Multiple Face Detection and Recognition Using HSV Histogram Matching and Principal Component Analysis Techniques
Authors: Alberto Bañacia, Mark Bryan Soylon, Sean Chris Conson, Jonn Dennis Regidor, and Florencio Cortes
Abstract
Multiple face detection and recognition system using HSV Histogram Matching and Principal Component Analysis was successfully implemented in this study. The system setup consists of a Canon VBC50i Network Camera that is used to capture the face images in a white background, perform face detection and recognition algorithms using Matlab installed in a laptop that also doubles as a database. The results show that HSV color space was successful in detecting a skin color with a white background resulting to a high precision and recall given that the positions of faces were varied. However, there were factors identified in this study that limited the systems detection capability, which include lighting conditions of the input image, the auto exposure feature of the Canon VB-C50i network camera, and the limitation of inverse cropping that was used for multiple face detection. The detected face(s) went through Principal Component Analysis. Euclidian distances were acquired to serve as the basis for face recognition. The system continued to loop until all the detected faces were recognized displaying a detected face with its corresponding profile. Recognition for multiple faces showed a higher rate in frontal orientation and fair rate on angled orientation of faces.