基於多重平面與軸線分析之即時多人定位系統

Abstract

在以視覺為基礎的人物定位與追蹤的研究中，人物遮掩是一個重要且具挑戰性的研究課題。而在此一研究計畫中，我們預計提出一多攝影機與多重參考平面的人物定位方法。先前被提出的方法是藉由將多個視角影像中的前景資訊投影至各參考平面來確認空間中不同高度的參考平面上是否有人物存在，因此比起僅使用單一參考平面，將能夠有效地處理人物遮掩之問題，然而這將使得計算量隨著參考平面的數量而大幅增加。為了減低上述投影所需之計算，本計畫擬進一步地將影像中的人物模擬成軸線，如此，在各高度參考平面上的人物定位將僅需計算軸線的交點，而能夠大量地減少先前的作法中需將前景資訊投影於多重平面的計算量。接著我們對這些交點進行分析後，將不同平面的交點進行連線即可形成3D的軸線（假設對於任意兩視角便計算出一條軸線）。這些軸線經過品質的評估，並淘汰掉不合適的軸線後，依據分群的演算法被分為數群，再依照各群內的3D軸線整合的結果推算出人物的位置及身高。然而，當人物間發生了嚴重遮蔽的情形時，我們將可能無法僅藉由上述的分析就可以獲得完成正確的人物定位，有時候甚至於可能失去所追蹤的人物。此時我們將透過考慮各視角影像中影像的前景資訊、顏色資訊及再加上時間的資訊，來快速地回復遺失偵測人物所在的位置。基於上述研究構想，在此二年研究計畫中，各年度將深入研究的主題包括： 第一年：快速與有效的人物影像軸線估計與其3D軸線之建構，包括研究更有效的多重參考平面建構方法。 第二年：多人遮蔽環境下的人物軸線估計與即時多人定位，包括偵測人物遺失時之快速回復方法。 透過本計畫的執行，我們期望由人物二維與三維軸線的角度切入，能夠對於人物定位的遮掩問題能夠提出有效且快速的解決方案，完成遮掩情況下的快速人物定位之方法建構與理論分析，並發展出一個實用的多攝影機整合監控系統。

Occlusion has been an important and challenging task in vision-based people localization and tracking. In this research project, we expected to propose a people localization method based on multiple cameras and reference planes. Some methods have been proposed to check the existence of people at reference planes of different heights by projecting image foreground from multiple views to these planes; such approaches can deal with occlusions better than using only a single reference plane. In order to reduce the amount of calculation due to image projection, especially for a large number of reference planes, this project aims to estimate a major axis for each person in different images and only project these axes to reference planes so that the computation time can be greatly reduced. For the subsequent localization process, these intersection points are analyzed and integrated to form some 3D axes (assuming one axis from any two views), and these 3D axes are then clustered and integrated to reconstruct the locations and heights of people in the scene. However, the above analysis might not accomplish correct people localization all the time, e.g., when serious occlusion occurs. Therefore, additional information will be considered such as information of foreground region, its color, and also its time-varying properties in multiple views to recover the miss detected (or lost) people. Thus, in this two-year research project, the following research topics will be carefully investigated: 1st Year: Efficient and robust estimation of foreground image and 3D axes of people, including more effective construction of multiple reference planes. 2nd Year: Estimation of multiple people axes under occlusion situation and real-time multiple people localization, including quick recovery for miss-detected persons. Through the execution of this project, we expect to establish an effective and efficient method to achieve real-time multiple people localization under occlusion via some theoretical analysis of 2D/3D axes of people. We are also looking forward to the development of a practical multi-cameras surveillance system in the near future.