基於幾何人臉特徵之智慧型頭部姿態估測

Abstract

近年來臉部特徵偵測和人臉辨識已被廣泛地研究,許多利用臉部特徵偵測的應用也隨之發展。本篇論文即利用幾何人臉特徵，針對頭部姿態的方向和角度，提出智慧型頭部姿態估測系統之設計。此系統能夠自動偵測影像中的人臉特徵，包括眼睛及嘴巴,進而判斷頭部姿態的方向和角度。本系統分成三個步驟完成智慧型頭部姿態估測系統設計,首先，利用膚色找出臉部位置後，使用幾何人臉特徵達到眼睛及嘴巴的高偵測率，第二，製作頭部姿態的人臉模擬立體模型，可調整臉部轉向角度及頭部傾斜角度，在轉盤上標有七個偵測點，根據不同的及，製作臉部模擬影像，並記錄每張影像的七個偵測點，作為類神經網路學習之用，第三，經由監督式學習的類神經網路設計達到頭部姿態的估測。本論文所提出的智慧型頭部姿態估測在於特定範圍內正確率可達97.3%。

Recently, facial feature detection and face recognition have been studied extensively and many applications using facial feature detection have been developed. This thesis is aimed at the development of head attitude estimation system (HAES) based on geometric facial features to detect the face orientation and angle. The HAES automatically detects eyes and mouth in the image as the facial features, and then determine the head attitude. There are three steps to complete the intelligent head attitude estimation. First, detect the human face based on skin color and use the geometric facial features to the detection of eyes and mouth in high accuracy rate. Second, build up a stereo facial model to simulate the head attitude which is able to adjust the face orientation and angle by seven detecting points marked on the face model. Record the seven detecting points on each image referring to a specific face orientation and angle, which will be used in neural network learning. Third, the HAES is completed by intelligent neural networks under supervised learning. The proposed HAES achieves a high accuracy rate up to 97.3%.