多人物及物件的自動化多機位協同拍攝與輔助取景

Abstract

在一個活動的現場為了拍攝，會需要有許多的攝影師和導播，要控制畫面得靠著攝影機與導播的配合，本篇論文把多個攝影機系統變成可以自動或半自動分配拍攝對象及取景，只要透過事先設定好的參數即可得到需要場景畫面，即使只有一個人也可以同時完成許多事情，在多攝影機系統下的三維定位，並去選擇攝影機進行拍攝，從三維的參考平面上去取得人物資訊，根據人物與攝影機的相對位置，人物的速度去選擇較佳的攝影機拍攝畫面，再根據拍攝規則，去調整攝影機的焦距縮放、左右移動、並且遇到特殊條件時，該如何去應對換手接手攝影機，透過這些規則實現，達到自動化的多攝影機協同拍攝與輔助取景。

Video shooting and activity need multiple cameramen and a director. In this thesis, we design a system in which a multi-camera system can automatically select the view to shoot. The director just operates on the user interface to control all the cameras. We will discuss how to choose the camera to shoot the people in the scene, and our rules to control camera motions (pan, tilt, and zoom). In order to choose the camera, we will track and localize people in the scene. We have designed a set of rules to control the cameras and achieve the automation of coordinated shooting.