三維資訊在人臉辨識上的應用

Abstract

本論文主要是研究三維資訊在人臉辨識系統上的應用,由二維照相機和三維高 解析度照相機所取得的二維影像及三維網格,對上述的資料進行臉部描述,其中 二維影像是透過紋理特徵分析的 LBP (Local binary patterns) 來取出每張照片的 特徵向量;三維網格的部份,是對於臉上的特徵點透過 Point signature 取得特徵 向量。進而利用機器學習中的 PCA (Principal component analysis)、LDA (Linear discriminant analysis) 以降低特徵向量維度並提升資料分類效能,最後使用機器 學習中的 SVM (Support vector machines) 來訓練分類器。數值結果顯示;透過 Point signature 所取得的三維特徵資訊對於人臉表情辨識系統是有幫助的。

This research is to study the 3D information on the application of face recognition system based on the 2D images and 3D meshes captured by the 2D camera and 3D high-resolution scanner. For the 2D images, we analyze the texture based on LBP(Local binary patterns) to get the feature vectors. For the 3D meshes, we define the feature vectors by using the Point signature of the selected feature points. Furthermore, we use machine learning techniques such as PCA (Principal component analysis), LDA (Linear discriminant analysis) in order to reduce the dimension of the feature vector and improve the performance of data classification. Finally, we train the classifier using machine learning based on SVM (Support vector machines). Several numerical results indicate the 3D information based on Point signature is helpful on face expression recognition system.