Bank note simulation

摘要

In order to train a bank note classifier such as for example present in a cash handling device, a plurality of electronic representations of bank notes are needed. To provide such representations in a fast and cost effective way, a reference bank note (1) is scanned (5) and a reference dataset (2) is generated. In a step defect simulation (6), a plurality of electronic datasets (4) is generated by taking a particular dataset (2) either from the scanner or from a reference library (3) and simulating one or several defects of the bank note such as for example dog-ears, folds, missing parts, soil, abrasion and/or many more. In the simulation step (6) a priori knowledge (8) about the different defects and/or the reference bank note may be considered. The electronic datasets (4) which represent virtual defective bank notes are stored in a database (7). A classifier may then be trained with these virtual defective bank notes.