3D face recognition based on volumetric representation of range image

Abstract

In this paper, a 3D face recognition system has been developed based on the volumetric representation of 3D range image. The main approach to build this system is to calculate volume on some distinct region of 3D range face data. The system has mainly three steps. In the very first step, seven significant facial landmarks are identified on the face. Secondly, six distinct triangular regions A to F are created on the face using any three individual landmarks where nose tip is common to all regions. Further 3D volumes of all the respective triangular regions have been calculated based on plane fitting on the input range images. Finally, KNN and SVM classifiers are considered for classification. Initially, the classification and recognition are carried out on the different volumetric region, and a further combination of all the regions is considered. The proposed approach is tested on three useful challenging databases, namely Frav3D, Bosphorous, and GavabDB.