Pengenalan Wajah Menggunakan Metode Support Vector Machine (SVM) Yudhie Suherdani / 9922109
Jurusan Teknik Elektro, Fakultas Teknik, Univeristas Kristen Maranatha Jln. Prof. Drg. Suria Sumantri 65, Bandung 40164, Indonesia Email :
[email protected]
ABSTRAK
Pada saat ini sistem pengenalan wajah merupakan salah satu teknologi biometrics yang paling populer. Kesulitan pada implementasi sistem ini terutama disebabkan oleh variasi gambar wajah yang kompleks. Perubahan ekspresi wajah, perubahan posisi dan ukuran wajah, perubahan arah sumber cahaya, penggunaan aksesoris, dan lain sebagainya, akan mempengaruhi hasil dari proses pengenalan wajah. Pada sistem pengenalan wajah menggunakan metode Support Vector Machine, perbedaan intensitas antara dua gambar (difference image), merupakan karakterisitik penting yang dapat menggambarkan variasi gambar wajah seseorang. Difference image dapat diklasifikasikan ke dalam dua class, yaitu intrapersonal variations class dan extrapersonal variations class. Jika difference image berada dalam intrapersonal variations class maka dapat dikatakan bahwa dua gambar pembentuk difference image merupakan gambar dari orang yang sama, dan sebaliknya jika difference image berada dalam extrapersonal variations class, maka dapat dikatakan bahwa dua gambar pembentuk difference image merupakan gambar dari orang yang berbeda. Berdasarkan hasil percobaan yang dilakukan, sistem pengenalan wajah menggunakan metode Support Vector Machine memiliki tingkat keberhasilan pengenalan sebesar 97,78 % untuk proses di dalam training set dan 96,67 % untuk proses diluar training set. Hasil ini diperoleh dengan menggunakan 11 gambar wajah yang berbeda untuk setiap subyek pada gallery dengan total subyek yang terdapat pada gallery sebanyak 11 orang, dan jumlah dimensi yang dipertahankan pada proses PCA sebanyak 30. Kata kunci : pengenalan wajah, Support Vector Machine, Fisherface, PCA, FLD i Universitas Kristen Maranatha
Face Recognition Using Support Vector Machine (SVM) Yudhie Suherdani / 9922109
Department of Electrical Engineering, Faculty of Techniques, Maranatha Christian University Jln. Prof. Drg. Suria Sumantri 65, Bandung 40164, Indonesia Email :
[email protected]
ABSTRACT Face recognition system is one of the most popular biometrics technology at this moment. Difficulties with implementing this system, due to complexity of face it self. Face expression, size and face position, light source changes, face accecories will have major effect to recognition result. Face recognition using Support Vector Machine method believes that the intensity difference between two images (difference image) can describe the variation of someone’s face. Difference image can be classified into two classes, mainly intrapersonal variation class and extrapersonal variations class. If difference image lies in intrapersonal variation class, then it could be said that the two images which construct the difference image belong to the same individual and if the difference image lies in extrapersonal variation class, then the it could be said that the two images which construct the difference image belong to two different individuals. Based on experimental results, face recognition system using Support Vector Machine method achieves successful recognition rate about 97.78 % for internal probes and 96.67 % for external probes. These results were achieved by using eleven different face images for each person in the gallery with total of 11 persons in the gallery and by keeping 30 dimensions while running PCA. Keywords : Face Recognition, Support Vector Machine, Fisherface, PCA, FLD
ii
Universitas Kristen Maranatha
Daftar Isi
Abstrak...................................................................................................
i
Abstract..................................................................................................
ii
Kata Pengantar.....................................................................................
iii
Daftar Isi................................................................................................
v
Daftar Gambar......................................................................................
vii
Daftar Tabel..........................................................................................
ix
BAB I Pendahuluan..............................................................................
1
1.1
Latar Belakang............................................................................
1
1.2
Identifikasi Masalah.....................................................................
3
1.3
Tujuan.......................................................................................
3
1.4
Pembatasan Masalah....................................................................
3
1.5
Sistematika Penulisan..................................................................
4
BAB II Landasan Teori...................................................................... 2.1
2.2
2.3
5
Deteksi Wajah (Face Detection)....................................................
7
2.1.1 Segmentasi Kulit (Skin Segmentation)..................................
7
2.1.2 Proses Morfologi................................................................
9
2.1.3 Connected Region Analysis..................................................
10
2.1.4 Ekstraksi Wajah (Face Extracting)........................................
11
Pengenalan Wajah (Face Recognition)...........................................
11
2.2.1 Linear Support Vector Machine............................................
12
2.2.2 Soft Margin.......................................................................
15
2.2.3 Non-Linear Support Vector Machine....................................
15
Klasifikasi Identitas.....................................................................
20
BAB III Perancangan Sistem.............................................................
22
3.1
Deteksi Wajah.............................................................................
23
3.2
Perhitungan PCA (Principal Component Analysis)..........................
28
3.3
Perhitungan FLD (Fisher Linear Discriminant)...............................
29
v
Universitas Kristen Maranatha
3.4
Klasifikasi Identitas.....................................................................
BAB IV Simulasi dan Analisa............................................................
31
34
4.1
Deteksi Wajah.............................................................................
36
4.2
Konstruksi Support Vector Machine..............................................
38
4.3
Klasifikasi..................................................................................
41
4.3.1 Metoda Jarak Euclidian.......................................................
44
4.3.2 Metoda Jarak Cosinus.........................................................
49
BAB V Kesimpulan Dan Saran.........................................................
56
5.1
Kesimpulan.................................................................................
56
5.2
Saran.........................................................................................
57
Daftar Pustaka......................................................................................
58
Lampiran Listing Program.................................................................. A-1
vi
Universitas Kristen Maranatha
Daftar Gambar
Gambar 2.1
Proses Face Recognition ..................................................................... 5
Gambar 2.2
Proses Deteksi Wajah………………………………………………......7
Gambar 2.3
Korelasi antara nilai Cb dan Cr yang menunjukan warna kulit………..8
Gambar 2.4
Nilai Cb dan Cr untuk warna kulit dan untuk seluruh gambar………...9
Gambar 2.5
Proses erosi dan dilasi………………………………………………...10
Gambar 2.6
SVM Hyperplane……………………………………………………...12
Gambar 2.7
Fungsi ߶………………………………………………………………16
Gambar 2.8
Jarak Euclidian dan jarak cosine…………………………………….. 21
Gambar 3.1
Diagram blok perancangan sistem ..................................................... 22
Gambar 3.2
Diagram Diagram alir sistem pengenalan wajah ................................ 23
Gambar 3.3
Diagram alir proses deteksi wajah........................................................24
Gambar 3.4
Diagram alir pembatasan warna dan proses morfologi……………….25
Gambar 3.5
Diagram alir proses region analysis…………………………………..26
Gambar 3.6
Diagram alir ekstraksi gambar wajah…………………………………27
Gambar 3.7
Diagram alir perhitungan PCA………………………………………..29
Gambar 3.8
Diagram alir perhitungan FLD……………………………………….31
Gambar 3.9
Diagram alir proses klasifikasi………………………………………..33
Gambar 4.1
Training set ....................................................................................... 35
Gambar 4.2
Gambar topeng hasil proses segmentasi kulit .................................... 36
Gambar 4.3
Gambar topeng hasil proses morfologi .............................................. 37
Gambar 4.4
Gambar topeng hasil proses connected region analysis. .................... 37
Gambar 4.5
Keluaran proses deteksi wajah .......................................................... 38
Gambar 4.6
Average face ..................................................................................... 38
Gambar 4.7
Set SVM………………………………………………………………39
Gambar 4.8
Rekonstruksi training set……………………………………………...40
Gambar 4.9
Rekonstruksi gambar wajah masukan………………………………...40
vii
Universitas Kristen Maranatha
Gambar 4.10
Bobot set SVM ……………………...……………………………….41
Gambar 4.11
Jarak Euclidian dan jarak cosinus………………………………….....42
Gambar 4.12
Beda jarak Euclidian minimum……………………………………….45
Gambar 4.13
False acceptance rate (FAR) jarak Euclidian…………………………47
Gambar 4.14
False rejection rate (FRR) jarak Euclidian……………………………47
Gambar 4.15
Equal error rate (EER) Jarak Euclidian……………………………….48
Gambar 4.16
Beda jarak cosinus minim…………………………………………….51
Gambar 4.17
False acceptance rate (FAR) jarak cosines……………………………52
Gambar 4.19
Equal error rate (EER) jarak cosines………………………………….53
Gambar 4.20
Hasil keputusan system……………………………………………….54
viii
Universitas Kristen Maranatha
Daftar Tabel Tabel 1.1
Fungsi kernel yang sering dipakai dalam……………………………………….
Tabel 4.1
Jarak Euclidian minimum untuk individu di dalam training set………………... 44
Tabel 4.2
Jarak Euclidian minimum untuk individu di luar training set…………………..
45
Tabel 4.3
Analisa kesalahan metode jarak Euclidian...........................................................
47
Tabel 4.4
Jarak cosinus minimum untuk individu di dalam training set………………......
50
Tabel 4.5
Jarak cosinus minimum untuk individu di luar training set…………………….. 50
Tabel 4.6
Analisa kesalahan metode jarak cosinus............................................................... 52
ix
17
Universitas Kristen Maranatha
