I-Corps: Sensors For Relightable Images

工业检测、执法和博物馆保护专业人员都需要捕获真实的对象的数字表示。这些数字表示后来被用于证据和分析，从而可以更有效地生产产品，更快地进行刑事调查，并保护无价的文物。现有的相机可以在单一照明条件下捕获单一视图，3D扫描仪可以捕获形状，但现有设备无法捕获表面反射特性。在该提案中，该团队开发了一种捕获照片的方法，该方法允许在捕获后交互地改变照明。此外，对数据的计算可以揭示肉眼看不到的细节。各种各样的传感器已经在研究实验室和商业上进行了研究。这些包括从根本上测量光强度的相机，以及从根本上测量3D对象几何形状的深度传感器。该提案涉及一种相关但不同类型的传感器，该传感器从根本上测量表面反射率，即表面如何对来自不同入射角的光作出反应。这是重要的，因为对象的表面反射特性可能包含重要信息，即使当用标准相机或深度传感器测量时表面没有明显的细节。除了表面反射率之外，这种新传感器还可以直接观察表面方向的微小变化，当使用三角测量或基于飞行时间的3D深度传感器测量表面形状时，这些变化低于噪声阈值。

Industrial inspection, law enforcement, and museum preservation professionals all need to capture digital representations of real objects. These digital representations are later used for evidence and analysis that allows more efficient product manufacturing, faster criminal investigations, and preservation of priceless cultural objects. Existing cameras can capture a single view under a single lighting condition, and 3D scanners can capture the shape, but existing devices do not capture the surface reflectance properties.In this proposal, the team has developed a method to capture photographs which allow the lighting to be interactively changed after they are captured. In addition, computation on the data can reveal details which are not visible to the naked eye.A wide variety of sensors have been investigated in research labs and commercially. These include cameras that fundamentally measure light intensity, as well as depth sensors that fundamentally measure 3D object geometry. This proposal regards a related but different kind of sensor that fundamentally measures surface reflectance, how that surface reacts to light from different incident angles. This is important because the surface reflectance properties of an object may contain important information, even when the surface has no apparent detail when measured with a standard camera or depth sensor. In addition to surface reflectance, this new sensor can directly observe small changes in surface orientation which are below the noise threshold when measuring the surface shape with triangulation or time-of-flight based 3D depth sensors.