標題: | 利用PTZ攝影機實現動態背景中特定物體之鎖定追蹤技術 Real Time Tracking System for Specific Moving Object under Dynamic Background Using Modified Mean-Shift Algorithm |
作者: | 林訓緯 Lin Hsun-Wei 林進燈 Lin Chin-Teng 電控工程研究所 |
關鍵字: | 追蹤系統;主動式攝影機;動態物偵測;tracking system;active PTZ camera;moving object detection |
公開日期: | 2006 |
摘要: | 本論文針對移動中的物體,提出一套基於改良後的Mean-Shift演算法,實現在主動式攝影機上的影像即時追蹤系統。在我們的系統架構下,可以區分成三個部分。第一個部分是以影像為基礎之動態物體偵測模組;第二個部分是將所偵測到的動態物體,進行鎖定以及連續平滑的追蹤;第三個部份則是針對主動式攝影機所開發的控制模組。首先,我們會記錄下所偵測到的動態物體,利用其色彩資訊,使用我們所改良後的Mean-Shift演算法,配合上主動式攝影機的控制,進行鎖定且不間斷的追蹤。因此,本論文所提出之技術不但可追蹤鎖定移動中的物體,使之保持在攝影機可視範圍的中間;並且在鎖定目標物的過程當中,若此目標物的尺寸小於我們所設定臨界值的時候,系統會自動調整攝影機焦距的倍率,以便得到清晰可辨識的影像,大幅提升追蹤的品質。 This thesis presents a study of a smooth tracking system for non-rigidity moving object using modified mean-shift algorithm implemented on an active pan-tilt-zoom camera. In our approach, the system is divided into three parts: moving object detection module, modified mean-shift tracking module including position tracking and size selecting algorithm, and an automatic pan-tilt-zoom camera control system. We detect the moving object by frame difference method plus some image processing techniques, such as erosion, dilation and image projection, and use HSV color space to reduce the illumination affection. We also develop a modified mean-shift algorithm by choosing a specific kernel function. Moreover we adopt a different position tracking and size selecting method for smooth tracking and locking target. |
URI: | http://140.113.39.130/cdrfb3/record/nctu/#GT009412541 http://hdl.handle.net/11536/80672 |
Appears in Collections: | Thesis |
Files in This Item:
If it is a zip file, please download the file and unzip it, then open index.html in a browser to view the full text content.