A study of features and classifiers for multiple environment face recognition system
Manita Boontua, Pattraporn Nam-asa, Sujitra Arwatchananukul, Nattapol Aunsri · 2018
This paper presents a study of multiple environment consideration for face recognition system in order to investigate the suitable pairs of features and classifiers of the system. The variation of the environment is devoted to the consideration for illumination of the working system and poses of the users. The images used in the dataset were prescribed as grayscale in 2D images. The performances of selected features were evaluated for each classifier. The experiments were designed in two main situations consisting of similar illumination setting for both training set and testing set, and different illumination of both sets. The experimental results demonstrate that the similarity in lighting of training set and testing set provides better accuracy than that from where the illumination settings between training set and testing set are different.