Dynamic 3D-Vision

摘要

Dynamic 3D-machine-vision has been realised, but fast and precise 3D-acquisition requires urgent improvements. This paper describes proved and promising new ways to achieve fast and precise three-dimensional capture of spatial scenes, finally enabling the long awaited "3D-vision chips". In contrast to 3D-imaging, the well-known 2D-imaging procedure evaluates only intensity projections of a scene and loses the depth information of 3D-objects. Dynamic 3D-vision captures all 3D-surface points of the objects. Fast 3D-vision is one of the most challenging tasks, e.g. in automotive safety, in robot vision, in safety and surveillance. A simple solution has been found and investigated in the PMD-principle and the photonic mixer device (PMD), providing a range of powerful contributions to dynamic 3D-vision. To gain the depth information it applies time-of-flight (TOF) of reflected signals with respect to modulation interferometry. PMD combines both detection and the first step of signal processing like mixing or correlation within a 3D-pixel, providing grey level, pixel coordinates x and y, and TOF=2R/c. This paper illustrates different implementations of PMD-pixels as a basis for new 3D-cameras.