SAFITRI, AISYAH (2026) IMPLEMENTASI ALGORITMA EIGENFACE UNTUK DETEKSI DAN PENGENALAN WAJAH TAMPAK SAMPING. Skripsi (Bachelor) thesis, Universitas Muhammadiyah Bengkulu.
![[img]](http://repository.umb.ac.id/style/images/fileicons/text.png) |
Text (Fulltext)
AISYAH SAFITRI_2255201022_Fulltext.pdf - Submitted Version Restricted to Repository staff only Download (2MB) | Request a copy |
|
Text (Turnitin)
AISYAH SAFITRI_2255201022_Turnitin.pdf - Submitted Version Restricted to Repository staff only Download (9MB) | Request a copy |
|
Text (ABSTRAK + BAB 1)
AISYAH SAFITRI_2255201022_ABSTRAK + BAB 1.pdf - Submitted Version Download (367kB) |
|
Text (BAB 2)
AISYAH SAFITRI_2255201022_BAB 2.pdf - Submitted Version Download (208kB) |
|
Text (BAB 3)
AISYAH SAFITRI_2255201022_BAB 3.pdf - Submitted Version Restricted to Repository staff only Download (495kB) | Request a copy |
|
Text (BAB 4)
AISYAH SAFITRI_2255201022_BAB 4.pdf - Submitted Version Restricted to Repository staff only Download (962kB) | Request a copy |
|
Text (BAB 5)
AISYAH SAFITRI_2255201022_BAB 5.pdf - Submitted Version Download (160kB) |
|
Text (Ref + Lamp)
AISYAH SAFITRI_2255201022_Ref+Lamp.pdf - Submitted Version Download (329kB) |
Abstract
Perkembangan teknologi kecerdasan buatan mendorong pemanfaatan pengolahan citra digital pada sistem pengenalan wajah, namun mayoritas sistem saat ini masih berfokus pada wajah tampak depan (frontal face). Pada kondisi nyata seperti kamera pengawas, wajah objek sering terlihat miring atau tampak samping (side-view), sehingga menghilangkan fitur geometri wajah dan menurunkan akurasi. Penelitian ini bertujuan mengimplementasikan algoritma Eigenface berbasis Principal Component Analysis (PCA) untuk mendeteksi serta mengenali karakteristik wajah tampak samping sekaligus mengevaluasi akurasinya. Sistem ini dikembangkan sebagai aplikasi berbasis web menggunakan JavaScript dengan pustaka TensorFlow.js untuk pemrosesan tensor citra secara client-side dan Math.js untuk perhitungan matriks nilai eigen secara real-time. Metode penelitian yang digunakan adalah pendekatan kuantitatif melalui eksperimen rekayasa perangkat lunak. Pengumpulan data primer dilakukan menggunakan kamera ponsel untuk menghimpun dataset citra wajah pada variasi pencahayaan dan sudut kemiringan objek. Pada tahap pra-pemrosesan, citra dikondisikan secara asinkronus dengan mengubah dimensi menjadi 100×100 piksel, dikonversi ke skala kelabu, serta dinormalisasi ke skala desimal 0 hingga 1 menggunakan fungsi tf.tidy demi efisiensi memori peramban. Ekstraksi ciri dijalankan dengan membentuk ruang wajah (face space) melalui perhitungan wajah rata-rata, matriks kovarian, dan vektor eigen. Proses pencocokan identitas dilakukan dengan mengukur jarak euclidean terkecil antara proyeksi fitur citra uji dan citra latih, lalu dikonversi menjadi persentase tingkat keyakinan dengan ambang batas minimal 50,0%. Pengujian sistem dievaluasi secara kuantitatif menggunakan Confusion Matrix 2×2 melibatkan 19 sampel citra uji wajah tampak samping. Hasil eksperimen menunjukkan sistem berhasil mengenali 12 citra dengan status benar dan gagal pada 7 citra dengan status salah akibat interferensi bayangan ekstrem serta rotasi sudut yang tajam. Dari analisis performa, sistem mencapai tingkat akurasi sebesar 63,16%, presisi 100,00%, dan sensitivitas 63,16%. Penelitian ini membuktikan algoritma Eigenface dalam ekosistem JavaScript efektif untuk reduksi dimensi citra wajah non-frontal secara interaktif pada peramban web, meski performanya dipengaruhi fluktuasi cahaya dan stabilitas sudut orientasi objek. The development of artificial intelligence technology encourages the use of digital image processing in facial recognition systems, but the majority of current systems still focus on frontal faces. In real conditions such as surveillance cameras, the object's face often appears tilted or side-viewed, thus eliminating facial geometry features and lowering accuracy. This study aims to implement the Eigenface algorithm based on Principal Component Analysis (PCA) to detect and recognize the characteristics of side-looking faces as well as evaluate their accuracy. The system was developed as a web-based application using JavaScript with a TensorFlow.js library for client-side processing of image tensors and Math.js for real-time calculation of eigenvalue matrices. The research method used is a quantitative approach through software engineering experiments. Primary data collection was carried out using a mobile camera to collect facial imagery datasets on variations in lighting and angle of inclination of objects. In the pre-processing stage, the image is conditioned asynchronously by changing the dimensions to 100×100 pixels, converted to grayscale, and normalized to a decimal scale of 0 to 1 using the tf.tidy function for browser memory efficiency. Feature extraction is carried out by shaping the face space through the calculation of the average face, covariance matrix, and eigenvector. The identity matching process was carried out by measuring the smallest Euclidean distance between the projection of the test image feature and the trained image, then converted to a percentage of the confidence level with a minimum threshold of 50.0%. The system test was evaluated quantitatively using Confusion Matrix 2×2 involving 19 side-appearance test image samples. The results of the experiment showed that the system managed to recognize 12 images with the correct status and failed on 7 images with the wrong status due to extreme shadow interference and sharp angular rotation. From the performance analysis, the system achieved an accuracy level of 63.16%, precision of 100.00%, and sensitivity of 63.16%. This study proves that the Eigenface algorithm in the JavaScript ecosystem is effective for the interactive reduction of non-frontal facial images in web browsers, although its performance is affected by light fluctuations and the stability of object orientation angles.
| Item Type: | Thesis (Skripsi (Bachelor)) |
|---|---|
| Additional Information: | Pembimbing : Yuza Reswan, S.Kom., M.Kom. |
| Uncontrolled Keywords: | Principal Component Analysis (PCA), TensorFlow.js, Confusion Matrix, Jarak Euclidean, Pengolahan Citra Digital. |
| Subjects: | Universitas Muhammadiyah Bengkulu > 02-Fakultas Teknik > 55201-(S1) Teknik Informatika 02-Fakultas Teknik > 55201-(S1) Teknik Informatika |
| Divisions: | 02-Fakultas Teknik > 55201-(S1) Teknik Informatika Subjek Terkait > 02-Fakultas Teknik > 55201-(S1) Teknik Informatika |
| Depositing User: | Mrs Atika Ledia Anggraini |
| Date Deposited: | 14 Aug 2026 02:40 |
| Last Modified: | 14 Aug 2026 02:40 |
| URI: | http://repository.umb.ac.id/id/eprint/2901 |
Actions (login required)
 |
View Item |