PDA

View Full Version : Fixing Camera "overload"


Crtcalmss
01-19-2006, 08:54 PM
Our team had been having problems with the color registration for the camera. Whenever we got closer to the illuminated target, the color "washed" out on the camera display and the computer couldn't recognize the color green. In addition, we were having significant problems with the camera trying to track the room's fluorescent lights and the light coming in from the windows. A previous thread suggested a UV filter since this kind of camera was sensitive to UV. Our team killed two birds with one stone... We went to Wal-Mart, bought the cheapest pair of dark (grey) sunglasses we could find, removed the lens, and mounted it in front of the camera's lens. We now have a camera that no longer responds to UV, has a nice, dark, background, even in full light, and sees the illuminated target as a nice, rich green color that is easy to track. Some tweaking of the camera controls through the software was needed, but its now working great and it tracks very well and re-acquires quickly when signal is lost.

The sunglasses were in the jewelry department and cost a whopping $3.95. They were a pair of "kids" sunglasses with 100% UV protection.

Hope that helps someone...

devicenull
01-20-2006, 10:44 AM
If you go through the calibration routine in the camera handbook, you wouldn't have had to do this :)

Crtcalmss
01-20-2006, 11:07 AM
The way our team operates, we often leave a lot of the work up to the students... They become the "experts" on their particular system of the robot. According to the programmer, who is in charge of the camera/robot interface, he had already completed this step and was still having significant problems. I will show him your post and remind him to double check!

Thank you very much!

Steve Rugoletti
team 350
Plaistow, NH

devicenull
01-20-2006, 11:40 AM
http://kevin.org/frc/CMUcam2_workbook.pdf
Page 23 begins something like "If the target is white, or mostly white with bits of color around the edge"

june_2008
01-20-2006, 07:52 PM
Our team had been having problems with the color registration for the camera. Whenever we got closer to the illuminated target, the color "washed" out on the camera display and the computer couldn't recognize the color green. In addition, we were having significant problems with the camera trying to track the room's fluorescent lights and the light coming in from the windows. A previous thread suggested a UV filter since this kind of camera was sensitive to UV. Our team killed two birds with one stone... We went to Wal-Mart, bought the cheapest pair of dark (grey) sunglasses we could find, removed the lens, and mounted it in front of the camera's lens. We now have a camera that no longer responds to UV, has a nice, dark, background, even in full light, and sees the illuminated target as a nice, rich green color that is easy to track. Some tweaking of the camera controls through the software was needed, but its now working great and it tracks very well and re-acquires quickly when signal is lost.

The sunglasses were in the jewelry department and cost a whopping $3.95. They were a pair of "kids" sunglasses with 100% UV protection.

Hope that helps someone...

Exactly what was your camera doing before you put on the sunglass lens? Was the tracking light on, but the terminal window showing maximum error?

Crtcalmss
01-21-2006, 11:39 AM
Devicenull - I'll point that out to the student in charge. Its beginning to sound like he didn't read the manual fully, especially if the page starts with a description of one of the exact problems we were experiencing.

June - We had three separate problems that were compounding each other. We were able to capture frames and see the target but the target was all white in the middle and we couldn't get a good lock on the color. The second was that the camera was fixating on other light sources like the shop window and the overhead fluorescent tubes. The third turned out to be a servo that was installed backwards.

The sunglasses I mentioned in the first post fixed problems 1 and 2, but the third problem was discovered after we realized that the software was giving the correct instructions to track the target, but kept losing it because the signal was sending the camera platform in the opposite direction. Everything is now working, but based on Devicenull's post, I think we'll probably eliminate the need for the sunglasses once my programmer actually reads the manual more thoroughly!

We do appreciate the help and concern -
Steve Rugoletti
Team 350
Timberlane HS. Plaistow, NH