognition accuracy of the proposed method is close to 100%in euro coin data set,which is superior to the traditional LBP and HOG features,moreover,the proposed method is efficient and requires only 0.045 ms for single point computation.(2)Considering the particularity of the year detection and recognition of euro coins,this paper presents a digital detection method based on Faster-RCNN model,and the year-sort algorithm based on clustering and priori rule.Through training data augmentation processing,the retrained Faster-RCNN network model can adapt to the various pose and size changes of the digital in the coin,then by using K-means clustering algorithm,the obtained digital candidate boxes can be grouped into 4 categories,and the most confident candidate boxes are selected in each category,finally,according to the pre-determined year arrangement pattern of different country coins,the correct year information can be obtained by proper sorting algorithm.The experimental results show that the detection accuracy of the method is 89.62%,and the calculation time is about 215ms,which basically meets the requirements of accuracy and real-time.(3)The computer vision detection experiment system for euro coins was designed,which can realize the country identification and year detection of euro coins as well as meets the requirements of principle verification and algorithm verification.The experimental system includes detection scheme design,hardware equipments selection and image detection system program.Our research has an impetus to the technical progress of the circulation management of the euro coins,in which the algorithm aiming at studying the characteristics of the euro coins,which not only solves the problem of the detection and identification of the euro coins,but also has reference value for the identification of other circular image and digit recognition in specific situation.