利用環場攝影機於車內做基於擴增實境技術的室外園區導覽

Abstract

本研究利用架設在車頂上的環場攝影機，偵測空間中的垂直物體，再定位出車子所在位置後，將得到的資訊直覺地顯示於車內的平板電腦上，成為完整的擴增實境戶外園區導覽系統。駕駛者可以利用手持平板電腦看到目前車輛周圍的建築物及其名字的擴增影像。 為了實現整個系統，首先要建立環境地圖，地圖包含了導覽的路徑、周圍要偵測的垂直特徵物及周圍的建築物。系統會對環境進行學習，紀錄垂直特徵物的位置及特徵，並且記錄建築物的位置及名稱等供系統導覽時使用。 接下來，在導覽的期間中，系統會在車子行進時自動偵測車輛周圍的垂直物體，並分析偵測到的特徵物以及學習的資料來對車子定位，在定位完成後，系統可以確定建築物與車輛的相對方位。 再來，系統利用環場攝影機所取得的影像經由轉換後，可以模擬出副駕駛所看出車外的景像，我們將建築的在影像上的位置計算出來，並且疊加此建築物的名稱在這影像上，使用者可以經由這擴增實境的影像，直覺得獲得導覽的資訊。 最後，上述方法的實驗結果良好，顯示本研究所提出的戶外園區導覽系統確實可行。

A new tour guidance system for use in a vehicle driven in park areas based on augmented reality and omni-vision techniques is proposed. A passenger in the vehicle can get tour guidance information from a passenger-view image generated by the system and displayed on a passenger-held mobile device. The passenger-view image simulates what is seen by the passenger through the front vehicle window. Nearby building information is augmented on the passenger-view image for convenient inspection. To implement the system, an environment map is generated in the learning phase after driving the vehicle through a pre-selected path. The map includes along-path vertical-line features in the environment (light poles, building edges, etc.) which can be detected for vehicle localization. A technique for detecting vertical-line features appearing in acquired omni-images and a method for vehicle localization are proposed, by which detected line features are analyzed to compute the vehicle position by the longest common subsequence algorithm. Also proposed are a technique for generating the passenger-view image from acquired omni-images as well as a method for augmenting building information on correct positions in the passenger-view image using the building positions yielded by the vehicle localization process. Good experimental results showing the feasibility of the proposed system are also included.