The classification of pen ink aging by machine learning and deep learning technique using Raman spectrum
Yazarlar (3)
Kübra Gürbüz Göçmen
Prof. Dr. Mustafa Cem KASAPBAŞI İstanbul Ticaret Üniversitesi, Türkiye
Sinan Bosna
| Makale Türü |
Özgün Makale
(Diğer hakemli uluslarası dergilerde yayınlanan tam makale)
|
||
| Dergi Adı | Journal of MillimeterwaveCommunication, Optimization and Modelling | ||
| Dergi ISSN | 2021-9999 | ||
| Dergi Tarandığı Indeksler | ROAD | ||
| Makale Dili | Türkçe | Basım Tarihi | 01-2025 |
| Cilt / Sayı / Sayfa | 5 / 1 / 20–24 | DOI | – |
| Makale Linki | https://www.jomcom.org/index.php/1/article/view/84 | ||
| UAK Araştırma Alanları |
Bilgi Güvenliği ve Kriptoloji
Veri Madenciliği
Makine Öğrenmesi
|
||
| Özet |
| Forgery of valuable documents generally constitutes falsification methods based on altering a previously written document by using similar or identical ink. In the event of the aforementioned situation, forensic science experts conduct various technical examinations on the relevant document using different devices. One of the main purposes of these examinations is to determine the differences in the aging levels of the inks relative to each other. Raman Spectra, which is also used for different purposes in forensic sciences, is one of the methods that can be used in this field. The Raman spectrometer provides information about molecules' vibration energy levels and presents the analyzed region's spectral signature values. Experts can observe the time-dependent changes that occur in the substances in the region under investigation relative to each other and in the substance content through the information obtained. Utilizing this information, sample data were created at different times using the same pen on A4 paper in our study. These data were divided into two groups old and new data. Raman spectra were taken with a 785 nm laser on both sample data. Sequential Keras model, KNN, and SVM algorithms were used to detect ink aging on paper. The k-fold cross-validation method was used to determine the classification performance more accurately. The results showed that the classification performance was 98.71% for the neural network and 100% for the KNN and SVM. |
| Anahtar Kelimeler |
