Counterfeit Indonesian Banknote Detection Using Hybrid HOG-GLCM Features and Support Vector Machine

Abstract

Advances in digital printing and imaging technologies have increased the visual similarity between counterfeit and authentic banknote, reducing the reliability of manual inspection. This study proposes an automated ultraviolet (UV) image-based framework for detecting counterfeit Indonesian rupiah IDR 100,000 banknote using hybrid Histogram of Oriented Gradient (HOG) and Gray Level Co-occurrence Matrix (GLCM) feature extraction combined with Support Vector Machine (SVM) classification. Banknote images were acquired under UV illumination to highlight security features, such as fluorescence patterns and texture irregularities. The preprocessing stage included grayscale conversion, CLAHE-based contrast enhancement, gaussian noise reduction, normalization, and rotation-based data augmentation. HOG and GLCM descriptors were fused into a single feature vector and classified using an SVM with different kernel functions. The experimental result indicate that the linear kernel achieves the most stable and reliable performance, demonstrating that the hybrid HOG-GLCM features space is highly discriminative and predominantly linearly separable under UV imaging conditions. The proposed method provides a computationally efficient and cost-effective solution for automated banknote authentication in the future.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint