Public Lab Research note


IR Pass Filters and a 2 camera NDVI Setup

by dlaflash | September 21, 2014 18:26 21 Sep 18:26 | #11167 | #11167

What I want to do

Purchase an IR pass filter (blocks everything but IR light) or more ideally purchase IR pass filter sheets to be modified for my Mobius 2 camera NDVI system. I want to see if using an IR pass filter on the modified (IR filter removed) Mobius camera will help with NDVI in any way.

My attempt and results

I have not attempted this yet. I am on a fairly fixed budget (yay for being a student) and was hoping to get some input before pursuing this any farther.

Questions and next steps

I have found amazon sells glass IR pass filters for SLR cameras and I found some online retailers that sell the small filters. I am hoping someone here might be able to chime in on the pros and cons of using this in a 2 camera NDVI system.

I noticed that all the one camera systems use a filter to block certain bands of light so the NIR shows up better but never have I seen a mention of using an IR pass filter on a 2 camera system. Would this be beneficial? Would it make better images? Can the current Fuji plugin be used with the IR pass filter? Also if it is useful or worth a try what IR filter would be best for vegetation? They come in a variety of nm such as 760, 850, 950....

Why I'm interested

Just curious if this might be beneficial.


7 Comments

Public Lab promoted dual-camera IR systems more than three years ago before they found that the one camera systems worked well enough to get people started collecting plant health photos. So there is a lot of information at this site about dual-camera NDVI systems (e.g., this note). There is also lots of information about pure NIR photos.

Glass IR pass filters can be used if the camera allows attaching them in front of the lens (the internal IR block filter has to be removed first). Some PowerShots allow attaching a filter tube, and such cameras have been used in dual-camera IR rigs. The cut-off wavelength to use can be anywhere from 700 to 950 nm, but somewhere around 700-750 nm seems to be a good compromise. Good quality filters can cost $100, but it is now possible to find glass IR filters for under $20.

Polyester long pass filters are also available in standard photographic types. The Wratten 87 filter is sold as a 3 inch square of polyester by Lee Filters, and can be found for under $20. The cut-off in the Wratten 87 is at about 720 nm. A little piece of this can replace the IR block filter inside the camera as in this rig. Small pieces of filter are available here.

In general, a dual-camera system will produce better NDVI results than a single-camera system because it produces both an NIR photo and visible light photo that are not cross contaminated (e.g., no red in the NIR photo). Dual-camera systems also produce normal color photos of the scene which is really nice. Having all four bands to work with (RGB and NIR) allows good false color IR images to be produced, and these are traditionally important for vegetation remote sensing.

Ned Horning's Fiji plugin was designed for manipulating the photos from dual-camera systems and makes that job easy even for directories full of matched VIS/NIR photos.

For aerial photos when the cameras are moving, the shutters must be triggered on both cameras simultaneously. This is straightforward with PowerShots, but I don't know how to do that with Mobius cameras.

Chris

Reply to this comment...


Hello Check my notes for daul cam system maybe they can help http://publiclab.org/notes/gpenzo/02-16-2014/ndvi-720nm-850nm-filter-two-camera-setup-result http://publiclab.org/notes/gpenzo/07-10-2014/inconsistant-two-cam-ndvi

Use them at a patato farmer and he liked the result. Grayson.

Reply to this comment...


Hi gpenzo, I read all your articles and would like to know if you have completed your work. Do you had better results with the filter 850 or 720? What software did you use for processing? I am building my drone and also will not use Gimbal to increase the autonomy of flight, any comments? I intend to use two Canon SX230 HS.

Thank you

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

Reply to this comment...


Hi tagorecaue. I did not try different lenses any more. I checked my camera result with a farmer who uses a ndvi camera of a university and my result where close to his. But I stopped this project. Reason I want something smaller and lighter for my drone. And I do not need 12-16 mega pixel resolution. I'm currently working on a cape for a beaglebone black which will house two 5 mega pixel camera. The idea is to house it in a 3d printed housing which will hang under my drone. Working on using opencv for image overlay. My idea is that my new camera system should be stiff enough to not to do a image matching every time. Only need to do a calibration once. This will make the software much faster. Also did not use a gimbal to save weight. But watch out with very strong winds. The result will not be uniform.

When I have some results with my beaglebone will post it. But still alot of coding to do. Linux kernel space can be a pain in the but. Grayson.

When I have some results I will post it here.

Reply to this comment...


Hello,Sorry for the newbie question, i am trying to experiment on NDVI, i just remove the IR filter on a EOS 1000D so i am looking for filter i can mount on a 58 mm thread. Should i get on filter to get only the IR ( above 700 nm ) and another just for the Red ( 650 mm ( no so sure about his value )). Then calculate the NDVI base on this two value ? Or does it will be interesting instead of the RED filter just a IR block so i still have my three RGB channel ?

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

Reply to this comment...


vdiallonort,

If you are going to use a single camera with the IR block filter removed, there are two options. If you can take two photos without moving the camera, one photo can be taken with an NIR pass filter (e.g., 720 nm or 750 nm) and another can be taken with an IR block filter just like the one you removed. Then the two photos of the exact same scene can be aligned and NDVI can be computed for each pixel. It might be hard or expensive to buy a 58mm IR block filter. The other option is to use a filter that blocks all blue light. The blue channel of the camera will capture only NIR, and the red channel will capture red and also NIR. The blue and red channels can be used to compute NDVI. The red channel is contaminated with NIR, but at least you don't have to align the two photos.

Another option is to use a second camera that has not been modified to get a normal photo of visible light. Use your modified camera with an NIR filter (720 or 750 nm) to take a photo of the exact same scene then align the two photos and compute NDVI with one channel from each photo.

You can use a narrow band filter to capture just pure red light (e.g., 650 nm). However, a normal camera has Bayer filters over each pixel and some of them pass only red light. So a normal camera already provides a mostly red channel (but with a much broader range and usually also some NIR). So a narrow band filter can produce better data for NDVI, but exposure times can be long and the filters are expensive.

Chris

Reply to this comment...


Thanks for your answer, i found some not to expensive filter that block above 700 nm,and some that block under 700 nm so it's not the perfect solution but should be a start.A red filter will be amazing but no idea how pricey it can be.I will also try the blue filter.My goal ( learning purpose ) is to compare the different options of calculation vegetation index.

VIncent

Reply to this comment...


Login to comment.