No software development or programming is required for this inspection automation processor. Traditionally, texture inspection and pattern analysis have been problematic for machine vision processors.

Traditionally, texture inspection and pattern analysis have been problematic for machine vision processors. To remedy this, Sightech Vision Systems Inc.'s (San Jose, CA) PC Eyebot combines selectable imaging functions with a trainable format to provide advanced inspection and recognition. No software development or programming is required to implement this inspection automation processor.

The PC Eyebot works in color or monochrome, with patterns and textures, ignores clutter and confusing backgrounds, and accommodates variable lighting and reflections. It requires limited or no product fixturing and frequently needs no triggers or strobe lights for high-speed production lines.

This new inspection device can function in most places human inspectors work. "It's effective on production lines from fluids to pharmaceuticals, from hardware devices to food processing," says Tom Seitzler, marketing manager at Sightech Vision Systems Inc. "If a human inspector can inspect it, generally, so can the PC Eyebot, but faster and more effectively."

Operator's ease

No software is required for this new system to work. Operators plug a camera, a temporary keyboard, a mouse and a display into the back of the new system. It is menu driven, but the menu's functionality is immediately workable and provides instant feedback for the operator. "An operator can create a feature definition, which can be used to judge an item for acceptance or rejection," Seitzler says. "Operators select the desired functionality, and then look to ensure correct feature definition or processing result. This provides instant feedback. Operators select functionality, they don't have to create or program it."

A user interface provides enhanced and optimized versions of the software tools found in previous-generation inspection processors. These are fully integrated and operational out of the box. The drop-down menu under each title contains numerous algorithms, which may be selected to enhance the feature definition. The effect of the algorithm is immediately viewable on the camera data, allowing the operator to decide if it improves the feature definition.

When the operator has adequate feature definition, he must then ensure that the processing mode selected will provide desired results: pass/fail or absent/present. Like the feature definition modes, the feature processing modes are generally provided for the specific application.

After selecting feature definition and processing modes, the operator trains the vision system by putting known good products in front of the camera connected to the system and it "learns" their features. "Operators can test the training by injecting some bad product examples into the flow of good products," Seitzler says. "A correctly trained system will accept the good products and reject the bad ones. This learning involves correctly selecting feature definition and processing parameters. For training, the vendor assists with this initially; the operator can experiment later if desired."

To interface with the production line equipment, the system sends a 12-volt rejection signal to the output connector on the back of the system's box when bad products are encountered. It also provides a programmable delay to accommodate production line and ejector characteristics. Good, marginal and bad enunciation signals are settable outputs. The operator's PLC may need adjustments to ensure reaction by the operator's system on a timely basis to the PC Eyebot's output signals.

Given the correct selections and a working environment, operators may experiment with the settings and fine-tune their performance as lighting, processing speeds and product changes occur. No software has to be generated or developed to have a functioning inspection system. Operators select the options and view the result.

This new system can perform numerous applications previously unfeasible with older generation equipment such as frame grabbers, video processors, smart cameras and software libraries. With no programming, it can recognize patterns and textures in biological filters; overcome or even make use of reflections in packaging; see slag or poor casting residue in injection molding products and work in variable lighting in liquid flow. It's doing all of this with just the hardware and internal software, no software development, often no triggers or strobes, and generally within a few minutes of the presentation of the problem to the PC operator.

For more information on the PC Eyebot, contact:

Sightech Vision Systems Inc.
6580 Via Del Oro
San Jose, CA 95119
(408) 282-3770
E-mail: [email protected]
URL: www.sightech.com

QUALITY SPECS

1. The PC Eyebot works in color or monochrome, with patterns and textures, ignores clutter and confusing backgrounds, and accommodates variable lighting and reflections.

2. A user interface provides enhanced and optimized versions of the software tools found in previous-generation inspection processors.

3. This new system can perform numerous applications previously unfeasible with older generation equipment.