Dewanto Harjunowibowo, Sri Hartati, Aris Budianto
Publication year: 2015

This research is aimed to test a paper currency counterfeit detection system based on Linear Vector Quantization (LVQ) Neural Network. The input image of the system is the dancer object image of paper currency Rp. 50.000,- fluorescent by ultraviolet light. The image of paper currency data was taken from conventional banks. The LVQ method is used to recognize whether the paper currency being tested is counterfeit or not. The coding was carried out using visual programming language. The feature size of the dancer tested object is 114×90 px and the RGBHSI was extracted as the input for LVQ. The experimental results show that the system has an accuracy 100% of detecting 20 real test case data, and 96% of detecting 22 simulated test case data. The simulated case data was generated by varying the brightness of the image data. The real test case data contains of 10 counterfeit paper currency and 10 original paper currency. The simulated case data contains of 11 original paper currency and 11 counterfeit paper currency. The best setting for the system is Learning Rate = 0.01 and MaxEpoh = 10.