Fast 3D Imaging

这项研究工作将实现一个“智能相机”（快速3D成像相机，或F3D1），它将实时生成制造场景的全范围地图，几乎不需要计算。同样的相机也将继续作为一个正常的基于强度的相机，可以通过切换单个二进制线从强度模式切换到范围模式。今天的机器人利用摄像头来捕捉需要机器人操作的物体的图像，并试图使用这些图像数据来学习零件的确切位置，识别问题，检查零件，检测接近度等。限制机器视觉在制造环境中使用的问题之一是相机图像是二维的，而必须操作的物体是三维的。三维信息在大多数现实应用中是必不可少的，因此已经探索了许多技术来从二维相机图像中生成深度信息。这些技术都不适合工厂应用，因为它们要么速度慢，成本高，要么不准确，要么无法处理镜面对象。为F3D1相机开发的新技术可能在制造业和其他环境中有重要应用。

This research effort will implement a "smart cameras" (Fast 3D imaging camera, or F3D1) which will generate a full range map of a manufacturing scene in real time with virtually no computation required. The same camera will also continue to function as a normal intensity based camera and can be switched from intensity to range mode by switching a single binary line. Today's robots exploit cameras to capture images of the objects which require manipulation by the robot, and attempt to use that image data to learn the exact position of parts, to identify problems, to inspect parts, to detect proximity, etc. One of the problems which limits the use of machine vision in manufacturing environments is that camera images are two dimensional and the objects which must be manipulated are three dimensional. Three dimensional information is essential in most realistic applications, hence numerous techniques have been explored to generate depth information from 2D camera images. None of these techniques is ideal for factory application because each is either slow, costly, or inaccurate, or fails to work on objects that are specular. The novel technology being developed for the F3D1 camera could have important applications in manufacturing and other settings.