基於人體測量學模型使用角度辨識臉部年齡分類之研究

基於人體測量學模型使用角度辨識臉部年齡的研究，由於人體測量學方法為測量臉部比例方法做研究，但是對於兒童接近成年人階段臉部分辨特徵不明確，所以大多適用於分辨幼年，較不適用於成年人。本文利用非單純背景的圖像去獲取人臉器官相對位置與比例，並提出一個將器官特徵轉化以角度為特徵研究的方法。研究基礎使用哈爾特徵，通過歐式距離訓練分類成人或小孩。此研究方法對於辨識壯年人口效果較差，但對學齡前與青少年的年齡層較有效，對於辨識非成年年齡層的系統較嚴謹，擁有不錯的辨識率能避免將非成年誤判為成年。而這個研究方法對所有年齡層的總辨識率最高可達85.02%，可惜此研究方法對於辨識壯年人口效果較差，但對辨識學齡前與青少年的年齡層頗有效，最高辨識率可達98.37%以上。

關鍵字

人臉年齡；哈爾特徵；歐式距離

並列摘要

In this thesis, an age classification method, anthropometric method is measuring the face ratio method and do research. Distinguished the face features away from child to adulthood is not easy. Anthropometric method can apply to distinguish for young, not usually apply to adults. Based on shape free texture and learning is introduced. After a face is detected, face alignment extracts facial landmarks by which the face image is complex. Furthermore, of local feature, Haar-like feature is extracted from the shape free texture. Euclidean distance method is used for training child or adults classifiers, and will try to propose a transformation of organ identity is characterized by the angle of approach. Through experiment it is shown that, more efficient for discriminating pre-school child and youth aged people. The research method for all age groups total recognition rate is up to 85.02%. Unfortunately this research method identify middle-aged population less effective, but identify pre-school age and young people quite effective, the highest recognition rate is 98.37% above .

並列關鍵字

Face age ； Haar-like ； Euclidean distance.

參考文獻

[1] N. Ramanathana, R. Chellappa, and S. Biswas, “Computational Methods for Modeling Facial Aging: A Survey,” Proceedings of Journal of Visual Languages and Computing, vol. 20, no. 3, pp. 131-144, 2009.