Signature Verification System Using Machine Learning and Image Processing

Abstract

This paper presents a robust and automated signature verification system that integrates advanced machine learning and image processing techniques to enhance security and accuracy in signature authentication. The system undergoes a series of preprocessing steps, including grayscale conversion, noise reduction using Gaussian blur, adaptive thresholding for binarization, and feature extraction using the HOG technique. The extracted features are then utilized to train multiple classification models, including SVM, Random Forest, and XGBoost classifiers, to evaluate their effectiveness in signature verification. A GUI has been developed to facilitate seamless user interaction, allowing individuals to upload and verify their signatures in real time. The experimental results demonstrate that our proposed method achieves an accuracy exceeding 80%, making it a viable solution for secure and reliable authentication in applications such as financial transactions, document verification, and identity validation.