Public Lab Research note


NDVI exposure setting

by gpenzo | July 13, 2014 21:04 13 Jul 21:04 | #10684 | #10684

What I want to do

I want to understand how others which do NDVI with one or two cameras decide which exposure time to use.

My attempt and results

Setup Two cameras, one normal and the otehr IR lock filter removed and a 720NM filter in front of the lens. Result images are converted to jpeg to save disk sapce. Both chdk raw output Vis: shutter 1/200 iso 200 NIT: shutter 1/200 and 1/400 iso 200

VIS for NIR 1/200 CRW_0307.jpg

VIS for NIR 1/400 CRW_0309.jpg

NIR 1/200 CRW_2647.jpg

NIR 1/400 CRW_2649.jpg

NDVI 1/200 CRW_0307_NDVI_Color.jpg

NDVI 1/400 CRW_0309_NDVI_Color.jpg

Questions and next steps

As you can see, the 1/200 NDVI picture has better indexes. My question is: How do we determine the correct exposure value for both cameras. Maybe 1/100 was better. Does anyone have an idea.

One solution which just pop up after my post: Make a picture both visible and NIR and let the camera decide what are the best iso and shuter settings for current lighting conditions. With chdk lock these settings, ISO and shutter speed on both cameras. With these "calibration" every picture made on diffrent days should yield almost the same NVI result. And Hope the shutter speed is fast enough for the movement on my drone or pictures will be blured. Comming days only rain .

Why I'm interested

If I want to use this NDVI pictures with farmers they cannot differ to much between diffrent days.


1 Comments

I often use platforms that are moving so I use a fast shutter speed to reduce blur. I often shoot somewhere around 1/1000th of a second. You should be able to compensate for different shutter speeds so the resulting NDVI images are similar even if different shutter speeds are used. If you have a reference target in images with different shutter speeds you could calculate the pixel value difference and then adjust the other image by applying an offset. It's possible that correct would require more than a simple offset but that's a good place to start.

Reply to this comment...


Login to comment.