Machine-vision software offers many functions that can enhance or extend the capabilities of mechatronic equipment, or make new designs possible. Typical vision functions include 2D and 3D pattern matching, edge detection, counting and classifying objects, contour plotting, character recognition, bar-code reading, and so on. But in most applications, machine-vision software requires a desktop or industrial PC. MVTech Software GmbH, based in Munich, Germany, has adapted its HALCON vision software so it will run on a Beagle Board single-board computer.
The Beagle Board provides a Texas Instruments OMAP3530 processor that uses an ARM Cortex-A8 CPU. The board includes video and audio interfaces and a DSP chip, so it gives developers a lot to work with–all for $149. You can read more about the Beagle Board hardware and software at: beagleboard.org. At this Web site you’ll find information about independent projects, support groups, and discussions.
MVTec made its first performance measurements by running the HALCON 9.0 machine-vision library with the Angstrom Linux operating system on a Beagle Board. That led to the development of the HALCON Embedded software specifically for the Beagle Board. Now, the board can run typical image-processing algorithms such as filtering or blob analysis in a few milliseconds. Sub-pixel measurements require less than a millisecond. (Sub-pixel measurements use intensity information in individual pixels to determine the location of an edge, for example, to a resolution of better that one pixel. Sub-pixel measurements also let vision equipment detect objects that occupy less than one pixel in an image.) For more information about HALCON Embedded, visit: www.halcon.com/embedded.
The HALCON Embedded software includes applications, C/C++ interfaces, the HALCON Image Processing Library, and interfaces for frame grabbers, the devices that capture image signals from a video camera in analog or digital form. If you want to try the HALCON software, MVTec offers several trial versions for a variety of computers and operating systems. (Registration required.)
If you’re not familiar with image processing or analysis, I recommend the book, “Digital Image Processing,” by Gregory A. Baxes, 1994. The book is out of print, but I found many copies available from used-book dealers. You’ll get an gentle (mostly non-math) introduction to image processing. The book came with a 3¼ floppy disk that relied on now-obsolete software, so if you buy a copy without the disk, it’s no loss. For a more detailed treatment, try “Handbook of Machine Vision,” Alexander Hornberg, editor, 2006.
If you would like to know more about machine vision as it related to mechatronics systems, let me know. –Jon Titus