Clarification of Visual System Rules

[jsteele]

We are working on our robot's vision system and are confused about the rules regarding lighting for the vision system. Are we allowed to put a small ring of lights around the camera to illuminate the vision target? What kind of lights are allowed? Thanks for all your help.

[The Lucas]

The lights (whichever ones you choose to use) can regarded as a legal custom circuit so long as you inhere to all the rules regarding custom circuits, including but not limited to:
R03
R38
R39-F
R43
R50
R59
R62
R63

Our team is experimenting with LED XMAS lights similar to those seen in the Vision Target Tutorial.

[DonRotolo]

A single LED of about 3000 mCd might be suitable for a first experiment. I would avoid white.

At kickoff, they showed their demo using a 2x AA MagLite

[Greg McKaskle]

One thing to consider is that camera sensors have a lower dynamic range than the human eye -- thus the popularity and difficulty of HDR photography.

The green CCT used several years ago will look super-bright-intense green to your eye. To the camera with normal exposure, it will look white with a green edge. The sensor is overexposed and the color of the light saturates. To correct for this, you need to lower the time of exposure, add a neutral density filter -- camera sun-shades, or in some other way lower the amount of light the sensor collects.

The reason I mention this phenomenon, is that if you see a ring of LED color, and a white interior where the retroreflective tape is, it means you have too much light. Since the tape reflects 600X as much light as white paint, this is easy to do.

In the end, it is the placement of the light near the camera that causes the shine, and it actually doesn't take that much.

Greg McKaskle

[Alan Anderson]

Originally Posted by Greg McKaskle

The reason I mention this phenomenon, is that if you see a ring of LED color, and a white interior where the retroreflective tape is, it means you have too much light. Since the tape reflects 600X as much light as white paint, this is easy to do.

Our initial tests of having the robot detect the retroreflector showed exactly this effect. The final bitmap of recognized color displayed a perfect silhouette of the target surrounded by a halo of "match". We put a diffuser over our LED to eliminate the spotlight effect and it works much better now.

[jfox100]

Can you please post the labview vision settings (HSV, RGB, etc) that work best for detecting the retro reflector tape.

Thanks

Gatorzillas, 2936

[Alan Anderson]

The settings will change dramatically depending on the color and intensity of the light source you're using to illuminate the retroreflector.

[jfox100]

Can you post screen shots of the settings that worked for you? For now, we've been using an LED flashlight. Our robot will have a ring of LED's surrounding the camera. With the flashlight, we weren't able to get the settings to exclusively detect the reflector. Did you also change the frame rate? exposure? etc.?

[Greg McKaskle]

The 3.1 update for LV contains an example for 2011 vision. It runs on the desktop and you give it a folder of images to process. On the panel, it lets you change the HSL numbers and interact with other parameters. That may be helpful when used on your own images.

Even more helpful would be to load some images, or connect your camera to Vision Assistant and analyze the reflection to see what color it is. You draw a line through or a rectangle around the color you want to know about and it will tell you what the values are.

If you are still not quite there, choose a photo to upload, and someone here will help you analyze it.

Greg McKaskle