Bank note clear classification algorithm version determination method and device, and storage equipment

Abstract

The invention discloses a bank note clear classification algorithm version determination method comprising the following steps: building bank note clear classification models used for determining banknote clear classification degrees in financial equipment, wherein each bank note clear classification model comprises feature variables, each bank note clear classification model is formed in multiple numbers, and the feature variable value of each bank note clear classification model is different; respectively counting real work indexes of the same bank note clear classification model under different feature variable values; comparing the real work index with a preset work index, and keeping the feature variable optimal value of the bank note clear classification model when the real work index is smaller than the preset work index. Aiming at each algorithm needing to adjust the parameter or threshold corresponding to the feature variable in the financial equipment, a plurality of algorithm models with different feature variable values are formed in the engineering; compared with the laboratory environment, the method is sufficient and complete in bank note samples, thus easily determining the optimal bank note clear classification algorithm version.

Description

technical field [0001] The invention relates to the technical field of banknote identification, in particular to a method, device and storage device for determining the version of a banknote sorting algorithm. Background technique [0002] Banknote recognition technology is an important application of machine learning and pattern recognition technology. As a relatively independent module, banknote image recognition system has been widely used in ATMs, unmanned vending machines, and automatic ticket vending machines. [0003] In current financial machines, each machine uses a series of algorithms to identify banknotes, and each algorithm pre-determines fixed parameters or thresholds after simulating real-life transactions in a laboratory environment, so as to ensure that after the machine is put into operation, it will be used in the actual banknotes. The application can bring accurate recognition effect. [0004] However, banknote recognition technology often involves a lot...

AI Technical Summary

Problems solved by technology

[0004] However, banknote recognition technology often involves a lot of threshold adjustments. Since the laboratory environment is often inconsistent with the production environment, too strict or too loose thresholds will have an adverse impact on customer experience: too strict will lead to the rejection rate of end users High, too loose will cause bank customers to receive a large number of unqualified old and damaged banknotes, etc.

However, in the production environment, sometimes the banknotes are relatively new, and sometimes the quality is relatively poor, and the old and new grades of banknotes fluctuate greatly.

In the laboratory environment, due to insufficient banknote samples and it is impossible to cover all possible sample situations, there is no guarantee that the parameters or thresholds are optimal

Images