Customer Solutions
NI Vision and LabVIEW Extend the Limits of Machine Vision Accuracy
Author(s):
Shahzad Sarwar, Averna Technologies, Inc. ; null null, null
Industry:
Research
Product:
LabVIEW, Vision
The Challenge:
Creating a visual strain measurement application, which would require a camera with 400 mega-pixels, to image a 20 mm wood sample and trace its deformation with a resolution of 1 micrometer when a mechanical load is applied.
The Solution:
Using an innovative application of National Instruments software to obtain a sub-micron tracing resolution while still maintaining a large field of view and using standard off-the-shelf hardware.
Time Evolution of Strain
|
Creating a Vision System for Researchers
Our customer, a university group involved in material research, needed a challenging vision system to measure sample strains with sensitivity of 50 microstrains or better. The group had been doing the research with traditional strain gage sensors. The research scope required the sample characteristics to be measured in an environmentally controlled chamber with operating conditions prohibitive for strain-gage use. The usual calibration requirements and noise vulnerability of strain gage measurements were additional incentives for the researchers to look for a visual strain measurement method.
The visual-measurement challenge was the maintenance of the large field of view and high tracing resolution. We could use low-cost machine vision cameras, provided we made efficient use of image correlation to trace feature displacements with sub-pixel resolution. It turned out that we could implement a PC-based solution using the National Instruments Vision Development Module that was well suited to perform the image correlation tasks needed for the application.
System Design
A PCI-1428 frame grabber was used to configure and acquire images from a progressive scan CCD camera. The monochrome camera had 1400 by 1000 pixels and could provide a maximum of 24 frames per second. Camera gain, shutter speed, and image transfer were all controlled through the PCI-1428 housed in a desktop running Windows 2000 and NI LabVIEW.
The camera, optical, and illumination hardware were mounted on a tripod, so the researcher could conveniently locate the imaging components close to the rest of the test setup. With the LabVIEW application, the researcher specified the acquisition parameters including frame rate and type of test to be performed. The researcher followed the guides to verify the image focus and quality and calibrate the vision pixels in physical distance. The system provided an intuitive and interactive user interface to define the image features for which tracing and relative displacements needed to be measured.
When researchers began a test by activating the load, they acquired and processed images to obtain a live strain development of sample under study. A typical test had a duration of a few minutes for image acquisition. Researchers could use the software to see, in real-time, the evolution of strain along axial and vertical directions.
At the end of the test, researchers could store all image data and displacement measurements to disk for offline analysis. The LabVIEW image analysis application also worked in offline mode during which images were loaded from a previously logged session of acquisition. The analysis and display capabilities of the system were identical for disk data or live acquisition.
The application started with the acquisition of an initial reference image. As the researcher defined the list of image features to be traced, nodes, a pattern was created for each node, and the initial location of these were memorized. With every new image acquired, a pattern matching and displacement calculation was performed for every node. The software logged node locations in an Excel file and stored test images on the disk for future use.
Obtaining a Resolution with Higher Sensitivity
We validated the visual strain measurement system with performance tests in which the sample was moved by a known amount between successive images. We found the vision system to reproduce the displacement measurements with excellent resolution that approached one-twentieth of pixel size. The implemented solution demonstrates how a LabVIEW-based visual strain measurement system can be used successfully to measure strain in the region of 50 strains or better over a large field of view. The solution is vibration and noise insensitive and very easy to configure and use. Based on off-the-shelf machine vision components, this high-resolution tracing solution remains cost-effective and easily extendable.
For more information, contact:
Shahzad Sarwar
Director of Industrial and Real-Time Solutions
Averna Technologies, Inc.
275 Slater Street, Suite 900
Ottawa ON K1P 5H9
Tel:(613) 230-0283
Fax:(613) 236-3754
E-Mail: shahzad.sarwar@averna.com
http://www.averna.com
