Camera Oversaturation Issues?

[114Klutz]

I'm using Kevin Watsons streamlined code, and am running into massive oversaturation issues. The light from the Green target, along with the glare being reflected off the rack, are the same color, according to the CMUCam2 GUI.

I loaded in the configuration values given by Alan Federman at the SVR regionals, and it still seems to have issues. I've tried common solutions, such as putting a polarized lens over the cmucam. This does not seem to help. I have tried multiple rotatations of the Lens.

[Tom Line]

We never did manange to get the Java Applet to work with the Kevin settings. The labview application was much easier for us.

Are you sure you are using the correct settings with the GUI? Have you changed it to CYK or is it still in RGB?

[Alan Anderson]

Originally Posted by 114Klutz

I'm using Kevin Watsons streamlined code, and am running into massive oversaturation issues. The light from the Green target, along with the glare being reflected off the rack, are the same color, according to the CMUCam2 GUI.

The "oversaturated" value from the camera is 240:240:240. If you see that in the GUI, it's because you're using the RGB color space. Switch to YCrCb instead, and you'll see much better numbers.

[114Klutz]

Even after switching to ycbcr, the values for the green light and the glare seem to be exactly the same.

[Alan Anderson]

Originally Posted by 114Klutz

Even after switching to ycbcr, the values for the green light and the glare seem to be exactly the same.

In order to see what the values are using the LabView GUI tool, you need to

1. change to the YCrCb color space
2. send the new settings to the camera
3. capture a frame

Leaving out any of these steps means that you'll still be viewing the RGB representation of the scene. If your camera is working properly, switching to YCrCb will result in a nearly uniform purple picture, with subtly darker bluish patches for light sources and bright reflections. The green target light might not look any different in the picture, but its values are definitely distinguishable from the other lights. You can turn on highlighting and/or use the "color tracking" window to prove it.

[pafwl]

I just created a new thread that discusses this exact problem. It is called "Free Sunglasses for your Camera".

http://www.chiefdelphi.com/forums/sh...ad.php?t=56654

I use neitral density filters you cna get at any theatrical supply house or photography store.

Good Luck

[Vogel648]

Can't you just decrease the brightness control? I know with the VI this is possible I am not sure what you are using though.

[114Klutz]

Originally Posted by Alan Anderson

In order to see what the values are using the LabView GUI tool, you need to

1. change to the YCrCb color space
2. send the new settings to the camera
3. capture a frame

Leaving out any of these steps means that you'll still be viewing the RGB representation of the scene. If your camera is working properly, switching to YCrCb will result in a nearly uniform purple picture, with subtly darker bluish patches for light sources and bright reflections. The green target light might not look any different in the picture, but its values are definitely distinguishable from the other lights. You can turn on highlighting and/or use the "color tracking" window to prove it.

I did switch to YCbCr successfully. I meant, that even afterwards, the green light and the glare off the backgrounds were still exactly the same as each other.

[Alan Anderson]

Originally Posted by 114Klutz

I did switch to YCbCr successfully. I meant, that even afterwards, the green light and the glare off the backgrounds were still exactly the same as each other.

So you saw the purple with blue blotches? That's steps 1 and 3, but I'm still concerned that you might have missed step 2. What source did you use for the camera settings? The defaults in the LabView program are all appropriate for RGB, but they're far from the values used in the code provided by Kevin Watson.

What color values are you getting for the target light and reflections? Are they anything like the range Kevin's code is looking for?

[114Klutz]

The entire thing actually looked greenish to me. But yes, I am familiar with how YCbCr looks like.

I think I will be following the idea of using multiple sunglasses.

[Alan Anderson]

Originally Posted by 114Klutz

The entire thing actually looked greenish to me...

I've never heard anyone describe it as greenish before. With the appropriate settings, it's unmistakably purple. It sounds like either your camera is malfunctioning or something was severely misconfigured.

[114Klutz]

Perhaps the program I was using to display the camera output puts puts chroma into a different channel then labview - I was using the Java GUI.