Pengenalan wajah merupakan salah satu teknologi biometric yang telah banyak dikembangkan oleh peneliti dibidang Computer Vision. Hasil penelitian dibidang ini juga sudah banyak digunakan dalam sistem keamanan seperti lingkungan intelijen dalam kepolisian serta dalam sistem keamanan ataupun verifikasi di berbagai perusahaan-perusahaan besar di Negara maju khususnya. Riset Tugas Akhir face recognition ini menggunakan citra ORL dan citra buatan. Dimana jumlah citra untuk ORL sebanyak 400 citra yang terdiri dari 40 kelas dan 10 pose, masing-masing citra mempunyai dimensi 112 x 92 piksel. Sedangkan, untuk citra buatan sebanyak 150 citra yang terdiri dari 15 kelas dan 10 pose, masing-masing citra mempunyai dimensi 180 x 150 piksel. Di skenario uji coba sistem ini terdapat 2 skenario pada masing-masing database uji coba, yaitu skenario 1 (7 pose citra training dan 3 pose citra testing) dan untuk skenario 2 (5 pose citra training dan 5 pose citra testing). Dalam proses uji coba, sistem ini membandingkan hasil akurasi antara penggunaan Histogram Equalization, Local Binary Pattern (LBP) dan tanpa proses Preprocessing, serta membandingkan hasil akurasi penggunaan jumlah fitur dalam proses klasifikasinya. Dalam proses klasifikasi menggunakan metode Canberra Distance dan Manhattan Distance dengan pendekatan penampilan (Appearance based). Hasil uji coba yang diperoleh membuktikan akurasi tertinggi terdapat pada citra ORL skenario 1 dengan histogram equalization yaitu 96,67%, sedangkan hasil terburuk pada citra ORL skenario 2 dengan LBP yaitu 33%.

Face recognition is one of biometric technologies that have been developed by researchers in the field of Computer Vision. Results of research in this area has also been widely used in security systems such as the police intelligence environment as well as in security systems or verification at various large companies in developed countries in particular. Research Thesis face recognition using ORL image and artificial images. Where the number of images to ORL 400 images consisting of 40 classes and 10 poses, each image has a dimension of 112 x 92 pixels. Meanwhile, for the artificial image of 150 images consisting of 15 classes and 10 poses, each image has the dimensions of 180 x 150 pixels. In scenario testing this system, there are 2 scenarios for each test database, ie scenario 1 (7 pose 3 pose imagery training and testing images) and for scenario 2 (5 pose pose imagery training and 5 testing images). In the process of testing, the system accuracy comparing results between the use of Histogram Equalization, Local Binary Pattern (LBP) and without Preprocessing process, as well as comparing the accuracy of the results of the use of the number of features in the classification process. In the classification process using Canberra Distance and Manhattan Distance to approach appearance (Appearance based). Test results obtained prove the accuracy is highest in scenario 1 ORL image with histogram equalization is 96.67%, while the worst result on ORL image of scenario 2 with LBP is 33%.`