Hi all,
I just wanted to see/read your impressions about this simple camera I have created to do some simple conservation related/work.
The multispectral camera is based on the compute module of Raspberry Pi. Uses two raspberry pi cams with a M12 holder modification and 8 mm lenses. One of the cameras uses a NIR filter (PiNoir), while the other one is the original RGB camera. Control over the compute module is done over UART connection with a 433Mhz 3-DR Radio. A 6 A battery powers the array.
It is a very simple array, and is intended as to enable conservationists to carry out simple tasks that could make a hugue difference while surveying. This camera, airborne with a small drone could help characterizing endangered species habitats that are restricted to very isolated areas; help recognising plant species dominance over canopy, plant health, and so on...
Any ideas, suggestions and critics and extremely welcome, since the idea is to improve the prototype.
Many thanks to all in advance.
Replies
Arnoldo,
I always wanted to play with the RPi cams but never found some time to do so...
My 2 cents: the best thing I guess would be if it would be powered by the main battery/BEC and if it would be based on a Raspberry Pi Zero. For mapping purposes it would be ideal if the cameras could be triggered from the AP and if the images could be saved to an SD card. Then it would have less weight and could be treated as "normal" p&s camera.
So some questions:
What about the rolling shutter effect with these cameras? Have you tested it on a UAV?
Can you provide separate IR and RGB images?
What lenses are you using?
Thanks and best regards,
Thorsten
Hi Thorsten,
Powering the camera from the main battery should not be a problem. Synchronization with the main PA is something that can be done, but since I don't have one, I can't really comment on that.
The raspberry Pi Zero does not have camera port. It's got a micro USB port though so in theory you can connect one USB camera; that in itself renders the RPI Zero useless for this application since you need at least two cameras synchronized. The RPI compute module has two camera ports; and triggering the cameras is something easy. I do it using a serial port multiplexer (tmux), which makes sync easy. Once the camera gets back to a place with Wifi, it will automatically upload the pictures to dropbox, using different folders for each camera. The pictures are named using date-time (including seconds) preceeded by the camera number.
Rolling shutter. Images need to be registered in arcmap or similar. The georeferencing tool does a good, but not impecable, job by aligning the pictures. Using the shift and rotate tools before, renders the automatic registration more accurate. There are several algorythms available to carry this process automatically. I am doing some research into Matlab to be able to apply some algorythms and make this automatic.
Yes, I can provide separate NIR and RGB pictures. This is all about two separate cameras. Find attached the originals used for one of the pictures posted above.
I am using cheap M12 8mm lenses, and a pretty expensive NIR filter from Andover.
cam0nir.jpg
cam1rgb.jpg
https://www.raspberrypi.org/blog/pi-noir-infrared-camera-now-availa...
http://raspi.tv/2013/pinoir-whats-it-for-comparison-of-raspicam-and...
$10 car video camera comes with battery, SD card video recording
Just install 4 car video cameras as array as payload.
Remove IR filter from one lens, remove from another and attach other filter
...
play 4 recorded video streams, sync them into matroska container
and postprocess 4 video channels off-line
IR night vision system featuring SD card recording can be purchased at $200.
For thermal camera go to
http://diydrones.com/profiles/blogs/workswell-airvision-thermal-ima...
False color images
false color.jpg
Multispectral Pi2.jpg
Some more pictures
IMG_20151215_192938.jpg
Multispectral Pi3.jpg
NDVIMultispectral.jpg