Sistem Pengenalan Wajah Berbasis Video untuk Monitoring Pintu

Ahmad, Luqman (2018) Sistem Pengenalan Wajah Berbasis Video untuk Monitoring Pintu. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

[img]
Preview
Text
05111440000187-Undergraduate_Theses.pdf - Accepted Version

Download (2MB) | Preview

Abstract

Sistem pengenalan wajah merupakan suatu sistem yang berguna untuk mengidentifikasi wajah seseorang melalui citra wajah orang tersebut. Sementara itu, video merupakan sarana multimedia yang mengkombinasikan citra secara berurutan sehingga membentuk citra yang seolah – olah bergerak. Sistem pengenalan wajah berbasis video merupakan sistem pengidentifikasi wajah seseorang bukan hanya melalui satu citra melainkan juga memperhatikan beberapa citra yang diperoleh melalui video. Tugas akhir ini mengembangkan sistem untuk mengidentifikasi seseorang ketika memasuki ruangan. Sebuah kamera diletakkan menghadap pintu masuk untuk meperoleh data citra wajah seseorang. Sistem akan memproses data tersebut dan kemudian memberikan hasil pengenalan. Sistem ini terdiri dari tiga tahapan utama, yakni (1) tahap deteksi wajah, (2) tahap ekstraksi fitur, dan (3) tahap klasifikasi. Tahap deteksi wajah bertujuan untuk memisahkan citra wajah dari citra input keseluruhan menggunakan algoritma Seeta Face dan Kalman Filter. Tahap ekstraksi fitur bertujuan memperoleh fitur representatif citra wajah menggunakan Discrete Cosine Transform. Tahap klasifikasi adalah tahap pengambilan kesimpulan suatu citra wajah menggunakan algoritma k-Nearest Neighbour. Uji coba meggunakan 31 video, terdiri dari 18 video untuk training dan 13 video untuk testing. Tiap video terdiri dari satu orang yang sedang memasuki ruangan. Data video tersebut memiliki 8 kelas yang merupakan nama orang tertentu. Uji coba pengenalan wajah berbasis video menghasilkan akurasi 100%, sementara uji coba pengenalan wajah berbasis frame menghasilkan akurasi 70,73%. ============================================================================================= Face recognition system is a system used for identifying face through the image of that person. Meanwhile, video is multimedia source combines sequences of image to form a moving picture. Video based face recognition system is face identifier system not only from one image but also considering sequence of image extracted from video. This final project develop system identifying someone when he is entering a room. A camera placed in the room facing the entrance door for collection face image data. System will process the data and then give the recognition result. This system consists of three main processes, (1) face detection process, (2) feature extraction process, and (3) classification process. Face detection process intended for separate face image from a whole input image from camera using Seeta Face Detection and Kalman Filter. Feature extraction process intended for obtaining representative feature of face image using Discrete Cosine Transform. Classification process is process of concluding face image using k-Nearest Neighbour. Testing use 31 videos, consist of 18 videos as training set and 13 videos as testing set. Each video consists of one person entering the room. Video data have 8 classes which is name of the person. Testing of video-based face recognition system give 100% accuracy, while testing of frame-based face recognition system give 70,73% accuracy.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: deteksi wajah, pengenalan wajah, seeta face detection, kalman filter, discrete cosine transform, k-nearest neighbour
Subjects: T Technology > TA Engineering (General). Civil engineering (General) > TA1650 Face recognition. Optical pattern recognition.
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems
Divisions: Faculty of Information and Communication Technology > Informatics > 55201-(S1) Undergraduate Thesis
Depositing User: Luqman Ahmad
Date Deposited: 21 Jun 2021 01:27
Last Modified: 21 Jun 2021 01:27
URI: https://repository.its.ac.id/id/eprint/53985

Actions (login required)

View Item View Item