Novel Feature Extraction and Representation for Currency Classification
In an era marked by the rapidly growing levels of international trade and tourism, the accurate recognition of various currency notes has become a necessity. This paper presents research on an image processing technique for classifying the origin of currencies. Individuals are hardly distinguishing...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
Semarak Ilmu Publishing
2023
|
| Subjects: | |
| Online Access: | http://ir.unimas.my/id/eprint/44811/1/Novel%20Feature.pdf http://ir.unimas.my/id/eprint/44811/ https://semarakilmu.com.my/journals/index.php/applied_sciences_eng_tech/article/view/3338 https://doi.org/10.37934/araset.33.1.275284 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In an era marked by the rapidly growing levels of international trade and tourism, the accurate recognition of various currency notes has become a necessity. This paper
presents research on an image processing technique for classifying the origin of currencies. Individuals are hardly distinguishing between different currencies from
various countries. Therefore, it becomes necessary to develop an automated currency recognition system that helps in recognition notes easily, accurately and efficiency. The
methodology consists of five stages, which are image acquisition, image preprocessing, feature extraction, classification, and, lastly, results and analysis. The
currency image will be pre-processed in grayscale and split into 100x100 blocks at selected regions of interest (ROI) on the currency. Next, binary matrix image features
and representations will be extracted. Lastly, the similarity percentage of the binary matrix will be calculated and compared with all currency image matrices. The highest similarity percentage will be chosen as the currency's origin. The proposed algorithm successfully classified the currency and improved the accuracy of currency classification, achieving a 93.4% accuracy rate from the experimental results. The proposed method could be useful for various applications, including financial
institutions, security agencies, and automated currency processing machines. |
|---|