Whiteout/Yellowing of photos in research applications

[Tobolo]

I'll provide a TL;DR later so skip ahead if your the impatient type.

I am trying to take pictures of turfgrass with a constant light source. In order to this we have a large box with four florescent lights and a small hole on top where we mount the camera. We have been taking pictures for about a year and have started having problems that I cannot figure out how to fix.

These photos are analyzed for their green pixels in order to produce a percentage of green color within a plot. Therefore the pictures need to be taken in standardized format so they are judged equally. Lately, we have been getting these results:

(Please note these are from the same day)

You can see how dark this one looks.

And the grey one.

And the horrible yellow one.

Can anyone think of any reason why the camera should be doing this?

Specs:
Nikon Coolpix P90
ISO 200
Shutter Speed 1/15
F5.6
1280x960

Tl;DR
Camera need to take pictures in a standard way
It doesn't
Check out the pics!
Any help?

 

[fralexandr]

those are pictures of different plots at roughly the same time of day right?
midday or maybe night-time should be the best to keep ambient light as constant as possible

some questions that hopefully help:

your camera's flash is off right?
what's the box made of?
is the box completely sealed?
have you tried using a different light source?
maybe your fluorescent lights are flickering. ex: http://www.youtube.com/watch?v=l3iXja3B8To
though i'd expect the longish shutter speed to average it out a bit better
what power source are you using for the lights?

 

[GWestphal]

Well first of all, this is in a box that sunlight can't pass through? If sunlight can pass through it then the camera's auto white balance will shift based on whatever the algorithm is. Second, the white balance of lights can change over time, old bulbs vs new bulbs.

I would say use a DSLR or some camera with full manual controls. You want to set your white balance either manually with a white balance card/cap, set a default temp and stick with it, or something of that effect so you're at least consistent.

The biggest issue would be in actually controlling your light though. Block all other light sources and make sure your light temperature has changed, maybe use LED lighting, not sure if the change temp significantly over time. Perhaps try strobes, they shouldn't change by that much over a couple hundred flashes.

Alternately, you could use some calibration standard each time and do some color correction in post processing.

 

[Tobolo]

those are pictures of different plots at roughly the same time of day right? Afternoon, but again it is pitch black in the box when the lights are off
midday or maybe night-time should be the best to keep ambient light as constant as possible

some questions that hopefully help:

your camera's flash is off right? Yes
what's the box made of? Galvanized Steel
is the box completely sealed? Yes
have you tried using a different light source? I have taken snapshots without being in the box and it still is yellow
maybe your fluorescent lights are flickering. ex: http://www.youtube.com/watch?v=l3iXja3B8To Checked them repeatedly and rewired and replaced recently.
though i'd expect the longish shutter speed to average it out a bit better
what power source are you using for the lights? Honda Generator in the field. I have been using a outside outlet for some of those shots.

See responses in the bolded sections.

 

[Tobolo]

Well first of all, this is in a box that sunlight can't pass through? If sunlight can pass through it then the camera's auto white balance will shift based on whatever the algorithm is. Second, the white balance of lights can change over time, old bulbs vs new bulbs. I haven't had a problem before I will check some old bulbs out

I would say use a DSLR or some camera with full manual controls. You want to set your white balance either manually with a white balance card/cap, set a default temp and stick with it, or something of that effect so you're at least consistent. I think I can set a upper and lower boundary

The biggest issue would be in actually controlling your light though. Block all other light sources Taken multiple photos with light and they are all pitch blackand make sure your light temperature has changed, maybe use LED lighting, not sure if the change temp significantly over time. Perhaps try strobes, they shouldn't change by that much over a couple hundred flashes. Do you mean actual light temperature?

Alternately, you could use some calibration standard each time and do some color correction in post processing.I've thought about it, only problem being that I would have 100-200 daily from now until October. I don't have much man power to change every picture.

See bolded responses.

 

[Tobolo]

Update:
I reset the camera back to factory defaults and changed all the settings again. I checked the white balance and tried the Florescent option and it completely changed the way it looked. It now has a blue hue to entire photo. Which is bad when I take pictures of Kentucky bluegrass!

It's a step in the right direction. I will check at work tomorrow about a custom white balance.

 

[ElFenix]

fluorescent lights flicker. 1/15th should be slow enough to average out the flicker out but maybe not.

here's a bit of an illustration.
http://photo.stackexchange.com/ques...hutter-speed-create-a-problem-with-color-cast

white balance will most likely solve your problem, though. ideally you'd have a grey card placed into the scene each time so that you can check the balance each time as the temp may change over time. not sure if a grey card can fade from intense light.

 

[slashbinslashbash]

Yes, florescent lights flicker at the frequency of the AC power, usually 60Hz. So if your exposure is 1/15 then you really shouldn't see too many effects, but hmmm...

First things first, full manual mode. Exposure setttings (shutter speed, ISO, aperture) need to be absolutely consistent in every single shot. White Balance should be set to a given point, never "auto". It doesn't entirely matter since you can change the WB in post, but from a scientific standpoint it is absolutely best to standardize within the camera itself. It doesn't really matter that much what the WB point is set to, just as long as it's consistent.

Next, the lights. As others have pointed out, flourescents have their downsides. Ideally you would get a few camera strobes and use them in place of the flourescents. There are some fairly cheap solutions that would actually be significantly better than dragging a generator around the field, IMO. E.g.

http://strobist.blogspot.com/2010/06/lumppro-lp160-quad-sync-v20.html

You can buy 4 of these, install them inside the 4 corners of the box with some velcro, set the power levels of all 4 equally, and play with your exposure until you get a good pic. The camera's built-in flash will set off the other 4 flashes due to the built-in slave feature, which detects other flashes and triggers the unit. There are plenty of other strobes out there (I think some well-known Sunpak models that go for like $50) but you may have to buy external slave triggers (like $10) for them.

Anyway, the point is that photographic strobes are designed to be super consistent, in terms of both power level and color temperature. Now that I think about it, 4 strobes might be overkill.... how big is this box? If it's galvanized steel, the inside is shiny, right? Really one strobe ought to be plenty; maybe even the camera's built-in flash would be sufficient. The biggest problem will be evenness of light, which can be alleviated by using multiple strobes, and by aiming them to reflect off the walls to scatter more light.

I don't know, I could elaborate more (and I'd be happy to if you have any questions), but I know you're probably sort of locked-in at this point and not wanting to change your light source. But for sure, set the WB of the camera to a given point.

EDIT: I thought of the reason why this might be happening. It is likely that the Auto-WB is set at a single point during the exposure; likely at the beginning or end (or even slightly before the beginning, or slightly after the end). So even if the spectrum of the flourescents over the whole 1/15 second exposure might be fairly even, during the split-second when the camera samples for WB, it might be at one extreme or the other. Setting the camera to a particular WB (again, shouldn't matter exactly which setting) will hopefully fix the problem and give consistent results. It is possible that the problem is only coming up now due to aging of the flourescent tubes and/or ballasts.

 

[GoSharks]

Is the lack of a control/standard "green" bothering anybody else, in terms of actually getting usable data out of the experiment?

If you had one, different color temperatures, exposures, etc, wouldn't be a problem.