Hello everyone.
We are looking to develop a UAV for crop surveying where the stress identification will be performed mainly by an infrared camera. We are not looking to spent too much money on the sensor and have been considering modifying a normal RGB camera. Are there any major advantages for using an off-the-shelf infrared camera vs getting a normal camera converted?
Thanks!
Replies
What region of the infrared are you working?
NIR(700-900nm), SWIR(900-2500nm), MWIR(2.5um-7um), LWIR(7um-14um)
There are integrated solutions for all regions, with Visible sensors, Including Image Fusion.
You say you don't want to spend too much? Well that may be a real issue. To be successful, the imaging system is of the utmost importance.
The issue at hand is that off the shelf consumer grade cameras used are not designed for data analysis, they are made for taking pretty pictures. The senors used in consumer grade cameras are the lowest grade of sensor, Grade 3. These sensors have the most bad pixels and are made to the lowest of standards in quality and quality control.
There are Grade 2, Grade 1, and Military/Space Grade(probably more expensive for a single sensor than the house you live in?) Most good industrial cameras use Grade or Grade 2, never a Grade 3.
As well, the Grade 3 sensors are usually of poor low light ability, have noise in the images, have high Dark Current & Dark Noise, Small pixel wells, and low Dynamic Range, just to mention a few issues.
I have seen the consumer cameras converted using Dual Bandpass filters. A region of Visible up to around 600/650nm and a region of NIR from 700-850nm. Beyond 850nm Silicon based sensors drop off quickly in QE and spectral response, usually terminating around 1100nm. They all perform horribly even with such software from Pix4D and others. Again the sensors are not designed for this and will perform poorly. Seen it a number of times.
So, I guess it depends, do you want to do this professionally or are you doing this as a hobby?
Have a look at the attached images. This are prism cameras, designed for UAV use and for Agricultural data collection. There are also hybrid Multi-Spectral/Hyper-Spectral, and Hyper-Spectral available.
telecamere-multicanale02.jpg
telecamere-multicanale10.jpg
telecamere-multicanale01.jpg
Could you please provide a link where these prism cameras for UAVs are sold ? Because what you provided is the sensor only..
Thanks
Michal,
You are correct. One prism was without imaging sensors, the other two had the sensors attached to the output faces.
The prisms in the images were designed and manufactured by Optec, SpA(www.optec.eu). Complete prism cameras using Optec, SpA's arrays are integrated in to cameras built by Quest Innovations(www.quest-innovations.com). Currently every one is at the Stuttgart Vision Show in Germany.
Along with Aphesa(www.aphesa.com) and Awaiba(www.awaiba.com).
Contacts: Aphesa - Arnaud Darmont, Awaiba - Martin Waeny & Stephen Voltz, Optec - Giuseppe Cilia, Quest Innovations - Richard Meester.
I am a distributor in the US for all four companies. Since you are in Europe, you are are welcome to contact each direct.
My take is that there is a significant difference between a converted camera and something like a Tetracam. The Tetracam will give you 3 clean, calibrated (relatively) narrow bands. The converted camera will give you (typically two if one of them is NIR) broad overlapping bands that are not calibrated unless you go through the trouble to do that. The converted camera image is also recording through a Bayer filter. Most people seem to be interested in seeing how vegetation vigor varies across an image and if that sort of relative image is good enough for your application then the converted image is fine. If you go with a converted camera I suggest you use a red filter like a Wratten 25.
Dear Benjamin,
I have done some trials both with modified canon and Nikon.
The best result was obtained with Nikon Coolpix P6000. Removing the IR blocking filter is only one part of the work.
You must work quite a lot on a correct calibration and choosing a correct filter to put in front of your camera objective (or inside replacing the IR -best solution). Tetracam offers a quite expensive out of the shelf solution, but basically uses the same principle (Delgado).
See some results obtained: www.agriculturaluav.eu
Best
Marco
Thanks so much Knut, very useful information. We will look to use a modified Canon to start off with.
It was unclear from the way you asked the question whether or not you were aware that you need a visible AND a NIR band to calculate NDVI. Having both on one sensor makes image processing much easier. Be careful of "IR" conversions that only take out the NIR blocking filter. You also have to block red, and the high end of NIR. If you need good data it is worth getting a camera from a professional with a good track record. MaxMax.com is well proven. Event 38 may be worth looking at too.
Baie dankie Deon! We are aware that we need both types of sensor, however we were considering mounting the two cameras on a single gimbal. We do need good data and both options you suggested look great. Would you happen to know of any South African alternatives, or places where we could get the conversions done locally?