Greg asked about getting useful information from photos taken with an Infragram camera that had n...
Public Lab is an open community which collaboratively develops accessible, open source, Do-It-Yourself technologies for investigating local environmental health and justice issues.
Public Lab chatroom
Reset your password
Read more: publiclab.org/n/9038
Greg asked about getting useful information from photos taken with an Infragram camera that had not been custom white balanced. If photos are taken with such a camera, plants and most other things appear purplish. The red and blue histograms of these photos are mostly overlapping, so computing NDVI, which highlights the difference between the red and blue channels, often produces unhelpful results. However the red channel still records mostly near infrared light, which by itself has very useful information about plant health. Healthy foliage appears bright in images of the isolated red channel of these photos. To highlight this information, you can colorize the NIR image by applying a color table like the ones we have been using to colorize NDVI images. This makes a "heat map" with more intense colors where there is more near infrared light being reflected.
I used the new Infragram Sandbox to try different variants of NDVI to find one that seemed to distinguish plants from non plants in these Infragrams. By entering (R-B)/(R+B) in the formula box, you can produce an NDVI image of the Infragram you have uploaded. In the formula, R, G, and B refer to the red, green, and blue channels. This formula does not produce a very meaningful image for Infragrams from cameras that have not been custom white balanced. One formula that seemed to produce a somewhat meaningful image is ((R*3)-(B-G))/(R+(B+3)). This formula is completely arbitrary and is probably quite hard to defend biologically. But foliage in the resulting image is lighter than some other things. To highlight this pattern, I colorized the images just as we colorize NDVI images. The result looks superficially like an NDVI image, and just might have as much biological meaning as some.
This exercise would have been much harder without the new Infragram Sandbox. It's a great way to experiment with Infragram photos. Jeff has posted a video about how to use it. The monochrome images downloaded from there were colorized in Fiji. Click the image above for a bigger view.
And do a custom white balance on your infrablue camera.
Chris, it occurs to me that we could try to come up with a way to generate an adaptation of NDVI based on a given un-white-balanced camera's photo of a white reference. What if we looked at the balance of a white reference and used the ratios to build our mod of NDVI?
It might help to have Infragram Sandbox show the color balance of any region you hover over.
Is this a question? Click here to post it to the Questions page.
Maybe try this brute force version with an infrablue photo with healthy plants in it:
1) Select an area of foliage.
2) Calculate means for both the red and blue channels for some or all of the pixels.
3) Based on the difference between those two means, select one of several NDVI formulations.
4) If there is minimal difference, assume NDVI will fail, so resort to using the red channel (NIR) as a proxy for leafiness.
The difference between an NDVI formulation for good infrablue photos (well separated red and blue values) and bad ones (partly overlapping red and blue values) is two-fold. For bad photos, resulting NDVI values will be below the traditional range (0.2 to 0.8) and span a small range. So the adjusted NDVI formulations should result in two types of adjustment.
For some poorly white balanced infrablue photos with completely overlapping red and blue channels, using the green channel is called for. So step two above should probably include the green channel to determine if it differs from red and blue more than red and blue differ from each other. If so, an NDVI equation including green may be the only hope.
You must be logged in to comment.
updated 8 days ago
updated 5 days ago
Public Lab partnership with NASA AREN project brings new easy-to-use tools for classroom use
updated 4 days ago
computer vision color detection
updated 14 days ago
Camera exposure matrix
updated 21 days ago
Use Image Sequencer for NDVI plant analysis with 2 images
updated about 2 months ago
NDVI workshop - Barcelona
updated 2 months ago
Red vs. blue filters for NDVI
updated over 4 years ago
Prototype: filter map tiles in real-time in a browser with ImageSequencer (NDVI Landsat)
updated 2 months ago
Great Neck, Ipswich, Massachusetts
updated about 4 years ago
Fort Mason Community Garden NDVI False Color. San Francisco, California.
updated about 5 years ago
Timbertown NDVI. Chelsea, Michigan.
updated over 5 years ago
Engineers Without Borders Regional Meeting at Cal Poly (NDVI)
Salisbury Swamp NDVI. Salisbury, Vermont
updated almost 6 years ago
Save The Bay Infrared. Providence, Rhode Island.
updated about 6 years ago
Foothills Community Park (NRG), Boulder, Colorado
updated over 6 years ago
Lima, Peru (Infrared)
updated about 7 years ago
Gowanus - Brooklyn, NY - NRG
Ag Park. Sunol, California (near infrared)
Ag Park. Sunol, California (ndvi)
Gowanus - Brooklyn, NY - IR