Israa Bashir Mohammed; Bashar Saadoon Mahdi; Mustafa S. Kadhm
الملخص الانجليزي
The authenticated handwritten signatures play a critical role in many daily lives operations. Many organizations
and banks around world depend on the handwritten signatures in their works. Forged signatures may cause big problems
that lead to serious crimes. Therefore, in this work, a reliable handwritten signature authentication mechanism is proposed.
The system employs an effective image preprocessing stage to improve the resolution of input handwritten signatures and
eliminate any picture noise that could impact the system's outcomes. Besides, a modified MobileNets model architecture
is used for the classification process by train input handwritten signatures image to identify the right writers of the desired
signatures. The K Nearest Neighbor (KNN) is used of authentication process by matching the features of the trained
handwritten signaturesimage and the testing once. The solution presented attained a 99.7% authentication rate when tested
with the CEDAR dataset.
أ.م.د مصطفى سلام كاظم | Mustafa Salam Kadhm | 16300