New tools are always exciting! In July this year a new ANAFI thermal arrived to replace / collaborate with a couple of slightly broken ANAFI work aircraft. It’s taken a long time to put together a post about it because – its complicated! Adding the thermal camera means a whole new learning curve / capability assessment. A few months in now, how’s it all going?
Nuts and bolts
Parrot’s description of the aircraft is the best source for all the physical characteristics – I can confirm that they’re all true. The ANAFI thermal does fly a little noisier than the ANAFI work – I suspect because of the slightly more rigid legs – but can’t be sure. Something for Parrot engineers to work on. It is still almost silent from 50 metres away.
Here are the ANAFI thermal and ANAFI standard (with yellow tape tags) head to head (yes, the yellow tagged ANAFI has been crashed, and has a slightly wonky gimbal).
Standard imaging and video capture
In its RGB imaging configuration, the Sony 21MP sensor does a wonderful job of collecting both still and moving images. Here’s a short sample flying over a coastal scene. Like any aircraft without an independent gimbal control, smooth rotation is up to the pilot and practice…
Still imaging is quite capable also. I generally shoot everything in JPEG wide + DNG mode when I’m flying manually. I’ve found the JPEG versions more than adequate for large prints – this one was printed out on canvas at 60 x 40 cm, and looks pretty nice (if I do say so myself).
Parrot’s freeflight 6 application acquired a GPS lapse mode earlier this year. This would have made my initial cliff flights a lot easier – and also allows a great ‘just fly around’ approach to mapping stuff. The ortho-image and digital surface model below were made by collecting over 1000 images in ‘GPS lapse’ mode, using the aircraft trajectory overlaid on the map. While this is pretty intense flying, it does allow a pilot to follow terrain and capture consistent image resolutions in a mapping mission. It also shows that my flight geometry was not incredible – blue dots are camera stations, with width and height showing GPS accuracy and colour showing elevation. Ellipse orientation comes from the camera pointing angle.
In November 2019 Pix4D Capture started working with the ANAFI thermal – making map flight planning a little simpler. This orthophoto was acquired in two flights over about half an hour; again with camera stations mapped as blue ovals showing GPS uncertainty ellipses, aligned to the camera pointing angle.
Straight out of the ANAFI, I get between 0.5 and 1m GPS accuracy in 3 dimensions. This is pretty good – and far better than GPS alone. I suspect (and may confirm) that Parrot use some inertial sensor integration to constrain positions. Which is pretty standard practice for mobile instruments, and if true, nice to see. I’ll get back to this…
I bought the ANAFI thermal with business capability expansion in mind – based on my proof-of-concept asset inspection work and thinking about insulation assessment / heat leaks / solar panels. Animal finding and SAR capability was also on my mind. I’ve tested some of those use cases – with some great and some mixed results.
Rooftops, thermal escape, vegetation
This is a great use case for the ANAFI – examining where heat escapes from a building. Coupled with mapping capabilities to derive measurable areas, it’s a really neat and compact tool for assessing where insulation might be needed or not. It’s also easily able to discriminate thermal properties of different vegetation cover types.
Finding cows and cats
Animals are a less distinct use case. Given the low sensor resolution small things are hard to see. Cow sized animals are better! The video below cycles through the different thermal modes, while observing cattle from around 50m away. Note a calf wandering into then out of thermal visibility as it moves from leaf cover to open to cover again.
For this demonstration I adjusted the thermal emissivity specifically for animal fur (yes! the thermal emissivity of animals is well studied), which helped a little. Previous experiments at the default settings did not turn out so well.
I mentioned finding cats. This doesn’t work so well – the sensor pixel resolution makes smaller things hard to detect. Here’s a cat, at night, from about 10 m – using the ANAFI as a handheld thermal imager
Long story short, for finding small creatures you’d probably enjoy a higher-definition thermal sensor. The thermal resolution of the FLIR lepton unit aboard the ANAFI is great! For some tasks, it just does not have enough pixels.
SAR – mixed results
Finding people using thermal imagery is a thing – if you have a really amazing thermal setup. The ANAFI’s little FLIR unit does OK at finding people-like objects. My initial tests were awful – however, I realised I wasn’t using the sensor very well. Just like cows, humans don’t stand out well below canopy cover. Here is me, behind a pine tree, about 50m from the camera.
..yep, I’m invisible!
Without interference, we do OK. This image is me, from 30m away. I can find me! So there’s hope.
Another post will look at a few specific SAR scenarios – how long until a person becomes thermally indistinct from a river; and whether we can spot a person in good outdoor clothing at all, including sparse treecover and in the open
Just for fun I sent a quick thermal imagery flight through Opendronemap, and made a thermal reconstruction! Unfortunately only RGB values are kept in the resulting output – it would be wise to carefully register thermal values to RGB if you wanted to interpret the model colours later.
The ANAFI thermal is a lot of instrumentation in a tiny and robust package!
It is a joy to fly. I generally don’t use automated cinematography modes and my fingers are slowly getting to smooth pan and tilt shots. Using GPS lapse is a great tool for modelling – making life easier, although even finer distances would be useful in some scenarios. It’s fantastic to be able to combine structured flight plans (eg Pix4D capture) with freestyle captures to pick on details.
Thermal imaging capability is an amazing extra – it opens up a lot of new tasks for a tiny, portable instrument with great flying characteristics and a wide range of operating conditions.
I like Parrot’s Freeflight app – it’s easy to use and straightforward. The only major issue some people might have is its lack of collision detection. it certainly pays keeps your pilot senses on! I’d like little signal lights in the arms – but that’s a minor quibble for this small-but-powerful flying instrument.
I bought this one from DronesForHire – and we’ve grown a great relationship since. Happy to recommend them as a place to buy from!
The sales pitch
Spatialised is a fully independent, full time consulting business. The tutorials and write-ups here are free for you to use, without ads or tracking.
If you find the content here useful to your business or research or billion dollar startup, you can support production of more words, ideas and open source geo-recipes via Paypal or Patreon; or hire me to do stuff; or hire me to talk about stuff; or just give me a seat on your advisory board and a 1% stake. Enjoy!