High-speed, High-accuracy 3D reconstruction application
ANAFOCUS Posted 07/05/2010
Application Overview
Eye-RIS™ Vision Systems perform complex adaptive acquisition and processing tasks, as well as robust and efficient event-driven readout in order to construct a 3D image of a moving object by processing laser profiles. In-pixel analogue and mixed-signal processing and memory circuitry allows achieving astonishing speed and accuracy within 3D reconstruction.
Eye-RIS™ Vision Systems perform complex adaptive acquisition and processing tasks, as well as robust and efficient event-driven readout in order to construct a 3D image of a moving object by processing laser profiles. In-pixel analogue and mixed-signal processing and memory circuitry allows achieving astonishing speed and accuracy within 3D reconstruction.
Different optics can be combined with light sources to adapt the application to each specific case.
Several interfaces are available within Eye-RISTM Vision Systems, such as USB 2.0, GigE and Camera Link (PoCL).
Algorithm Overview
A laser stripe is generated using a solid state laser with a cylindrical lens yielding a Gaussian power profile.
A laser stripe is generated using a solid state laser with a cylindrical lens yielding a Gaussian power profile.
3D reconstruction is achieved using triangulation: Considering a cross-section of the stripe and taking the first derivative in a noise-free experiment, the point at which the first derivative crosses the Y=0 axe is the peak location. However, the discrete-nature of the image sensor pixel array makes accuracy to be as high as ½ pixel in case no interpolation is used.
AnaFocus’ solution comprises both sub-pixel (up to 1/40pixel accuracy) and low-accuracy (around 1/2pixel accuracy) modes. An example is shown at left: A rounded shape has been scanned using both methods (green dots for the former method and red dots for the later).
The triangulation algorithm flow diagram is shown at right. If no sub-pixel accuracy is required, all processing is performed by in-pixel processing blocks, leading to astonishing operating speed.In all cases, image acquisition, pre-processing and point interpolation (if required) are performed in parallel, leading to a three-staged pipeline processing structure.
Performance
Eye-RISTM v2.1 is capable of analyzing more than 2,500profiles/sec with sub-pixel accuracy (1/40) and over 12,000profiles/sec with ½ pixel accuracy.
The distance range covered by the Eye-RISTM v2.1 in the X axis (laser direction) depends on the lens, which must be selected following the specification of each application. Then, the range and accuracy in all axes, referred to the range in the X axis are shown in the table, along with an application example.
Eye-RISTM v2.1 is capable of analyzing more than 2,500profiles/sec with sub-pixel accuracy (1/40) and over 12,000profiles/sec with ½ pixel accuracy.
The distance range covered by the Eye-RISTM v2.1 in the X axis (laser direction) depends on the lens, which must be selected following the specification of each application. Then, the range and accuracy in all axes, referred to the range in the X axis are shown in the table, along with an application example.
