目前大多數的數位監控系統,都只有側錄動作與畫面分割等基本功能,無法有效的進行目標搜尋及物件追蹤等智慧型分析。因此,本研究以主動式PTZ(Pan-Tile-Zoom)攝影機,運用數位影像處理(Image Processing)技術,連續性地偵測場景中人物,建立一套具有強健性的人物追蹤系統。 本系統針對序列影像做出運動估測與目標追蹤分析,以影像相減法擷取出場景中移動物件的特徵,利用影像前處理技術強化其物件特徵,偵測出該物件在影像中的位置與大小。系統判斷出優先追蹤的特定目標,將其目標的顏色資訊回饋至系統,使用CamShift演算法計算出色彩機率分佈並不斷的疊代收斂後,便可找出該目標的中心及尺寸。最後,配合其PTZ攝影機的控制模組,驅動攝影機進行影像平移與縮放動作,以達到即時偵測與追蹤的效果。
Up to date, most of the digital monitoring systems only provided in the basic functions of recording and frame division instead of performing intelligent analysis, such as searching and tracking objects. Therefore, the purpose of this research is to build up a robust targets tracking system, which apply the technique of Digital Image Processing to continuously detect objects or persons in the targeted scene based on the active PTZ camera. The designed system estimates the motion directions and analyzes the tracking targets in sequential images. Next, the system extracts the features of the moving object based on the method of temporal difference, then uses the image preprocessing method to strengthen detected features of object, and finally measures the position and size of the image. Based on the warning rules, the system will be able to prioritize the tracking targets, and feed the color information back to the system. In the meantime, the CamShift algorithm is used to compute probability distribution of detected color and iterative-converged, the system will find out the center and size of that target. Finally, with the PTZ camera’s active module, the system will drive the camera to pan, tile and zoom the image in order to reach the real-time detection and tracking.