以俯視式環場電腦視覺及行動裝置作擴增實境式室內導覽

Abstract

本論文提出了一個結合電腦視覺及擴增實境技術在行動裝置上使用的室內導覽系統。此系統以在室內環境天花板上安裝的魚眼攝影機作為基礎硬體架構。在人物定位方面，提出了一個以電腦視覺為基礎的方法，藉由分析魚眼影像來偵測使用者的活動資訊。為了得到影像中人物的真實空間位置，我們也提出了一個空間映射的方法，來進行影像座標與真實空間座標的轉換。此外我們也整合了三項技術來進行人物方向的偵測，分別為(一)分析使用者的移動路徑、(二)利用行動裝置上的方向感測器、以及(三)藉由行動裝置上所貼一長條色彩標記，在魚眼影像中分析該標記來進行方向偵測。另亦提出一適用於室內路徑的規劃方法，藉由分析建築平面圖來得到障礙物區域，並以此為基礎得到障礙物迴避方向來進行路徑規劃。伺服器會將導覽資訊傳送至行動裝置上的使用者端，此資訊包括了定位資訊、周遭環境地點及導覽路徑。使用者端接收到的導覽資訊會被覆蓋在行動裝置影像中對應的真實物件上，來提供擴增實境導覽介面。此外本研究也提出了一個方法來估測行動裝置上攝影機的可視角，並以此建立一個轉換矩陣來將真實空間中的點轉換到影像平面上。最後，實驗結果也顯示出了本研究所提出方法的可行性。同時，定位資訊的精確測量結果也顯示了此系統在提供精確導覽資訊的能力。

When people visit new indoor places or complicated indoor environments, there usually needs a navigation system to guide them to desired destinations. In this study, an indoor navigation system based on augmented reality (AR) and computer vision techniques by the use of a mobile device like an HTC Flyer or an iPad is proposed. At first, an indoor vision infra-structure is set up by attaching fisheye cameras on the ceiling of the navigation environment. The user’s location and orientation are detected at a server-side system, and the analysis results are sent to the client-side system. Furthermore, the server-side system also sends the surrounding environment information and the navigation path to the client-side system, which runs on the user’s mobile device. The client-side system then displays the information in an AR way, which provides clear information for a user to conduct the navigation. For human localization, a vision-based localization technique is proposed, which analyzes images captured from the fisheye cameras, and detects human activities in the environment. In order to transform coordinates of image points into the real-world space, a space-mapping technique is proposed. Furthermore, three techniques are integrated together to conduct human orientation detection effectively. The first is analysis of human motions in consecutive images. The second is utilization of the orientation sensor on the user’s mobile device. The last is localization of the color edge mark attached on the top of the mobile device using omni-images. These techniques are integrated together to provide a reliable human orientation detection system. A path planning technique for use to generate a path from a spot to a selected destination via the use of an environment map is also proposed. The environment map is constructed from a floor plan drawing of the indoor environment. An obstacle avoidance map is created from the floor plan drawing, which is used to determine the avoidance direction when a path collides with an obstacle in the environment. Finally, the navigation information is overlaid onto the image shown on the mobile device to provide an AR navigation interface. A method for estimation of the field-of-view of the camera on the mobile device is proposed. The field-of-view is used to construct a transformation matrix, by which real-world points can be transformed into the screen plane, so that the navigation information can be overlaid onto the corresponding real-world objects in the images to accomplish the AR function of the system. Good experimental results are also presented to show the feasibility of the proposed methods for real applications. Precision measures and statistics showing the system’s effectiveness in producing precise data for accurate visiting target displays and environment navigations are also included.