RPi Multispectral Camera for Agriculture

4+ custom bands multispectral camera based on RasPi 3 and raspi camera multiplexer with geotagging and environmental montoring capabilities

Similar projects worth following
This multispectral camera based on Raspberry PI3 uses 4X NoIR V2 Raspberry Pi cameras (8Mpx), an IVPort V2 Camera multiplexer and is capable of geotag the captured images using an external GPS module (Currently using U-Blox 7 GPS).

The RPi Wi-Fi is configured in AP mode and is running a web server wich displays the user interface when connected, including the visualisation of external sensor values, making the camera an ideal aerial environment monitoring system.

Enjoy this project?



bloom wrote 03/10/2018 at 02:26 point

It's capable of geotag the captured images using an external GPS module.

GoMovies streaming site

  Are you sure? yes | no

houseonline wrote 12/24/2017 at 10:10 point

Say drone moving at 12kph
Which is 12000metres/hour
Or 3.33333metres/second
10ms delay in the multiplexer = 0.010s

Displacement over 10ms=33mm (1.3”)

This is probably much smaller than the tolerance of the GPS. Also if the delay is known and predictable then it would be possible to batch edit the metadata for each stream of the images to correct for this. As to the cameras, yes I agree that filtered monochrome cameras would be better but I would love to see the results (going to test with multiple Pi Zeros myself!).

  Are you sure? yes | no

Andres Felipe Gomez R. wrote 11/21/2017 at 21:33 point

@makye the project is in hold while I get proper filters. I had good results with image registration and mosaic generation without filters, and saving raw images instead of jpg ones. Trust me, the software available now is powerful enough for that kind of processing. You can't say what does work and what doesn't without knowing the background research, and only by reading a short description and seeing a couple of photos. Problems are made to challenge us as engineers, and I as an engineer will overcome them by any means, and that's the way you apply science.

  Are you sure? yes | no

Mayke wrote 11/23/2017 at 12:12 point

Hi Andres, in no way I'm discouraging research/science outcomes. In fact, I would love to see some meaningful results from your project. Now, based on your last post, I concluded, probably a bit early, that your project was going nowhere, or you may have dropped it altogether.

Having clarified my point, how do you describe 'good results'? How is registration 'good'? Taking into account that the multiplexer does not send commands to the cameras at the same time but, instead, it sends the commands in an orderly fashion. What is actually the latency between cameras? 7 milliseconds? My experience is that the multiplexer gives more than 10 milliseconds between cameras. That's ok if your drone stops at each waypoint and triggers the camera(s), but constant movement would be a real challenge for this type of camera to work. 

Second, What sort of filters are you planning to use? short-bandpass or long bandpass? What's the centre-point of your filters? I guess you are planning to use blue, green, red and a sort of NIR filter, right? Besides, you made no mention about the camera sensors, which I guess, means that you are happy with the PiNoir cameras. If that's so, in terms of processing, what format are you saving the images on? 

Lastly, can you update your post so as to reflect the real stage of your project? Can you post your registered images and mosaics? What you mean by proper filters? The ones on the pictures look ok; aren't they? what about those: 

  Are you sure? yes | no

Mayke wrote 11/21/2017 at 20:54 point

Logically the project didn't work. And it didn't work because you can't use a multiplexer mounted in a moving object. These things would never trigger the camera at the same time, therefore you'll have all sort of problems with image registration, alignment and consequent mosaic production. Besides, raspberry pi cameras are colour sensors with a lot of processing happening in the background in order to give 'crisp' images. Pi Noir are colour sensors without the IR filter. You need a monochrome sensor, with good quantum efficiency after 800nm. Monochrome sensor-based cameras are not cheap because they are not produced en mass for mobile phones. If you really want to create your own multispectral camera, be ready to create your own pcb design, create drivers, and get hold of a very good monochrome sensor. In other words, buy a commercial system.

  Are you sure? yes | no

Tomi Piriyev wrote 11/14/2017 at 08:33 point

Hi Andres, did you get any result ? can you share with us?

  Are you sure? yes | no

chadlathe wrote 11/09/2017 at 19:41 point

Super interested in this project, will be watching closely!

  Are you sure? yes | no

Sebas wrote 10/26/2017 at 04:17 point

I really want more information of this project.

  Are you sure? yes | no

Guy Fraser wrote 04/23/2017 at 13:09 point

Very interested in this project as I've been trying to work out how to do multispectral stuff for a while now. Are there any samples of the images taken by the system?

  Are you sure? yes | no

Andres Felipe Gomez R. wrote 04/21/2017 at 21:36 point

Project details and logs soon

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates