I am interested in any and all feedback related to measuring color value with mobile phone technologies. We have developed and tested an application that uses a single color reference (WhiBal card) to calibrate/generate RGB and Munsell values. When matched to known values via spectrophotometer readings, we have a lot of trouble in direct sun. We can accommodate some level of low precision, but would obviously like to get as close as we can to actual color. Anyone out there have experience with this?
I used " soil colour chart" android application in field. Before I match soil sample with colour chart I made a little bite wet to soil. Accuracy was rather good. So wetting process makes more accuracy...
Ugur - yes, I'm familiar with the soil color chart app. Thanks for the feedback, though searching for an app that uses the camera capability of phones to generate the color values. Attempting to go beyond person in field matching color from existing charts (virtual or Munsell). By the way, did you find your assessments to be accurate across multiple color families? Did you test your assessment against spectro readings? Just curious...
Dear Dr.Adam Beh,
It is a good idea to use smart phone camera for soil color.
These are the parameters affecting actual soil color as far as I know:
1. Soil particle size distribution,
2. Soil Texture,
3. Roughness,
4. Moisture content in the soil.
5. Salinity, organic matter, fertilizer content, minerals in the soil, etc.
If you keep these facts in mind, hopefully you may succeed.
Here's what we've tested so far......subset of 5 sample soils with know spectrophotometer values. Test 6 devices in 3 lighting conditions (controlled indoor, outdoor full sun, and outdoor full sun with full body shade on photographed area) using 3 image property settings (default camera, grab "almost raw" image in preview and transform to linear RGB, and modified preview image using new algorithm to convert to linear RGB). Some values for non-Samsung devices do not have all of the data due to glitch with settings functionality. Feel free to have a look at the data to see where our biggest challenges are (sorry, no data visualizations just yet). I can only include one data file at a time though...
Some take-home messages: 1) of major concern is how to take accurate measurements with irradiance-independent algorithms. Direct, ambient, and reflected light all wreak havoc on our results. 2) higher variance and weak accuracy for darker soils, and in indirect light. 3) single color reference off WhiBal gray is inadequate. Need multiple color references. 4) camera photos do not produce linear RGB values, so need a fancy algorithm to "lineralize."
Any color scientist out there want to give this a shot? We've got funding.....:)
I am not sure where you are going with all this 'measurement by camera', but FYI I include a Mozambique flag (it has three primary colours) in all soil photos and if the three primary colours of the flag are correct in the download then the soil colour in the download is assumed to be correct also. Then you can get the soil Munsell colour by comparing the photo with a book of standard Munsell colours. If the colours of the flag are not quite correct in the downloaded photo you should be able to adust the colour settings of the image until the flag is correct. Then the soil colour will also be correct.
Alternatively you can carry a colour printout of Munsell charts to the field, get the soil colour in terms of this print, and then get the more accurate Munsell colour back at home by looking at the colour identified on the printout and comparing it with a book of standard Munsell colours.
All this is done to preserve the integrity of the standard book of Munsell colours by not having to expose it to sun and field conditions.
Relevant or not?
Adam,
Sorry to be late to the party, but are you still working on this? I am a color scientist and spent several years trying to get accurate color out of RGB cameras. As you already know, you need to control, the environment, ensure that the camera is giving you data that you can trust - they do all kinds of magic in the software, and are constantly adjusting stuff so you can get "point and click. Another roadblock is the fact that the spectral response of cameras is different from that of the eye. I have a blog post that explains that.
http://johnthemathguy.blogspot.com/2014/07/rgb-into-lab.html
- Badges
- Science topic