多台攝影機監視系統下的多角度人臉偵測

Abstract

在本篇論文中，我們提出一套應用於多台攝影機之多角度人臉偵 測系統。此系統可根據多攝影機擷取的影像偵測出影像中的人臉位 置，並得到在三維空間中人臉方向的鳥瞰圖。有別於以往的作法，我 們並不在二維的影像中直接作搜尋與偵測、或是將這些在二維的影像 中的偵測結果投影到三維空間中作結合；反之，我們的系統直接在三 維空間中進行搜尋，並將三維空間投影回二維影像中加以比對處理， 以判斷在這個三維空間中是否存在著某種方向的人臉。這樣的做法使 得我們的系統可以有效地結合多攝影機中的二維影像資訊，並可避免 以往因在二維影像中的錯誤偵測所導致的不明確資訊整合。

In this paper, we propose a multi-view face detection system, which is capable of detecting all targets’ faces in the given images and is able to illustrate the bird-eye view direction of each face in the 3-D space in a multi-camera surveillance system. Unlike existing approaches, the proposed system does not directly detect targets over the 2-D image domain nor project the 2-D detection results back to the 3-D space for correspondence. Instead, our system searches for the targets over small cubes in the 3-D space. Each searched 3-D cube is projected onto the 2-D camera views to determine the existence and direction of human faces. This approach can help us to efficiently combine 2-D information from different camera views and to suppress the ambiguity caused by 2-D detection errors.