Analisis Akurasi Similarity Wajah Menggunakan Manhattan Distance Melalui Fitur Bentuk, Warna dan Tekstur

Abstract

Face is a human biometric characteristic that can be categorized as unique. Every human being
has different faces or differences despite being born twins. These differences make the face unique. This
makes the face image as one of the unique fields that can be used as key field on the detection of personal
information based on the image. In the process of verifying the attendance of lectures manually through
signature detection is still possible in the presence of counterfeiting and takes a long time. This condition
can be anticipated through the verification of the attendance of the lecture using face detection.
Conditions at lectures that affect the accuracy of face detection include students alternating seating
position, facial pose variation, class lighting and camera distance to the object. This requires methods to
improve detection accuracy in order to verify attendance. The method used in this research is facial
recognition through matching similarity of shape, color and texture. Form feature extraction uses
invariant moment, color feature extraction using color moment and texture feature extraction using
statistical texture. In this research, we developed a model of college attendance verification through the
introduction of many class faces or classrooms using stereo vision camera. The test is done by combining
two images from the stereo vision camera. The test will produce a system that has a detection accuracy
level that supports valid lecture verification with various seating variations, pose variation, brightness
illumination variations and camera range variations with objects. The result of matching similarity test
reaches more than 75% accuracy.