Analysis and Representation of Face Robot-portrait Features. Issue 3 (3rd July 2018)

• Record Type: Journal Article
• Authors: Heenaye- Mamode Khan, M.; Nagawah, Nivedita; Bheechook, Lalesh
• Is Part Of: Journal of applied security research. Volume 13:Issue 3(2018)
• Page Start: 332
• Page End: 346
• Publication Date: 2018-07-03
• Subjects: Biometrics -- feature representation -- robot-portrait; Police, Private -- Training of -- Periodicals; Police, Private -- Training of -- United States -- Periodicals; Private security services -- Periodicals; 363.289
• DOI: 10.1080/19361610.2018.1463138
• Languages: English
• ISSNs: 1936-1610

Abstract

ABSTRACT: Biometrics is being increasingly implemented and used in systems, mainly in security. Many works are still being carried out by researchers to improve existing biometric systems. One current important ongoing research in the field of security is the use of face biometrics in police departments to recognize criminals via portrait robot images described from crime witnesses. However, the correct identification of an individual is still a challenge. This is because the features required for the identification or the authentication might not have been accurately drawn or generated. The need of efficient robot-portrait image applications makes this study of vital importance and thus, a face recognition application using robot-portrait images is being developed in this research work. To extract and represent the features from the images, two techniques namely, Principal Component Analysis (PCA) and Local Binary Patterns (LBPs) have been adopted. To cater for the high computational load of the face features, PCA has been improved through the Gram-Schmidt orthogonalization process and the Rayleight-Quotient iteration scheme. From the results generated from the experiments conducted, it was concluded that modified PCA and LBP are two promising techniques that can be adopted to develop robot-portrait recognition applications.