目前的監控系統，由於取像設備成本越來越低，使得日常生活中取得設備也變的容易。有別於以往只能提供事後資訊的錄影監控，利用電腦視覺的方法，能讓監控系統可以自動的分析影像，以達到即時警報與追蹤的目標。 本論文提出一個使用PTZ攝影機追蹤限制區域非法入侵者的系統，不同於過去限制區域非法入侵偵測只能侷限在固定視角，限制區域範圍太小的限制，利用上衣色彩為特徵判斷進入限制區域的人員是否合法，再使用PTZ攝影機的旋轉視角功能持續追蹤非法入侵者。首先利用背景相減法偵測出目標，使用一個分割演算法將目標的上半身區域分割出來，針對目標上衣區域色彩與預先設定合法上衣色彩資訊比對，判斷出是否為非法入侵者，若是非法入侵則開始追蹤目標物，目標物超出追蹤畫面就使用旋轉功能持續追蹤。 實驗結果證明，本論文所提出的系統是可以使用在較大的限制區域進行非法入侵者追蹤。 Unlike the traditional surveillance systems which only consist of one or several capturing devices, in this thesis we incorporate the PTZ camera with the computer vision techniques. By doing so, the surveillance system can analyzes the moving objects automatically and provide real-time alerts if necessary. In this thesis, we propose a particle filtering based method to identify and track illegal invaders in restricted area using a PTZ camera. The advantage of our method is that can track illegal entrants in wider space, unlike the detecting illegal entrants in restricted area over the past, which is limited to a fixed perspective, the scope of the restricted area is too small constraints, First of all, we employ a robust background subtraction method to detect the moving object. Then we use the watershed segmentation algorithm to extract the upper body of the moving object. Afterwards we compare the color features of the upper body with the template to determine the legality of the entrant. Once the illegal entrant is detected, the system will track the moving object using PTZ camera. Since no background information can be obtained after panning the camera, we employ the particle filtering technique to cope with this problem. Experimental results show the feasibility and validity of the proposed method under several environments. The proposed method provides satisfactory results which prove that the system can be applied to some places such as post office or storage.
