Surface reflectance acquisition for finished materials - phase 2

成品材料的表面反射率采集 - 第 2 阶段

• 批准号: 522789-2018
• 负责人: Lalonde, JeanFrançois
• 金额: $ 0.79万
• 依托单位: Université Laval
• 依托单位国家: 加拿大
• 项目类别: Engage Plus Grants Program
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=656435
• 关键词: Surface; reflectance; acquisition; finished; materials

Accurately representing the appearance of materials is of paramount importance in many applications. For**instance, finished materials (e.g. hardwood, melamine, etc.) manufacturers must convincingly convey the "look**and feel" of their products to potential customers, which must currently be done with physical samples.**Replacing those physical samples with immersive virtual tours would be greatly beneficial for demonstrations**of interior design, architecture, real estate projects and the like. However, in order for those virtual tours to be**convincing, they must accurately replicate the appearance of the various materials present in the premises.**While systems for capturing rich appearance models exist and have been proposed in the literature, many of the**solutions are either too complicated or restricted to very small material samples.**In this project, we will develop a data-driven capture approach that will overcome these limitations. By**leveraging the current database through deep learning, we will extrapolate the appearance of new materials**using very few images captured in an uncontrolled environment using a single consumer camera. This will**speed up tremendously virtual catalogs updates, strengthening the industrial partner's position as a leader in**software-based visualization tools.

在许多应用中，准确地表示材料的外观是至关重要的。例如 **，成品材料（如硬木、三聚氰胺等）制造商必须令人信服地将其产品的“外观 ** 和感觉”传达给潜在客户，目前必须通过实物样品来完成。用身临其境的虚拟图尔斯之旅取代那些实物样品，将大大有利于室内设计、建筑、真实的房地产项目等的演示。然而，为了使这些虚拟的图尔斯是令人信服的，他们必须准确地复制的各种材料的外观存在的处所。虽然存在用于捕获丰富外观模型的系统，并且在文献中已经提出，但许多解决方案要么过于复杂，要么仅限于非常小的材料样本。在这个项目中，我们将开发一种数据驱动的捕获方法，以克服这些限制。通过深度学习 ** 利用当前的数据库，我们将使用单个消费相机在不受控制的环境中拍摄的极少数图像 ** 来推断新材料的外观。这将大大加快虚拟目录的更新速度，加强工业合作伙伴在基于软件的可视化工具中的领导地位。