室內環境之人物辨識

Abstract

本篇論文以開發智慧型居家照護系統為目標，對於居住者之居住環境提供監控之服務，而人物辨識則為本系統中的一個重要部份，藉由人物辨識技術，得以監視環境內每個人物的身分。在本篇論文中，我們開發出一個人物辨識系統，由兩個子系統所構成，分別為人臉辨識系統以及色彩模型辨識系統。兩個系統皆使用獨立的攝影機搭配特定的監視角度，並且透過網際網路分享資訊。 人臉辨識系統用以擷取進入者臉部影像進行身分的辨識，其使用一基於特徵空間轉換與標準空轉換的方法進行識別。而在實際的應用裡，並非所有時間的影像皆適合進行人臉辨識，接著我們使用衣著色彩模型辨識系統進行人物之辨識。當人臉辨識系統辨識進入者身分的同時，亦建立進入者之色彩模型，再將色彩模型以及人物身分資訊傳送至色彩模型辨識系統，最後透過色彩模型辨識系統對於房間內每個人物的身分進行識別。

We aim to implement an intelligent home-care system for nursing home purpose, in which person recognition is an important part in the system. In this thesis, we develop a person recognition system which monitors the room by the two sub-systems: face recognition system and dressing color model recognition system. Each sub-system utilizes a camera with the specific angle of view, but the two sub-systems share information with another system by the internet. The face recognition system uses the capturing face image and then recognizes the person when one enters the room. The face recognition technique is based on eigenspace transform and canonical space transform. In fact, not all of the image frames are suitable for face recognition, hence, we also propose the color model recognition technique to recognize a person by the color of one’s dress. Whenever the face recognition system is possible and done, we build this person’s dress on immediately. Then we pass this color model to the color model recognition system. Finally, each person will be identified by the dress color models generated in the system.