Perbandingan Metode Pengenalan Wajah Melalui Surveillance Berbasis Pengenalan Wajah

Abstract

Until now, the facial recognition method is still very difficult to do, especially when the facial confirmation process is accurate in real time. Facial recognition methods that have been tested, such as eigenface, Local Binary Pattern Histogram (LBPH), and fisherface, are feasible methods to be tested directly by applying these three methods to facial recognition-based surveillance systems. This study aims to compare the level of real-time accuracy in personal identification on the three methods through 4 parameters, namely accuracy, FAR (False Acceptance Rate), FRR (False Rejection Rate), and time condition, namely lighting conditions based on time, namely morning, noon, afternoon and evening. Based on the results of the tests that have been done, the average accuracy is obtained, namely the highest average with the eigenface method with an accuracy of 73.64%, FAR 0.11%, FRR 0.15%, the LBPH method obtains the highest average with an accuracy of 80, 91%, FAR 0.13%, FRR 0.07%, and finally fisherface got the highest average with 90.00% accuracy, 0.05% FAR, 0.05% FRR, in identifying personal. The results obtained by the Fisherface method tend to have the highest accuracy value based on the average both in terms of accuracy and the lighting conditions that have been tested.