Public Lab Research note

  • 0

Fort Bayou Near Infrared Map

by jrokoff |

Yesterday (9/10/14), a group of Wash U students, with Stevie and Scott from Public Lab, went out near Fort Bayou. We balloon mapped the vegetation in the area near the terraces -- using a near infrared camera. We stitched together some of the images we collected (see main image) using Map Knitter, but ran into an issue on converting the stitched image to NDVI. Has anyone attempted to convert a stitched image to NDVI before or is it best to convert the images before stitching?

gulf-coast balloon-mapping washustlouis washu-stlouis


Hi! With only the infrared image, and no visible light image, you can't do NDVI, unfortunately, as NDVI requires both visible and infrared. Did you not have an Infragram camera? If you have a pure NIR camera, you should consider modifying it with a red filter to be a full infragram camera and get both channels at once.

Is this a question? Click here to post it to the Questions page.

Ah.. learning here ok. So, we need an exact match for the images? Because this map we got as well, has some overlay with the location:

Is this a question? Click here to post it to the Questions page.

It's a challenge to match up photos that were not taken from exactly the same place. It really requires that all the photos or the mosaic be georectified first so they are orthophotos. If you have a lot of photos, and they are mostly vertical shots, you could look for the best matching pairs and use Ned's Fiji plugin to align them and make NDVI for an overlapping rectangle for each pair. But you might not get enough overlap for the algorithm to find the alignment matches. If the RGB and NIR photos are not aligned really well, NDVI is computed with values from two different places and the result can be meaningless.

That's why a dual camera rig uses an external timer and CHDK to take two photos (RGB and NIR) at exactly the same time of the same scene. You needed a Titan 2 IR rig.

Those rigs are beautiful, you know I was admiring them.. :)

The Infragram point and shoot would solve this though right?

Is this a question? Click here to post it to the Questions page.

I recently took near infrared photos in banana farm using DJI Phantom 3 Professional with Mobius ActionCam (modified infrared) which i bought from public labs...i took nearly 600 images from the field. Now i am having trouble stiching all those images. I tried to stich using Agisoft Photoscan Professional, however the results are unsatisfactory. Has anyone tried merging large numbers of infrared photos together? Looking forward to your suggestions.

Thank you


Is this a question? Click here to post it to the Questions page.

I have not tried to stitch Mobius photos nor have I used Photoscan. I have stitched many infrared photos including blue- or red-filtered (NGB or RGN) photos. All of these photos were from PowerShots with longer lenses than the Mobius. Here is an example. Here is another. Here is another. All of these panoramas were stitched with Microsoft ICE.

Failure to stitch could be due to inadequate overlap, motion blur, varying camera angle, lens distortion, or improper stitcher settings. You could try correcting the lens distortion before stitching especially if you have a lot of overlap (the corrected photos will be cropped compared to the originals).

Okay, I found some old Mobius Wratten 25 RGN photos taken from atop a 30 foot tall pole over some garden plots. I stitched 30 of them together in ICE and the results are mediocre. The Mobius photos are not very sharp, and the lens distortion seems to interfere with the stitching. But I made a reasonable stitched ortho image with lots of stitching errors.



Haven't tried a large number of images, but one thing that might be an issue is the fisheye lense on the infragram. I've had to correct for it before when putting images together. GIMP has a good protocol. Check out these posts:

You must be logged in to comment.