PENGEMBANGAN METODE PENGENALAN WAJAH BERBASIS MOMENT INVARIANT DAN LINEAR DISCRIMINANT ANALYSIS SERTA IMPLEMENTASINYA PADA ROBOT

Abstract

Face recognition is one of biometrics and computer vision application, implemented widely in the fields of security, robotics and entertainment. The development of face recognition is still facing several problems such as illumination/lighting, pose and image size. Many methods have been proposed but there is still no satisfactory methodology. This research focused on a face re1cognition method based on Moment Invariant (MI) and Linear Discriminant Analysis (LDA). MI method is known as a method of producing the geometry characteristic of the object that are independent of translation, rotation, and scaling. The LDA method is a method of classification of objects by minimizing the within class object and maximizing the difference equation object. Based on these two methods, a novel method called MILDA has been successfully developed by combining moment invariant method and linear discriminant analysis. Experiments has been conducted using data from the AR-database. The database consist of 50 faces basic images which are extended into variation images by cropping, equalizing, resizing, and
illuminating. The total images used in this research 450 images. Face image are grouped into two, consist of 150 training images and 300 testing images. The accuracy level of MILDA face image recognition is 92.00%. This method has been implemented on a robot using processed camera.