Scholarlink

By reducing human mistakes in manual verification, the Offline Signature Forgery Detection System personal authentication is automated. The preprocessing steps for scanned signature images (gray scale, Otsu’s binarization, segmentation) are followed by extracting handmade features (pixel density ratio, centroid, eccentricity, solidity, skewness, and kurtosis). A shallow MLP classifier (Tensor Flow 1.x) was developed to segregate authentic and forged signatures using these features. This method performs admirably, is easy to understand, and does well in some limited circumstances. We have demonstrated good accuracy on test/train, so this has potential to interface with deep learning or hybrid models in the future.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Automated Offline Signature Verification Using Deep Learning Techniques

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Abhinay Karthik Kaveti,
  • Ch. Nanda Krishna,
  • Megha Shaym Bugata

摘要

By reducing human mistakes in manual verification, the Offline Signature Forgery Detection System personal authentication is automated. The preprocessing steps for scanned signature images (gray scale, Otsu’s binarization, segmentation) are followed by extracting handmade features (pixel density ratio, centroid, eccentricity, solidity, skewness, and kurtosis). A shallow MLP classifier (Tensor Flow 1.x) was developed to segregate authentic and forged signatures using these features. This method performs admirably, is easy to understand, and does well in some limited circumstances. We have demonstrated good accuracy on test/train, so this has potential to interface with deep learning or hybrid models in the future.