使用Boosting之人臉年齡分類

Face Age Classification using Boosting

吳衍毅(Yen-Yi Wu)

指導教授 ： 陳文雄

國立暨南國際大學/科技學院/電機工程學系/碩士(2011年)

摘要

生物辨識技術中的人臉年齡分類在電腦視覺的領域裡，目前還沒有很好的研究成果，大部分的研究僅能分類出小孩與老人，多類別的人臉分類研究較少。因此，本論文結合人體測量學與生物學兩種論點並且使用Boosting演算法來做多類別的人臉年齡分類。
本論文的資料庫採用FG-NET所提供的人臉資料庫，其中共有82個人，1002張影像。人臉年齡分類的系統架構可分為三個部分，「人臉影像擷取模組」、「特徵萃取模組」和「分類學習模組」。人臉影像擷取模組的部分使用的是Haar-Like特徵並運用英特爾的OpenCV函式庫，而特徵萃取的部分則採用小波轉換進行紋理萃取，最後使用Boosting分類器將人臉年齡做分類。
實驗結果顯示，利用人臉皺紋紋理來分類年齡可以得到不錯的辨識率，其中青少年之前這段期間因為沒有皺紋，故依據人體測量學中的骨骼發展變化來分類，而壯年期開始到老年期這段時間皺紋的變化很明顯，因此得到較佳的辨識效果，以後因為皺紋發展已趨近穩定的狀態，故辨識率略微下滑。

關鍵字

人臉年齡 ； 年齡分類 ； 小波轉換 ； 提升演算法

並列摘要

The face age classification technology of biometric in computer vision field does not have excellent researches at present. The major researches only can classify the child and old person, but are fewer in face of multi-category classification researches. Therefore the thesis combines anthropometric method and biology, and uses Boosting algorithm to classify multi-face age.
In the thesis, all experiments are tested on FG-NET database. The FG-NET database includes 1002 face images from 82 people. The framework of our proposed age classification system has three modules in the following: face detection module, feature extraction module, and classification module. In face detection module, it is use the Intel’s OpenCV library and the Haar-Like features to locate facial tissues. The parts of the feature extraction module use wavelet transformation to extract the texture. Finally, the classifier module is using Boosting algorithm to classify facial ages.
The experimental results show that using the wrinkle to classify age can obtain high-performance recognition rate. Because of juvenile do not have wrinkle, they classify age according to the skeleton by anthropometry. Besides, in the mature stage to the old stage, whose wrinkle changes very obvious. For the reason, it could prove high-performance in these two stages. Moreover, the wrinkle has little change in elderly, and the recognition rate has been lower.

並列關鍵字

Face age ； Age classification ； Wavelet transform ； Boosting

參考文獻

[1] I. Craw, H. Ellis, and J. Lishman, “Automatic Extraction of Face Features”, Pattern Recognition Letters, 1987.

[2] A. Z. Kouzani and F. He, “Common-sense Knowledge Representation and Reasoning, and its Application to Face Detection”, Conf. on Intelligent Engineering Systems, 1997.

[3] B. Moghaddam and A. Pentland, “Probabilistic Visual Learning for Object Representation”, IEEE Trans. Pattern Analysis and Machine Intelligence, 1997.

[4] K. Sung and T. Poggio, “Example-Based Learning for View Based Human Face Detection”, IEEE Trans. on Pattern Analysis and Machine Intelligence, 1998.

[5] J. G. Wang and T. N. Tan, “A New Face Detection Method Based on Shape Information”, Pattern Recognition Letters, 2000.