camera trouble,cant tell the green-light from regular white light [Archive]

Nimmy

13-01-2006, 12:01

Rick TYler

13-01-2006, 12:13

(Rick whacks his head with the flat of his hand several times.)

I can't quite remember this conversation, but I think this is pretty close:

At the Seattle Robotic Society's Robothon event I was talking to the builder of the winner of the Robo-Magellan competition. This involved an autonomous robot covering rough ground outdoors to a goal. He mentioned that he finally had to put a filter over his CMU cam to make it work properly. Here is the part my memory fades on. I THINK he said that the camera was being oversaturated by UV light, and that adding a *framistat* level UV filter, it worked much better. Unfortunately I don't remember his name or the nature of the filter. One of our mentors is a SRS regular, I'll ask him if he knows how to find this out.

Sorry I couldn't be of more help.

Kevin Watson

13-01-2006, 12:19

our programming team has been working all week on the camera
and they can't seem to be able to tell apart the green light we got
for FRC from the regular light from lightbullbs in the room,
we tried it from diffrent distances but the green-light seems to just
be so powerful that the camera gets "blinded" by it and sees it as white

anyone else have this problem?
is it possible our green-light is to powerful? thanks in advance

(sorry for typos, in a hurry)You should try loading the camera.hex file from either of the .zip archives and see if it works. The default calibration parameters should work just fine if you're using the official FIRST green light.

-Kevin

Phreakuency

13-01-2006, 12:26

I believe that the reason is that the wavelengths are close enough that it is causing the camera to detect them as the same color. This could either be due to the wavelengths as I said before, or it could be due to a setting in the camera. What I am going to try when I get ahold of the camera again is I am going to dump the camera settings into a text file and I will try to find either a contast setting or jumper, or a good filter type.
I should be able to get back to you by saturday or sunday, if I fnd anything earlier, I will let you all know.
Sincerely,
Phreakuency

Nimmy

13-01-2006, 13:49

thanks guys, ill update my programming about the camera.hex
and see about that UV filter

is there a chance that the light is to bright for the camera
if we take out one of the green lights (or 2) will it make the green light
less bright? or will it then not catch it in the competition?

I was also thinking to put a layer of some whitish plastic on the light to make it less bright,

the problem is when we "grab image" and then select the green light
it highlights the greenlight and also highlights any other strong lights in the area (fire escape light for example)
it also shows the light in the RGB as green:240 blue:240, the red varies
but it is a high number

however in the edge of the green light, the green is more strong then the other colours, still not good for us though.

I will check back on saturday-sunday also

have a good weekend guys thanks :)

Matt Krass

13-01-2006, 20:09

How are you powering the light? We're using a 12 volt Robot battery and its just the right brightness.

Andrew Blair

13-01-2006, 20:39

I think that the problem lies with camera exposure values. Somehow, if you don't want to use Kevin's code, you have to adjust the exposure settings so thats its dim enough that it can distinguish.

Smithvillefirst

13-01-2006, 21:08

I am Anthony Peck from team1806 and My camera is not working properly the camera has Inverted Servos and WILL NOT be changed via jumpers, Is it fried?

Dman

14-01-2006, 11:19

b_mallerd

14-01-2006, 11:56

Mayb you just put in the pan servo upside down? Is the arm coming out of it closer to the top?

Just a suggestion...

Kevin Watson

14-01-2006, 12:12

I am Anthony Peck from team1806 and My camera is not working properly the camera has Inverted Servos and WILL NOT be changed via jumpers, Is it fried?What exactly do you mean by inverted? Are you using the pan/tilt mechanism from the kit of parts?

-Kevin

JoshJ

14-01-2006, 13:39

I am Anthony Peck from team1806 and My camera is not working properly the camera has Inverted Servos and WILL NOT be changed via jumpers, Is it fried?

I'd be willing to bet that you're camera is mounted upside down. Ours was.... :rolleyes: The serial port should be at the top, lense should be towards bottom.

Eldarion

14-01-2006, 13:47

I am also having the same trouble with the camera. The pan servo works correctly, but the tilt is inverted. I have tried repeatedly to use the jumper on the board to invert the tilt axis, but to no avail. Any advice in this matter will be greatly appreciated.

Do you have the servos connected to the camera's on-board servo drivers? If so, the tilt jumper is not the jumper identified in the instructions. To reverse the tilt, short the jumper marked as pan in the instructions.

If the servos are hooked up to the RC, just change the code :rolleyes:

Athenian Roboti

14-01-2006, 18:43

Mike

14-01-2006, 21:48

I was talking to Greg Marra today, who had the same problem. He finally fixed it. The problem was that he was using RGB instead of YCbCr.

Hope that helps
-Mike

ahecht

14-01-2006, 21:55

I was also thinking to put a layer of some whitish plastic on the light to make it less bright.

The light should have a sheet of whitish plastic mounted in front of it. There was a piece of HDPE in the kit (shrinkwrapped with the cold cathodes) that is designed to be used as a diffuser.

Eldarion

14-01-2006, 22:17

We have found theat the camera sensors are overdriven. To remedy this, we have (temporarily) put sunglasses over the camera, and this seems to have worked. We may put a dark filter over it in the future. Good luck.

Team 852-- The Athenian Robotics Collective

I also ran into this problem. To fix it (without filters, anyway), write a value of 2 to the saturation control register (register 3), I.E. send a "CR 3 2" to the camera.

AV_guy007

14-01-2006, 22:31

our programming team has been working all week on the camera
and they can't seem to be able to tell apart the green light we got
for FRC from the regular light from lightbullbs in the room,
we tried it from diffrent distances but the green-light seems to just
be so powerful that the camera gets "blinded" by it and sees it as white

anyone else have this problem?
is it possible our green-light is to powerful? thanks in advance

(sorry for typos, in a hurry)

my team was working on the camara today and had the same problem we have not found a solution to the problem. ours was reading 240 240 240 and did the same with the floresent lights in the room. we thought we could mabey have it track the wite light but then soon remembered how bright the arenas are.

let me know if you find a solution thanks

AV_guy007

14-01-2006, 22:32

The light should have a sheet of whitish plastic mounted in front of it. There was a piece of HDPE in the kit (shrinkwrapped with the cold cathodes) that is designed to be used as a diffuser.

thats how ours is set up but we still have the same problem

devicenull

14-01-2006, 23:17

Take a loot at the workbook here (http://kevin.org/frc/CMUcam2_workbook.pdf), page 6-7.. it tells you exactly how to fix that problem. (Or you can copy the values from the default code to the GUI and see how that works)

Greg Marra

14-01-2006, 23:23

I was talking to Greg Marra today, who had the same problem. He finally fixed it. The problem was that he was using RGB instead of YCbCr.

Hope that helps
-Mike

Yea, what we were doing was using last year's Java GUI to try to calibrate the camera. Of course, if I had read a little bit more before I started I would have known that the Java GUI is by default in the wrong color space, and thus no wonder it was acting differently than the default code.

(Interestingly, Track White in RGB tracks the vision target brilliantly. It also tracks the flourescent lights and paper :rolleyes: ).

Next time I get my hands on the camera, I just need to fix some pan/tilt backwardness and we should be in business.

drinkdhmo

15-01-2006, 00:10

I remedied the tilt problem for our camera by doing a little digging into the user manual. Somewhere in there it talks about the demo mode and the pan and tilt invert jumpers. It mentions that the tilt invert jumper needs to be installed while it is trying to track. We did this and it inverted the tilt axis, just as we wanted. The weird part is, ever since we did that, the board now responds to a jumper that is installed or removed while the power is off, where it did not before. I know this fix is counterintuitive, that is why I was a little hesitant at first. Make sure you do everything in your power to fix the inverted axis before trying this remedy. Then you can try this method at your own risk.

Also, concerning the difference between white lights and the supplied green light from the KOP, we are experimenting with different filters for the lense. We have more research to do, but at this point we are looking at using a polarizing filter and a neutral density filter. When we finish our testing, I will post the filter arrangement we find to be best. Good luck!

Eldarion

15-01-2006, 00:54

Also, concerning the difference between white lights and the supplied green light from the KOP, we are experimenting with different filters for the lense. We have more research to do, but at this point we are looking at using a polarizing filter and a neutral density filter. When we finish our testing, I will post the filter arrangement we find to be best. Good luck!

You really don't need to bother with that; I've been there and while it works, it is much easier to just do what I described in my earlier post. :D

When you crank the saturation control register way down, flourescent lights look red, the target looks green, and arc-lights look bluish.

bcieslak

14-01-2007, 10:23

We had to go into the camer code that Kevin supplied and changed the sign bit in tracking.h. otherwise the camera woul look away form the light.

BC

I am Anthony Peck from team1806 and My camera is not working properly the camera has Inverted Servos and WILL NOT be changed via jumpers, Is it fried?

bcieslak

14-01-2007, 10:32

We are developing code on an EDUbot while our team builds up the base of the robot. We had code from last year that works great with last year's camera. We use TTL to TTL port communications for this prototype. When we replaced the camera with this years model it doesn;t work. It searches but never finds the light. All the information in the T-packet is zero. I checked for the obvious like baud rate. I assume with no jumpers its 115200 baud. I get the impression from the assmbley drawing (the only docs IFI provided this year for the new camera) that the TTL port may be disabled.

As mentioned earlier, last years camera works fine on the prototype.

ANy Clues or do we have a bad camera?

PS we are using Kevin Watson's camera code.

BC

itsme

14-01-2007, 10:44

If you're calibrating the camera using the labview CMUcam2 demo file, then try to lessen your tolerance over there, maybe that's the problem.

JSonntag

14-01-2007, 14:53

If you are using the labview program that was supplied last year, select there should be an option to load light parameters from an external file. One of the files (there should only be one or two) holds the default color values for the lights supplied by First and will distinguish between green and white light.