應用於標記部份遮蔽或不均衡光線環境之可調動閥值擴增實境系統

Abstract

擴增實境(Augmented Reality, AR)是一種新興的人機互動技術，可以即時的將虛擬物件與實際的影像作結合。其中，多數AR系統使用標記(Planar markers)來判斷疊加虛擬物件所需的角度與位置。但是，這些標記會因為不穩定的光線環境，或是因為部分的區域被物體遮蔽，讓系統無法找到標記的正確位置。

在本篇論文之中，我們提出了兩種方式：動態廣域閥值(Dynamic global threshold)以及適應性閥值(Adaptive threshold)來偵測不穩定光線下的標記。使用者可以依據自己的需要，使用其中一種方式來偵測標記。另外，我們設計新的方形偵測法，使用邊界重建(Boundary reconstruction)來重建被部份遮蔽的邊界。接著，我們使用部份模板匹配(Partial pattern matching)演算法，提昇偵測部份遮蔽標記的穩定度。最後，透過一連串的實驗，證實我們的系統能在不穩定的光線環境下或是在標記被遮蔽的狀況下得到較好的偵測結果。相較現有的AR系統，我們的方法在兩倍亮度環境下或當標記的遮蔽面積達到1/3時，仍可以偵測到該標記。

Augmented Reality (AR) is a new human-computer interaction technology which exactly overlays physical image with the virtual objects in real time. Most AR systems use planar markers providing the positions and the angles information to correctly merge the visual objects. However, planar markers are difficult to detect under unbalanced light conditions or while the markers are partially occluded.

In this paper, we propose two methods, dynamic thresholding and adaptive thresholding, to cope with the marker detection under an unbalanced light environment. With the proposed methods, a user can select a proper one to get a better result according to users' requirements and purposes. We also propose new designed quadrangle detection and rebuild the boundaries of occluded markers by reconstructing boundaries, then applying partial pattern matching algorithms. Such a design improves the stability of maker detection when the marker is partially occluded. In the end of the paper, we conduct a series of experiments to demonstrate that our methods have better detection results in an unbalanced light environment or when the marker is partially occluded. Compared to the existing AR toolkits, the results show that the proposed methods can detect markers under double brighter conditions or when a marker is at most 1/3 occluded.