Object Shape and Surface Reflectance Modeling Based on 3D Measurement

基于 3D 测量的物体形状和表面反射率建模

• 批准号: 13480098
• 负责人: YOKOYA Naokazu
• 金额: $ 5.63万
• 依托单位: NARA INSTITUTE OF SCIENCE OF TECHNOLOGY
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (B)
• 财政年份: 2001
• 资助国家: 日本
• 起止时间: 2001 至 2002
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-13480098/
• 关键词: Virtualization of Real Objects; Object Modeling; Surface Reflectance Estimation; 3D Measurement; Mixed Reality; Dynamic Image Analysis; Computer Vision

We have obtained the following results in this project.1. Simultaneous Acquisition of Registered 3D Shape and Surface Color of Real Objects Using a RangefinderWe have developed a method for simultaneously measuring 3D shape and surface color of real objects. First, the minimum number of light sources and their positions are determined based on 3D shape of an object subject to the following conditions: (1) only diffuse reflectance components are observed at least once at each surface point, (2) strong specular reflectance components are observed at least twice at each surface point. The object is then measured under selected illumination conditions.2. Dense Estimation of Surface Reflectance Parameters of Objects from Registered Range and Color ImagesWe have developed a method for densely estimating surface reflectance parameters from registered range and color images. The method is based on substituting acquired range and color data into the Torrance-Sparrow model of surface reflection. … More First, the surface reflectance parameters are estimated without considering surface interreflections, and then the method is modified to consider interreflections. The new method is based on radiosity and has been proven to obtain more accurate results than the previous method. The proposed method has been successfully applied to interactive virtual relighting of virtualized real objects under arbitrary illumination conditions.3. 3D Surface Modeling by Registering and Integrating Multi-View Range ImagesWe have first proposed a method for acquiring multi-view range images from multiple sequences of images, which is based on applying a shape-from-motion technique to multiple long sequences of images captured by a moving video camera. 3D models with surface textures are then obtained by voting multi-view range images into a voxel space. Geometric registration of multiple range images is achieved by measuring a number of feature points in the scene in advance that appear in adjacent range images. Less

在这个项目中，我们获得了以下结果：利用测距仪同时获取真实物体的三维形状和表面颜色我们开发了一种同时测量真实物体的三维形状和表面颜色的方法。首先，根据物体的三维形状确定光源的最小数量及其位置，并满足以下条件：(1)在每个表面点上仅观察漫反射分量至少一次，(2)在每个表面点上至少观察到两次强镜面反射分量。然后在选定的照明条件下测量物体。从配准距离和彩色图像中密集估计物体表面反射率参数我们开发了一种从配准距离和彩色图像中密集估计物体表面反射率参数的方法。该方法基于将获取的距离和颜色数据代入表面反射的Torrance-Sparrow模型。首先在不考虑表面间反射的情况下估计表面反射率参数，然后对该方法进行改进，考虑了表面间反射。新方法是基于辐射度的，并已被证明比以前的方法获得更准确的结果。该方法已成功应用于任意光照条件下虚拟真实物体的交互虚拟重照明。本文首先提出了一种从多序列图像中获取多序列图像的方法，该方法基于将运动形状技术应用于由运动摄像机捕获的多长序列图像。然后通过将多视图范围图像投票到体素空间中获得具有表面纹理的3D模型。多幅距离图像的几何配准是通过预先测量场景中出现在相邻距离图像中的多个特征点来实现的。少

项目成果

Tomokazu Sato: "3D modeling of an outdoor scene by multi-baseline stereo using a long sequence of images"Proc.16th IAPR Int.Conf. on Pattern Recognition. 3. 581-584 (2002)

Tomokazu Sato：“使用长图像序列通过多基线立体对室外场景进行 3D 建模”Proc.16th IAPR Int.Conf。

Takashi machida: "Virtualizing real objects with surface interreflections"Proc.IEEE Int.Conf. on Multimedia and Expo. 1. 833-836 (2002)

Takashi machida：“通过表面相互反射虚拟化真实物体”Proc.IEEE Int.Conf。

佐藤 智和: "動画像からの三次元復元-マルチベースラインステレオ法を用いた屋外環境の三次元モデル化-"画像ラボ. 14・4. 1-5 (2003)

Tomokazu Sato：“运动图像的 3D 恢复 - 使用多基线立体方法的室外环境 3D 建模” Image Lab 14・4.1-5 (2003)

Takashi Machida: "Dense estimation of surface reflectance property by combining multiple illumination conditions"Trans. Of IEICE(D-II). J84-D-II-8. 1873-1881 (2001)

Takashi Machida：“通过结合多种照明条件对表面反射率特性进行密集估计”Trans。

Takashi Machida: "Dense estimation of surface reflectance properties for merging virtualized objects into real images"proc.5th Asian Conf. on Computer Vision. 2. 688-693 (2002)

Takashi Machida：“将虚拟对象合并到真实图像中的表面反射特性的密集估计”proc.5th Asian Conf.