10 tests help evaluate vision sensors

By testing a vision sensor under conditions that closely approximate those found on a production line, you enhance your chances of choosing a sensor that will work well in your application. The test results also help you determine how a sensor will operate when it faces unforeseen problems. Each of the general tests I've described below explain how problems can arise and how to determine whether a sensor can adapt to them. These tests don't provide quantitative results, but the qualitative results can help you "grade" sensor performance.

1. Parts don't always appear where you expect them. Loose fixtures and normal production-line movement or vibration can cause slight changes in a part's position. To test a vision sensor's ability to find a part and locate the features or areas of interest you want to check, rotate the part through 360° and acquire many images. The sensor should still find rotated parts and make the needed measurements. You may not need to rotate the part through a complete 360° circle, but your sensor should tolerate a full 360° rotation.

2. Variable lighting conditions due to ambient light, sunlight, or aging light sources can "confuse" a vision sensor. You can test a sensor under a variety of lighting conditions that approximate its eventual work environment to see how well it performs for each. A sensor should overcome light conditions that can provide uneven light on a part or that can vary over a wide range of intensities. For best results, set up a vision sensor so it works with a source of light that closely matches its characteristics. (Sensor vendors supply compatible light sources or recommend suppliers that offer compatible light sources.) You shouldn't let ambient light from overhead lamps, the sun, or other sources interfere with the vision system. Unfortunately, you cannot always control all possible light sources. You may have to add light baffles or shields to prevent ambient light from reaching your inspection station.

Fig 1 Fiducial marks range from (a) a "golden" mark to (b) marks that show significant differences caused by variations in process steps.

4. Scale variations arise from slight changes in the distance between a vision sensor and a part on a production line. To check how well a vision sensor tolerates some change in the size of an image, move the vision sensor toward or away from the part. The sensor should still find the specified features and make the necessary measurements even if the acquired image appears a bit smaller or larger than expected.

5. Changes in color or size or parts, or subtle process changes can slightly alter the appearance of parts. Present your vision sensor with a selection of parts that show some subtle variations in color, texture, reflectance, size, and so on (Fig. 1). You can determine how well the sensor will handle minor variations in parts without rejecting them.

6. Blur appears in an image when the vision sensor needs its focus adjusted. Sensors can get out of adjustment, so you should test to see how well a vision sensor works with some blur in an image. A vision sensor should perform well despite being out of focus though.

7. Shadows from robot arms or production-line equipment can interfere with some image-processing tasks because they alter how the part appears to a vision system. Project a shadow on parts as you evaluate a sensor to see how it processes images that contain shadowed areas. Some sensors may tolerate shadows while others may force you to eliminate the shadows. If your production line includes a robot or structure that casts a shadow on the part as it undergoes inspection, adjust lights, baffle light sources, or move your inspection station slightly to help eliminate the shadows.

8. Not all production lines provide fixtures that hold parts in place during inspection steps. If your production line can jumble parts so they overlap, set up a similar set of overlapping parts when you test sensors. The sensors should distinguish between the overlapped parts and still extract the information you need. If you can't separate parts on the line, choose a sensor that can handle overlapped parts.

9. Stray light from reflections can appear as bright spots in an image. You can test a sensor's susceptibility to reflections by shining a bright light on your parts as you inspect them. The vision sensor should tolerate even bright reflections on a part.

10. Vibrations due to conveyor-belt motion and the general movement of production equipment can shake parts undergoing inspection. A good sensor should have no problem overcoming changes in position. Let's face it; almost all production lines will have some associated vibrations. Thus, a vision sensor should have the capability to produce meaningful results even when vibrations occur, as long as a part remains in the sensor's field of view.