NDVI flight results

3689627980?profile=original

Following on from my last post on modifying the PowerShot S95 for NIR photography. I have ran a couple of test flights to collect some heathland data. The last flight produced 134 photos which were stitched with Agisoft Photoscan and then processed with the Fiji Photo monitoring plugin to produce the NDVI image shown. Original output files can be found here

Areas of green/higher chlorophyll containing vegetation can be seen on the processed photo as green and red areas.

The photos from the camera are a little blurred right now as the 2007 filter is thinner than the IR block that was removed from the camera causing some focusing issues. I have a replacement non blocking filter on order that should solve this problem.

I'm making progress, I now need how to analyse this to get some quantitative data...

E-mail me when people leave their comments –

You need to be a member of diydrones to add comments!

Join diydrones

Comments

  • Thanks for the info on the LUT Chris. It is relatively new to me so it's great to have some clarification and tips :) As you say, I think that for this LUT assigning 250 - 255 to a grey would be a good idea.

    I agree with you that it probably isn't the glass itself that is fixing the focusing issue. It is more likely to be due to the glass holding the sensor away from the lens in the same way that adjusting the screws does. I tried adjusting the the screws in my S95 a fair amount before opting for the glass option as I just couldn't get it right, I might have another go at some point though.

    The Rosco 2007 filter is sandwiched between the optical window and the sensor so at least it is being held in place pretty securely. I did wonder about the security of it when it was just rattling around in the space for the old IR block.

  • That NDVI_VGYRM look up table assigns magenta for only the values of 250 to 255. That is intended as a troubleshooting feature. If NDVI values are that high (equivalent to NDVI > 0.95), it is obviously due to an artifact. In dark shadows, both the NIR and the Red value will be very low, and for some reason Red is typically lower than NIR. So NDVI can compute to a very large value. Because it has no biological meaning, you could justify filtering out those high values and coloring them with a less conspicuous hue. But I like having the magenta remind me not to take the actual values too seriously. 

  • Did you sandwich the glass and the polyester filter together? If so it appears that no new problems were created.

    I am hypothesizing that the glass itself is not responsible for the better focus. Good focus can be achieved with just a polyester filter replacing the IR block glass. Below is a photo from an S100 with the IR block filter replaced with a piece of polyester Wratten 87 from Lee filters. It is not as crisp as a photo from an unmodified camera, but it is very close. So I am sticking with my position that adjusting those last screws can make all the difference. The photo is re-sized from 4000 to 2000 pixels across, but otherwise unaltered.

    3701888719?profile=original

  • I'm afraid I don't Mark. Hopefully it may be less of a problem with aerials?   I'm defiantly going to give that lut a try. 

  • Hi Jim, I used the NDVI_VGYRM LUT. I prefer it over the NDVIClasses_-1_1 for a quick assessment of vegetation indeces as it only colourises positive NDVI values. You can see the output and table colours in the image below.

    There are some shadows in the image that are showing v. high NDVI values. Do you have any tips on dealing with this?

    3701888586?profile=original

  • Above photo looks spot on. Will be interesting to see this processed. I'm using the NDVIClasses_-1_1.lut as this gives the healthy vegetation a deep green hue. Which lut did you use for your first image Mark?

  • This is the optical window I used. I ordered a 1.10 mm thick window and cut it to size. I was worried that it would be too thick (The IR block from the camera measured 0.6mm thick with my digital callipers) but there doesn't seem to be any issue with the thickness. I expect that the glass is helping to keep the rubber gasket and sensor away from the lens and this is what was causing the focus problems.

  • Chris as a follow up I have fitted an optical window with good RGB and IR transmittance (Borofloat in case anyone needs to know) and the images are nice and crisp now. Next step will be another survey flight to get some better aerial NIR data.3701888626?profile=original

  • Mark,

    The only S95 I converted works pretty well. I put a Wratten 87 polyester filter inside and the focus is sharper on the right side of the photos (below). So I assume I can improve focus by adjusting one or more of the sensor screws. I will be really interested to learn how the optical glass helps. The first test to do might be with only the new glass to see if focus rivals the unmodified performance. Getting those screws right will be crucial.

    3701886791?profile=original

  • Chris, I've played around with the screws a lot but can't improve the focus. They are currently tightened right down. The camera was focused at infinity so I'm expecting that the missing ir block is the problem. Have you got the S95 focus sharp with the IR block removed?

    I have an optical window with the same dimensions as the IR block ordered. This will give full NGB wavelength transmittance and I can install the filter on top of this. Hopefully this will sharpen things up.
This reply was deleted.