雙攝影機追蹤系統

Abstract

為了能追蹤移動物體並且紀錄特寫鏡頭做為備查，我們實作一個雙攝影機追蹤系統。這個系統可以分成兩個階段。第一階段，目標追蹤子系統從視訊攝影機讀取視訊資料，然後擷取前景，偵測及追蹤物體。我們使用背景學習、變化偵測、變化分類和前景分割的方法來擷取前景。我們使用輪廓和物體追蹤技術來偵測及追蹤物體。第二階段，座標轉換子系統使用planar homography mapping的方法轉換物體在視訊攝影機的座標到PTZ的座標。為了讓PTZ能追蹤移動物體並且給予特寫鏡頭，座標轉換子系統根據物體的座標及大小，自動地調整 PTZ的參數。

In order to capture a high resolution view of interested target region, we implement a dual camera tracking system that has two stages. In the first stage, the input to a target tracking subsystem is video streams from a single web camera. The subsystem analyzes the video content by extracting the foreground from the background, detecting and tracking the objects. The foreground is separated from the background by using background learning, change detection, change classification and foreground object segmentation methods. The objects are detected and tracked by using contouring and blobs tracking techniques. In the second stage, coordinate transformation subsystem transforms the coordinate of object on webcam coordinate system to PTZ coordinate system by using planar homography mapping method. In order to capture a high resolution view of interested target region from PTZ, the coordinate transformation subsystem adjusts automatically the PTZ parameters based on the coordinate and size of the object on PTZ coordinate system.