EAGER: Combining Sketching and Computer Vision Techniques in Cultural Heritage Applications

计算机图形学已成功地应用于广泛的文化遗产应用，在学科专家的研究和为公众教育准备展品方面都证明了这一点。每年，世界各地都会启动数百个基于计算机图形的文化遗产项目。然而，数字材料的使用仍然要求很高，需要技术专家和学科专家。这一研究项目将对用于向公众传达历史文化信息的展览的发展产生影响。对于学科专家来说，准备这样的展品将更容易，也将有可能创造出更吸引人的信息互动展示。这项研究的目标是改变在交流中创作和使用数字对象的方法。这项研究解决了如何生成令人信服的遗产遗址3D展示的问题，在这些遗址上，可以获得大量的档案照片文件。最近在基于草图的建模方面的创新，特别是心理画布？仅依赖3D笔划而不是全实体的框架将与计算机视觉技术一起开发。Mind Canvas框架将允许用户注释，以便促进应用图像特征检测和束估计方面的最新改进，以从档案图像计算3D点。由于当所有相机特性未知且照片不是从旨在促进3D重建的视图中拍摄时，当前方法经常失败，因此需要进行注释。拟议的系统的输出将是一个3D模型，可以很容易地被临时用户导航。然后，该系统将能够自然地在太空中导航，并浏览实体模型、档案图像和文本数据。耶鲁美术馆正在准备的杜拉·欧罗波斯遗址的文物和图像展览是这项研究的试验台。

Computer graphics has been successfully used in a wide range of cultural heritage applications.It has been proven both in the context of study by subject specialists and in the preparation of exhibits for education of the general public. Each year hundreds more computer graphics based cultural heritage projects are initiated around he world. However, the use of digital material remains demanding and requires technologists as well as subject specialists. This research project will impact the development of exhibitions used to communicate historical and cultural information to the general public. Preparing such exhibits will be easier for subject specialists, and it will be possible to create more engaging interactive displays of informationThe goal of this research is to change the methods for both authoring and using digital objects in communication. The research addresses the problem of generating compelling 3D displays of heritage sites, where large archival photographic documentation is available. Recent innovations in sketch-based modeling, in particular the ?Mental Canvas? framework that relies only on 3D strokes rather than full solids, will be exploited along with computer vision techniques. The Mental Canvas framework will allow user annotation in order to facilitate the application of recent improvements in image feature detection and bundle estimation to compute 3D points from archival images. Annotation is required since current methods often fail when all camera characteristics are unknown and photographs were not taken from views designed to facilitate 3D reconstruction. The output of the proposed system will be a 3D model that can be easily navigated by a casual user. The system will then be able to naturally navigate through space and browse solid models, archival imagery and textual data. An exhibit of artifacts and imagery from the Dura Europos site that is being prepared by the Yale Art Gallery is serving as the testbed for the research.