English  | Deutsch | Français
 
01.03.2005
NI Vision Debuts New Geometry-Based Matching Tool
For years pattern matching has been a machine vision staple in applications that detect, count, align, or inspect objects. National Instruments pattern matching is based on highly optimized cross-correlation algorithms that use stable pixel regions and edge information within an image to accurately locate matching templates down to one-tenth of a pixel and one-tenth of a degree. However, you may have difficulty using cross-correlation-based pattern matching to match templates that are overlapping or scaled to different sizes.

With the new Vision 7.1 Development Module, you can use a new matching tool called geometric matching to solve these more complicated vision application challenges. Geometric matching is a powerful algorithm for finding and sorting overlapping objects that are only partially visible, for matching components that contain distinct geometric information, and for guiding robots when objects in the field of view change scale as a camera moves closer to and further from them.

Pattern matching and geometric matching are both two-step processes. The first step is to learn a template and the second step is to search for a match to the template within an image. In correlation-based pattern matching, the learning step extracts large areas with similar pixel values and sharp edges to create a simplified model that speeds up the search. Geometric matching also builds a model of the template in the learning step, but rather than look for stable pixel regions and edges, it looks for 2D shapes such as circles and rectangles, linear primitives such as parallel lines, and other distinct geometric features such as corners and curves. The second step in both algorithms uses the abstracted templates from the learning step to search for a match in the image.

These two algorithms cover a wide range of matching applications and scenarios. Correlation-based pattern matching is effective in general matching applications for which features are defined by their grayscale values or textures. Examples are prevalent when inspecting consumer packaging and labels for cosmetic defects. You cannot easily define these labels, which often contain complex patterns or many shades of gray, by geometric shapes. Geometric matching works best when you are defining objects by geometric components such as stamped metal parts moving down an assembly line. With geometric matching, you can identify, locate, and sort these metal parts even if they are overlapping, dirty, or reflecting glare.


Learn more about NI vision algorithms and request an evaluation copy of the NI Vision 7.1 Development Module.

 

01.03.2005
National Instruments Names Select Alliance Partners for 2005

01.03.2005
NI Vision Debuts New Geometry-Based Matching Tool

08.02.2005
National Instruments Expands Suite of Camera Link Image Acquisition Boards

21.10.2004
Sony announces a new VGA IEEE1394b camera (XCD-V50)

19.10.2004
National Instruments today announced the industry’s first PCI Express-based image acquisition board

14.10.2004
Sony introduces full range of CameraLink-compatible video cameras

04.10.2004
New NI Vision Builder AI Tools Inspect, Identify Parts

13.07.2004
National Instruments Introduces Industry’s Fastest Embedded PXI Controller

17.05.2004
QUALIMATEST on the TV

05.05.2004
SONY launches new Fire-Wire (IEEE-1394) digital color cameras

04.05.2004
The new Compact Vision Systems CVS-1456 is available

20.01.2004
The new frame grabber Fire Wire (IEEE-1394) pour PXI is available

04.11.2003
The new Compact Vision Systems CVS-1455 is available

29.09.2003
QUALIMATEST is certified by CSIA

30.06.2001
QUALIMATEST is Select Integrator

design by Seeaytch Advertising Network & ddellaro.ch