智慧型監控系統一直是近年來相當熱門的研究項目，其目的大都是為了達到目標物之偵測與追蹤為主。監控系統依照環境不同可分成室內及室外兩種，室外監控系統其重點大都著重在行為分析上，而室內監控除了行為分析，有時還加上了身份辨識的目的。 大部分的室內環境監控系統，大都只用一台或多台相同的攝影機來取像，在此種情況下有時並無法發揮其最大的功效，本研究利用兩種不同功能的攝影機：環場攝影機(omni-camera)及PTZ攝影機建構室內環境監控系統，利用兩種攝影機不同特性的優缺點作互補，以期達到最好的效果。 在本篇論文中，並未使用到許多相當複雜的算式或演算法，大多利用基本的影像處理方法來達到監控的目的。在前景物偵測部分，利用對環場攝影機之影像作背景相減來找到目標物位置並持續追蹤，而背景建立的方法則使用漸進式的背景影像建構法，以期能快速建立並更新背景影像。在臉部正面判定部份，則是利用膚色資訊及幾何條件在PTZ攝影機之影像中尋找人臉位置，再利用人臉特徵中五官在臉部區域之相對位置判斷出人臉方向。實驗結果顯示本系統對於以臉部辨識為目的之室內監控系統有相當的可行性。 The development of unsupervised surveillance systems always attracts the attention of many researchers due to its importance in several applications. The main purpose of most surveillance systems is to detect and track the target. It can be classified into two categories according to the environment, which are indoor and outdoor surveillance systems. Most of outdoor surveillance systems focus on behavior analysis, whereas indoor surveillance systems emphasize not only on behavior analysis but also on identity recognition. Almost all indoor surveillance systems only use one kind of cameras to capture the images. However, the performance will not be optimal under this constraint. In this thesis, two different kinds of cameras, called omni-camera and PTZ camera, are utilized to build our indoor surveillance system. These two kinds of cameras can be the complementary of each other to accomplish the task in reaching the best performance. In this thesis, basic image processing techniques are developed to accomplish the surveillance goal. To accomplish the task of detection and tracking, we apply background subtraction method on the images captured from omni-camera to detect the target and track it continually. Progressive method is employed in this part to generate background image with an eye to generating and updating background image quickly. As to the deciding of face orientation, skin color information and the relating geometric features are firstly utilized to find all faces presented in the images captured from PTZ camera. Then, the relative positions of all facial characteristics are used to determine the orientation of faces. Experimental results demonstrate that our devised system is effective in indoor surveillance systems for face recognition purpose.