利用智慧型手機及特徵比對技術在室內藝術雕像展覽作擴增實境式導覽和資訊檢索

Abstract

在一個藝術品展覽館中，當參觀者想要欣賞某藝術品時，常會不知藝品所在位置或其樣貌而找不到，因此感到挫折。對此本研究提出一個應用於室內雕像展覽館的擴增實境系統來協助參觀者找到感興趣的雕像。此系統主要功能有二，一是根據參觀者所選擇的目的雕像，規劃出最短距離的路線，引導參觀者找到該雕像；二是根據參觀者所拍攝的雕像，在參觀者的智慧型手機上提供完整的雕像資訊。此二個功能皆是以擴增實境式的方式來呈現。 為了要能夠辨識本質為3D物體的雕像，本研究提出了一個具有技巧性的「影像特徵窮舉比對」方法，此方法透過將每個雕像不同角度的「加速穩健特徵(SURF)」擷取出來並建入資料庫，再根據參觀者所拍攝雕像影像的特徵與資料庫裡的特徵進行比對。 另外，本研究提出了一個利用「最近鄰居法(KNN)」(K為2)的方法來將錯誤的配對去除，進而提高特徵比對的準確度。本研究也採用了「K維樹(K-d Tree)」來加速特徵比對的過程。此外，本系統採用主從式的架構模式(Client/Server model)，以提高整體系統使用上的反應速度。 為了定位參觀者的所在位置，本研究也提出了一種方法，讓參觀者拍攝一座附近雕像，再藉由Server端程式利用SURF辨識所拍影像中的雕像，並從資料庫中讀取其位置，傳送回Client端，當成參觀者的當前所在位置。當參觀者的目前所在位置已知後，本研究接著以戴克斯特拉(Dijkstra)演算法來規劃出一條最短路徑。為了能以更直覺的方式引導參觀者，本研究亦提出以擴增實境的方式，產生一「指引箭頭」(guidance arrow)，顯示在參觀者的智慧型手機上，來引導參觀者，而且此箭頭會根據參觀者智慧型手機內建的電子羅盤的數值來動態地更新所指引的方向。 最後，本研究的實驗結果良好，證明本研究所提系統與方法確實可行。

Finding an artwork in an artistic exhibition is sometimes frustrating because a visitor usually does not know the location of the work they are looking for and they might even not know how it looks like. To solve such a problem, an augmented reality (AR)-based guidance system for use in indoor artistic-statue exhibitions is proposed to help the visitor find an artwork they are interested in. The system has the functions of guiding a visitor to a statue that he/she selects as the target; and providing the detailed information of the taret statue by overlaying the information onto the screen of the visitor’s smart phone. Both functions are provided in an AR way. In order to recognize a statue, which is a 3D object, a skillful exhaustive image-feature matching method is proposed, which “learns” an input statue by capturing the images of the statue from different viewing angles as a database, and recognizes the input statue by matching its image against the database via the use of speeded-up robust features (SURFs). Also, a method using two nearest neighbors of each of the matched keypoints in the SURFs is proposed to refine the resulting matches and improve further the matching accuracy. Moreover, a K-d tree algorithm is adopted to speed up the feature matching process; and a client/server model is used to improve the responsiveness of the proposed system. In order to locate the visitor, a method that recognizes a statue near the visitor and takes its location as the visitor’s location is proposed. Specifically, the visitor takes an image of a nearby statue and the image is sent to the remote server for statue recognition. The location of the recognized statue is then retrieved from the database for use as the visitor’s location. After the location of the visitor is known, a method based on the Dijkstra algorithm is proposed, by which the system can plan a suitable shortest path from the visitor’s current location to the location of the target statue. Also, in order to guide the visitor in an intuitive way, an AR-based guidance method which augments an arrow on the screen of the visitor’s smart phone is proposed. This guidance arrow is updated dynamically based on a technique that utilizes the reading value of the built-in digital compass of the visitor’s smart phone. Good experimental results showing the feasibility of the proposed system and methods are also included.