IMZO TASVIRLARINI TANIB OLISHDA HU, ZERNIKE VA GABOR XUSUSIYATLARI ASOSIDAGI YONDASHUV SAMARADORLIGI

Annotatsiya.Mazkur tezisda imzo tasvirlarini tanib olishda Hu invariant momentlari, Zernike momentlari va Gabor filtrlari asosidagi xususiyat ajratish yondashuvining samaradorligi tahlil qilinadi. Tadqiqotda imzo tasvirlariga dastlabki ishlov berish bosqichlari — shovqinni kamaytirish, binarizatsiya, kontur ajratish va normallashtirish amallari asosida barqaror xususiyatlar majmui shakllantirilgan. Ajratilgan xususiyatlar SVM, Random Forest va CNN modellarida sinovdan o‘tkazilib, kombinatsiyalangan descriptorlar imzoning geometrik va teksturaviy belgilarini aniqroq ifodalashi isbotlangan. Natijalar ushbu yondashuv biometrik autentifikatsiya tizimlarida aniqlikni oshirish va imzo verifikatsiyasi ishonchliligini ta’minlashda samarali ekanini ko‘rsatadi. Kalit so‘zlar: imzo tasviri, biometrik autentifikatsiya, Hu momentlari, Zernike momentlari, Gabor filtrlari, xususiyat ajratish, tasniflash, CNN, SVM, Random Forest. Abstract.This thesis analyzes the effectiveness of a feature extraction approach based on Hu invariant moments, Zernike moments, and Gabor filters for signature image recognition. In the study, a stable set of features was formed through preliminary processing stages applied to signature images, including noise reduction, binarization, contour extraction, and normalization. The extracted features were tested using SVM, Random Forest, and CNN models, and it was demonstrated that the combined descriptors represent the geometric and textural characteristics of signatures more accurately. The results show that this approach is effective in improving accuracy and ensuring the reliability of signature verification in biometric authentication systems. Keywords: signature image, biometric authentication, Hu moments, Zernike moments, Gabor filters, feature extraction, classification, CNN, SVM, Random Forest.

Перевод пока недоступен