Lighting in an identification tasks

[BruceAmmons]

Is the ring light centered on the lens of the camera?  I'm not sure why the reflection is at the top of the image.  Maybe taking a picture of your setup and posting it would help.  A side shot that shows the location of the camera, target, and light would be useful.

Perhaps try moving the light closer to the target.  If the ring light is larger than the target, the reflection will eventually get larger than the target.

If you want to diffuse the light, the diffuser has to be a good distance from the light.  The distance should be greater than the spacing between the LEDs.  If you position the light right, you won't need to diffuse it, though.

On the image you posted, you can clearly read the needle position.  If you set up an ROI in the clear area below the scale, you can easily find the line of the needle.  If you calculate its angle, you can do a direct conversion from angle to value.  There is a vi for measuring dial indicators like this already made for you.  All you need to do is plug in a couple of measurements.

If you are having problems with direct sunlight, you need to make sure your light is brighter than direct sunlight, probably by an order of magnitude.  You could probably create a measurement routine that is robust enough to work even with direct sunlight.  You would need to find the rising and falling edges of the needle.  If sunlight is giving you problems, post an image that shows the problem.

If you want to use two spot lights, you would want to position them to the left and right of the target, at about a 45 degree angle from the surface.  You will get two faint shadows, but the needle should be much darker than the shadows.  Once again, to overcome sunlight they would need to be significantly brighter that sunlight.

Bruce

[max_i]

Hello, Bruce!

Yes, ring light centered with the lens of the camera.
The ring light less than the target.
Location the particles of the system are in the attachment file.

You right. And here is necessary to try with the diffuser. He is located close to the LEDs 1-2 mm (between LEDs 1mm).

With the software just are no any problems. Thanks.
We use a standart function Machine Vision palette.

Apropos of the measurement routine. Do you mean to place the light sensor?
Search the rising and falling edges of the needle you mean software procedure or light quality?
Image that shows the problem:

***

Ie is the same image as in the previous post, only brighter.
After subtraction, the bright area gives this effect.

And what the brightness would be enough to overcome the typical brightness of sunlight?
Which a lighting model is preferable in this case?

Thanks a lot!

Max O.
Engineer and software developer,
TeSLa.

[BruceAmmons]

I see. The camera and light are not perpendicular to the surface of the dial, which is why the reflection is at the top.  Positioning the camera and light perpendicular would not help you anyway, because the light is smaller than the dial.  You would get a reflection of the ring light that would make analysis impossible.

If possible, I would move just the camera to perpendicular.  This will give you the most accurate readings.  If it can't be moved, it will work pretty well where it is.

If you moved the light up a little bit, just above the camera, it would eliminate the reflection.  The lighting would probably be even enough for analysis.

I can come up with a couple of good lighting options.  The first would be two spotlights, positioned 45 degrees to the left and right of the camera.  The second would be a long bar light (about twice the width of the dial) placed horizontally above the camera. The goal is to position the light where it does not reflect into the camera.  Above or to the sides is your best option.  You also want the light coming from as many different directions as possible to eliminate shadows.

In answer to your question, options a,c, and e are all pretty good.  Notice the lights are far outside the object of interest, so the light hits at about 45 degrees.

I was talking about IMAQ Get Meter under Machine Vision.  I haven't used it, but it is designed to read meters like this.  I suspect there is an example.

Don't overexpose the dial.  The picture you posted earlier would work pretty well.  You want grayscale, not binary.

Bruce