- 學位論文
即時影像追蹤在網路攝影機上的應用
The Implementation of Realtime Image Tracking on IP Camera
摘要
本論文主要目的為在網路攝影機上實踐影像追蹤之技術，當移動物體進入攝影 機拍攝範圍之內，攝影機會隨著物體移動而轉動，以維持移動物體在畫面中央，達 到追蹤影像之目的。 本文主要分為兩個部份，首先為影像處理，以判別物體中心位置，其次為攝影 機馬達之控制，藉由第一部份所判別之物體中心位置，轉換為馬達控制訊號，以控 制攝影機轉動至所需之位置。 影像處理部份又分為幾大步驟，首先為顏色座標之轉換，將RGB之顏色轉換為 YUV，之後做次取樣以進一步減少運算資料量，並且運用影像相減法在連續影像中 找出相異之部分並判別移動物體的中心位置，為了消除背景雜訊更使用型態學濾波 器做雜訊的消除。 本文章提及之所有影像處理運算完全是在嵌入式系統上完成，受限於嵌入式系 統之運算速度緩慢，實作上以節省運算為首要目標，在運算速度與影像追蹤準確性 之間取一平衡，以便達到即時影像追蹤之目的。
並列摘要
This thesis presents the technology of object tracking for network camera. The network camera will track the moving object so as to keep it in the center of camera view. This thesis includes two main topics. The first is to determine the center of the moving object using image-processing technique. The second is the motor control of the camera. The camera will sweep to the right position automatically so as to center the moving object. The main strategies of image processing are summarized as follows: i) converting the RGB color representation to YUV representation to retrieve luminance of captured image; ii) sub-sampling of image to reduce the computation complexity of image processing; iii) Find out the moving part in the continuous images by computing their differences and, then, using the so-called movement energy method to calculate central position of the moving object; and iv) using morphological filtering to reduce background noise.. This project was developed in an embedded system. Because such a system is intrinsically in a lower capacity, reducing the amount of computation needed is important. Furthermore, to get balance between operation speed and accuracy of tracking is another important consideration of this research.