Significance and Comparison of PCA in Removing Multicollinearity of Variables in Face Detection and Recognition
M. Tamilselvi
Abstract
One of the main biometric identification techniques, face recognition, has been used in a variety of industries, including e-commerce, security, the military, and economics. Its touchless identification method and optional nature for consumers make it indispensable compared to other techniques like fingerprint or iris recognition. Principal component analysis (PCA), the first face recognition method, continues to draw attention from academics due to its ability to reduce the dimensionality of data without sacrificing significant information and is most renowned for eliminating multicollinearity in variables. However, the process of constructing a PCA-based face recognition system still takes much time because real-world applications require taking into account many factors including lighting, facial expression, and camera angle. Moreover, integrating toolkit implementations into applications still requires a significant amount of work on the part of software developers. In order to help software developers efficiently customise their applications, this study offers a software outline for PCA-based face recognition. This study models a PCA-based face recognition system by calculating the eigen values of the photos in the LFW database and comparing the results to other techniques to determine accuracy.