Recognition Of Vietnamese Dong Based On Multispectral Images

The application of machine vision technology to the identification of banknote is a development trend of related technologies for banknote processing financial instruments.China's high-end banknote counterfeit detectors,counterfeit instruments and other equipment have been able to complete the RMB counterfeiting and crown recognition functions based on image analysis.With the increasing frequency of international trade and personnel exchanges,foreign currency identification function had been proposed for financial instruments.Vietnam is an important neighbor country of one Belt one Road.It is also the export destination of financial instruments.It is the market demand to study the image recognition of Vietnamese Dong and develop financial instruments compatible with the Vietnamese Dong is a domestic and international market demand.This thesis studies the Vietnamese Dong multispectral image recognition technology.The main work includes:Firstly,a Vietnamese Dong image recognition algorithm based on convolutional neural network is designed to achieve function of Vietnamese Dong's orientation identification,denomination and edition.The algorithm uses the downsampled image of the Vietnamese Dong image as a feature.The orientation recognition accuracy rate is 99.83%,and the denomination recognition accuracy rate is 99.65%.Compared with the traditional method,this method can identify the orientation denomination at one time,and improve the recognition accuracy and robustness.Secondly,the extraction and segmentation of the Vietnamese Dong crown characters is realized,and the crown recognition algorithm based on neural network is designed.The identification method which realizes the accurate identification of the Vietnamese Dong crown uses a binarized image of the crown.The accuracy of letter recognition reaches 99.31%,and the accuracy of digital recognition reaches 99.74%.Thirdly,according to the characteristics of the Vietnamese Dong image,a Vietnamese Dong authenticity recognition algorithm based on the characteristics of local neighborhood enhancement mode was designed.The the True Postitive Rate of this method is 99.42%,and False Positive Rate is 0.50%.Compared with the traditional method,the new method uses 10 anti-counterfeiting features,the True Postitive Rate are improved and False Positive Rate reduced.Based on the method of multispectral image analysis,this thesis realizes the recognition of Vietnamese Dong's denomination,orientation,edition,crown and authenticity.Related algorithms have been applied to foreign currency counterfeit detectors.After testing,the algorithm performance reaches the accuracy,reliability and computational performance requirements of the financial equipment to the banknote image.