Recognition Of Indian Currency Based On Multi-spectral Image

Banknote identification and anti-counterfeiting are related to the financial security of a country and the stability of the market economy.In recent years,banknote recognition and anti-counterfeiting based on multi-spectral images have become a research hotspot and have attracted much attention.At present,the Indian market is flooded with counterfeit currency,lacking mature Indian currency identification and anti-counterfeiting technology and tools.How to analyze the multi-spectral image of Indian currency to identify its version,denomination,crown size and identify the authenticity is a technical problem to be solved.Domestic research on the identification of multi-spectral Indian currency is still in its infancy,and traditional banknote recognition algorithms are difficult to apply to Indian currency.The difficulties faced by Indian currency identification mainly include: 1)printing is not standardized,there are deviations in banknotes;2)the size of banknotes of different denominations may be the same,and the image features are similar and difficult to distinguish;3)with a lot of versions and a variety of counterfeit currency types,it's very difficult to identify counterfeit currency;4)India is still a country dominated by cash transactions,which in the circulation of banknotes may cause serious wear,dilapidation,incompleteness,smearing,etc.,which makes the identification and authentication of banknotes more difficult.In view of the above problems,the following research work is carried out: Firstly,through the research and analysis of the characteristics of the Indian currency image,this thesis proposes an improved LBP feature combined with a random forest method to carry out the orientation and version identification of the Indian currency,and solves the problem that the Indian currency version is diverse and the features are similarly indistinguishable.Secondly,the authenticity features of Indian currency in multiple spectra are studied.The local feature classifier and global HOG feature classifier are proposed to solve the problem of low accuracy based on traditional local feature recognition.Finally,the characteristics of the Indian currency crown are studied,and a novel solution is proposed for the complex regional background.The optimized neural network is used to identify the crown number,which solves the problem that the traditional recognition method is not accurate and is insufficient generalization.The algorithm proposed in this thesis has been successfully applied in the money counter equipment.A large number of experimental results show that compared with the traditional method,the proposed method can achieve faster and more accurate recognition effect,and has stronger robustness and generalization.