Afifah, Imada Nur (2018) Implementasi Ekstraksi Fitur Jumlah Keypoint Descriptor pada Pengenalan Tanda Tangan dengan Algoritme Learning Vector Quantization. Sarjana thesis, Universitas Brawijaya.
Abstract
Pengenalan tanda tangan sangat penting untuk proses verifikasi tanda tangan. Salah satu metode pengenalan tanda tangan adalah implementasi algoritme Learning Vector Quantization (LVQ) pada pengenalan tanda tangan dengan ekstraksi fitur jumlah keypoint descriptor menggunakan metode Scale Invariant Features Transform (SIFT) dan fitur fraktal global. Dalam proses pelatihan pengenalan tanda tangan digunakan fitur-fitur seperti nilai maksimum piksel hitam dari histogram horizontal dan vertikal, massa pusat objek, normalisasi luas objek tanda tangan, rasio aspek, fitur three surface, fitur six fold surface dan fitur transisi serta tambahan fitur berupa jumlah keypoint descriptor yang didapatkan dari hasil penggunaan metode Scale Invariant Features Transform (SIFT). Dataset yang digunakan antara lain 100 citra untuk data latih dan 100 citra untuk data uji dari 20 kelas yang berbeda. Serta 25 citra diluar kelas yang dilatih sebagai data yang tidak dikenali. Hasil pengujian diperoleh 71,2% dari pengujian tanpa threshold, 56% dari pengujian dengan nilai threshold menggunakan nilai maksimum jarak terpendek antara citra dengan kelas, 45,6% dari pengujian dengan nilai threshold menggunakan nilai maksimum kedua dari jarak terpendek antara citra dengan kelas.
English Abstract
Signature recognition is important for signature verification process. One of the signature recognition method is implementation Learning Vector Quantization (LVQ) for signature recognition with number of keypoint descriptor features extraction method using Scale Invariant Features Transform (SIFT) and fractal global. In the train process, this research used some features such as maximum of black pixel in horizontal and vertical histogram, center of mass, normalized area of signature, aspect ratio, tri surface feature, the Six Fold Surface feature, transition feature and additional features called number of keypoint descriptors. Number of keypoint descriptors are output of Scale Invariant Features Transform (SIFT) method. The dataset used is 100 images for training data and 100 images for testing data from 20 different classes. And 25 images from out of trained class as unknown data. The result of algorithm testing is 71,2% from testing of nonthreshold process, 56% from testing process with maximum value of minimum euclidean distance between data and class as threshold value, 45,6% % from testing process with second maximum value of minimum euclidean distance between data and class as threshold value.
| Item Type: | Thesis (Sarjana) |
|---|---|
| Identification Number: | SKR/FTIK/2018/332/051804145 |
| Uncontrolled Keywords: | sift, Learning Vector Quantization, pengenalan, tanda tangan sift, Learning Vector Quantization, recognition, signature |
| Subjects: | 000 Computer science, information and general works > 006 Special computer methods > 006.2 Special-purpose systems > 006.24 Automatic identification and data capture (AIDC) |
| Divisions: | Fakultas Ilmu Komputer > Teknik Informatika |
| Depositing User: | Budi Wahyono Wahyono |
| Date Deposited: | 06 Mar 2019 01:49 |
| Last Modified: | 16 Oct 2021 05:48 |
| URI: | http://repository.ub.ac.id/id/eprint/13417 |
Preview |
Text
Imada Nur Afifah.pdf Download (3MB) | Preview |
Actions (login required)
 |
View Item |