Calibrating vision processing on the field

[Jared Russell]

Originally Posted by Dan252

Changing the parameters for the threshold to match the lighting in the field.

Is there a way to overcome the difference in lighting of the field from our workshop?

Here is the strategy that has worked for 341 using an Axis camera and LED ring(s). We never had to calibrate on the field in either 2012 or 2013 (both years, all of our shots were aimed by our vision system):

1. Turn off automatic white balance on the camera (can be done from the camera's web interface).

2. Get a very short exposure time so the image is mostly dark (from the web interface). You want a very dark image, except for the illuminated regions of the reflective tape. Set the exposure time to something very short. Put the camera in a bright scene (e.g. hold up a white card a foot or two in front of the lens) and then do a "hold" on exposure priority. Experiment with different settings. You want virtually all black except for a very bright reflection off of the tape. This is for two purposes: 1) it makes vision processing much easier (fewer false detections), 2) it conserves bandwidth, since dark areas of the image are very compact after JPEG compression. The camera doesn't know what you are looking for, so it will try to send you the entire scene as well as it can. But if it can't see the "background" very well, you are "tricking" the camera into only giving you the part you need!

3. Take a wide variety of sample images in and around our build space.

4. Design a threshold in RGB or HSL that works well on your sample images.

5. Your camera is now effectively much more sensitive to your LED ring than to ambient lighting, and all dynamic settings on the camera (white balance and exposure) are held constant. You likely will not need to do calibration (but of course YMMV).