SparkFun Forums

[gewoodfo]

Hi all:

I am designing a very mobile table top photo studio kit/case for taking high resolution photos of small objects like coins and jewelry. One of the things that I would like to do in this case is have the ability to capture the exact location of the lights when the picture is taken. I want to capture the x, y, z pitch, yaw ans roll ( i think i got these terms right) cooridinates of each individual light. Long term, as I build a database of cooridanates based on coin type and conditions, I want to build a targeting systems to help prepostion the lights. This will speed up productivity and efficiency.

I am assuming that I would need a 6 DOF sensor but I do not know. I am looking for a solution that will be capable of being attached to a light on the end of a flex arm. Something like a desktop reading lamp.

I would prefer to havea usb, bluetooth or wifi connection, but am open to other means to trasmit the data to the laptop computer whick will be controlling the camera as well.

Any ideas and suggestions would be greatly appreciated.

Thanks,

[theatrus]

An inertial unit won't help. First off there is no reference point, second off they will drift without getting a fix.

Your best bet is to make a rig with sliders which have encoders attached to read out the position. Think of your average set of digital calipers and multiply that by a lot of axis. You could also do it based on tensioned string and triangulate.

[Garak]

Simplest way would be a fish eye camera placed where the object is going to go. The picture could be used to calculate the intensity of the light coming from various directions. Depending on the type of light used it could determine the exact position.

Lighting is alot more art than something that can be defined by a set of coordinates. If your using a soft tent the light placement will make very little difference provided the tent is evenly lit up. Now maybe your using hard light because you want to bring the detail out. Your going to need a very high resolution system to measure the position of the light because the smallest amount of error will be the difference between a good picture and a bad picture.

Anyway the only way to find the exact position and orientation of each light automatically would be to have linear potentiometer(variable resistors) in each joint of the light fixture stand and have it mounted on a fixed point relative to the object being photographed. The potentiometer would then be connected to a microcontroller which would measure the voltage which is related to the joint angle. Using the joint angles and the lenght of the arms/booms of the light stands you could create vectors which could be added up to calculate the lights position in xyz. You could also use optical encoders rather than potentiometers to measure the angles.

My gut tells me that the data will be pretty meaning less. There will be some general trends but not much beyond what you know already. You will still have to tweak the lighting positions for each object being photographed. For example the difference in how an object has been polished will change the placement of the light to avoid reflections. Other than just flooding an object with very soft light there are no cookie cutter solutions to lighting.

Anyway I'm interested to hear otherwise. Post some of the results.

--
Chris McDonald
Audio Visual Technologist

[wiml]

Magnetic tracking is something you could look into also.

Another approach would be to put visible reference dots on the lights, take a few pictures with a webcam, and do some computer-vision processing to work out the location and orientation of the lights.

I think the most successful solution, though, would be the boring low-tech method of making a rigid frame that encloses the whole stage, onto which the lights clamp, with labeled marks (or holes into which pins fit) so that you can reproduce successful configurations. For orientation, I've seen small swivels with marks on them so that you can read off their angle (down to 15° or so).

[lyndon]

OK, I *have* to ask the obvious question. What's wrong with using three small angle protractors on each light and just writing it down after you take the shot? Low tech, no interference problems, no hardware/software to debug, no dead batteries, etc...

[bigglez]

I am assuming that I would need a 6 DOF sensor but I do not know. I am looking for a solution that will be capable of being attached to a light on the end of a flex arm. Something like a desktop reading lamp.

While this is an interesting idea its not practical. You
should rethink your needs to reduce the work and
data management involved.

Allowing unlimited movement of the lights creates
two problems (data generated and complexity in
the mechanism). How accurate must the lights be
to reproduce a previous arrangement to make a
second (presumably identical) PIX?

If you limit the lights to no movement, what percentage
of PIX are satisfactory? What if the lights only had
four positions, would that greatly increase the outcome?

You could make the lights fixed (if that works well)
or make the lights select one of a limited number of
positions (I said four, but it could be less).

Moving the lights by hand to latched positions and then
reading those electrically and storing the profile as a
text file on your PC, (with switches or photo interrupters)
would greatly reduce the complexity and the data
stored to recreate the same settings.

I'd suggest leaving the wizz-bang robotics and
fancy GUI to Hollywood.