[[alternative]]Design and Implementation of Moving Target Tracking System with Lens Control

碩士工業教育與技術學系[[abstract]]本論文是利用影像處理的方法來追蹤移動目標物及控制攝影機鏡頭的對焦(focus)、光圈(iris)及變焦(zoom)。影像伺服追蹤的方法是利用攝影機擷取移動目標物的一連串動態影像，由相鄰兩張影像做差分運算，將移動目標物與背景分離。同時將正、負差分影像分別做水平及垂直的投影運算後再做交叉相關運算，可以求出移動目標物的移動速度和移動向量去控制pan及tilt伺服馬達，使攝影機鏡頭的光軸對準目標物的中心，完成即時視覺影像追蹤。控制攝影機鏡頭的對焦、光圈及變焦，本論文是利用模糊控制的方法來達成。在鏡頭的對焦控制部份，以擷取影像的頻譜來設計輸入歸屬函數，控制攝影機鏡頭的對焦；在鏡頭的光圈控制部份，以目標物影像的灰階平均值和標準差來設計輸入歸屬函數，控制攝影機鏡頭的光圈；另外以目標物的大小、移動速度之變化與目標物外形改變時擷取之影像的可信任度來設計輸入歸屬函數，達成攝影機鏡頭變焦的控制。[[abstract]]This thesis based on image processing algorithms presents a moving target tracking system with lens focus, iris and zoom control systems. The vision tracking approach utilizes a subtraction operation on two sequential images to recognize the moving target in a scene.After the horizontal and vertical projection operates on the positive and negative images, the correlation method estimates the motion velocity to produce the motion vector of the moving target. This information is used to control the pan and tilt servo motors that drive the CCD camera in tracking the moving target in real time. This thesis proposes three fuzzy controllers for controlling the focus, iris and lens zoom. The spectrums of the target image are used to design the focus input membership function for the focus control.The mean and standard deviations of two sets of iris input membership functions from the target image are used for the iris control. The target sizes, the variances in target velocity and the target contour confidences are three sets of zoom input membership functions used for zoom control