First Therm-App Pro arrives (640 x 480 x 25Hz uncooled LWIR)

[Ultrapurple]

Native 640x480 thermal video from the Therm-App Pro looks pretty good. I've posted a 40-second sample video on the Flickr group taken from the Rock of Gibraltar overlooking the border with Spain. Vehicles can easily be seen a mile away and people at about half that distance. Impressive stuff - and the low noise level is remarkable, too. The video is exactly as-shot, with no post-processing.

[Vipitis]

This is truly amazing.

For Such a tiny device. In the past all thermal videos of this quality and range were huge cooled MWIR military imagers, expensive and inaccessible for civilians.

I am a fan and follower of your experiments with photography tricks to increase resolution and details.

Just now I invested into a better application that allows me to capture linear scaled images instead of interpolated flir images, has a feature for superresolution and working pallette/gamma locking and saves images will less watermarks. I will write to the developer again and ask for a way to save clean images for stiching.

My problem is that I am limited to high fov and low resolution with my phone. I can increase the resolution partially by super resolution. But lowering the fov will need a lens. And the offer I got from TPLogic is kinda expensive and might not work. So diy and a ZnSe solution might help to lower fov and get detail at range - for better stiching. A tripod with a panorama mount that fits my phone could help.... That might be printable.

I experimented with photogrammetry today, trying to capture one side of my house from different angles and focussing on some details. Sadly Autodesk is changing their photogrammetry platform and VisualSFM does not work with .pngs directly. So I have to postpone those experiments to the holidays next week.

I might went off topic here... But even with my Lepton sensor and some passion I were able to produce some results. Seeing what the ThermApp can do unprocessed gives me some dreams for the future... As I currently see myself going towards a university degree in computer vision or optical engineering.

[Ultrapurple]

Thank you for your kind words and all the details. I suggest you consider starting a new topic (something like 'Narrowing Lepron FOV'), but I am also happy to continue the discussion here if you prefer.

I know very little about the Lepton. I assume it's not possible to remove the lens assembly. That really only leaves you the option of making a telescope arrangement - which can work well. Try using a reflector telescope and putting a germanium or ZnSe lens in place of the eyepiece. You may be able to use it to form the light so it can be seen by the presumably fixed-focus Lepton. I use a simple astronomical telescope like this, except I do not need the extra lens for my setup. I suggest using a hot target such as a soldering iron a few tens of metres from the telescope. Set it up with its original eyepiece so you can be certain it's pointing accurately. Then start playing with lenses and the position of the Lepton to see if you can get anything resembling an image. I expect you so get poor sensitivity and high magnification - but once you have demonstrated that it works you can start optimising different bits.

Remember, some of the first thermal cameras were a single pixel, scanned to make an image. You have a lot more pixels - a small staring array - so if you can narrow the FOV you are on the right track for.hkgh resolution thermal imaging. It may just take time to build images from multiple exposures.

[IwuzBornanerd]

Side-by-side tests of the Therm-App 8.7Hz, Therm-App updated to 25Hz, and Therm-App Pro

...
In essence, within the limits of what I had available and the software that would drive the various cameras I tried for a level playing field. The scene is nothing special - a typical corner of an office, with a reasonable run-of-the-mill thermal contrast range.
...
What do you think?

I think what I would like is to see some images that actually show the camera's noise, however low it is.   You have 16 and 19 degree C temperature spans there.  Even my old Seek looks fairly good at those spans--at least with averaging on.

Will the software allow you to get those same (or similar low contrast) images with the full black-to-white gray scale spread over a mere 3 degrees C or so?  I know that my eyes can only readily distinguish 32 shades of gray, if that, so if you spread the whole gray scale over 3 degrees I know that I will be able to see 100mK of temperature difference as a different shade; and I expect most everyone would also see it.  So if we don't see any noise at that span noise is less than 100mK.  And if you don't see any noise at a 3 degree C span, can you zoom in to where you do see it??   

I too am looking for a level playing field...with all these "cheap" cameras.

[Ultrapurple]

@IwuzBornanerd

Yes, I am sure I will be able to do sensible tests soon. Driver software for the Pro is still evolving; it is now reaching the stage where comparisons will be meaningful. As I said in the earlier post, it was a quick-and-dirty first test. More will certainly follow, with various temperature ranges etc. My initial impression is that the Pro is much more sensitive than its predecessors, but I will report more once I have done the necessary investigation.

[Fraser]

This is all very interesting reading. Thank you.

Without any inside knowledge of the new Pro design it is hard to know what challenges the OEM faced when building such a camera. I find it interesting that the software appears to require significant modification to deal with the new sensor array and that the new array initially appears to have less noise than its lower resolution stable mate.

In the case of ULIS Microbolometers, each new generation of sensor array that they release is often more refined and lower noise than its predecessors. That is just evolution of the technology. The 640 x 480 pixel sensor may be a later generation than that used in its 'little brother' hence some changes to the software that drives it. I was interested that initially the Pro was not going to be capable of accurate temperature measurements. That lead me to believe that the temperature stability of the sensor die was presenting the OEM with some challenges. There seemed no other reason why such an advanced sensor array could not measure temperature accurately. The fact that the PRO was released with some Radiometric capabilities suggests that the sensor is more stable than expected, or some fine tuning of the design has improved stability.

The new dongle format 640 x 480 thermal cameras are cramming impressive performance into a very small form factor. It is to be expected that there will need to be some more fine tuning of the designs and associated software. They choose to not employ an FFC shutter and that, for me, is a disappointment. Yes it decreased the required complexity and lens block volume, but it also removes the very useful facility to auto correct the sensor array output if temperature drift is present.

Very interesting times though and I watch these new camera releases with interest

Fraser

[Vipitis]

yeah, that was my thinking as well. I believe you did exactly that with some of your ultra high resolution images.

I started a thread here: https://www.eevblog.com/forum/thermal-imaging/scope-for-my-phone/  but it got lost in a discussion on how the title is misleading - which is it. So I may just repost this thread with new information, some example images and a better title.

[IwuzBornanerd]

@IwuzBornanerd

Yes, I am sure I will be able to do sensible tests soon. ...

Thank you; I understand.  I will wait patiently.

[IwuzBornanerd]

...
They have now released the following minimum spec to run the Therm-App Pro:

System chip: Qualcomm Snapdragon 820 MSM8996
Processor   : Quad-core, 2200 MHz, Kryo, 64-bit
Graphics processor: Adreno 530
System memory: 4 GB RAM

They suggested something like a Samsung Galaxy S7, so I'm looking for one of those at a reasonable price on flEabuy. I'll continue to report progress.

What, if anything, is surprising to me is that this is all it takes.  I am certainly no expert on efficient software, but my current Seek code in C/C++ consumes 4%-5% of CPU on my dual core, hyperthreading "Intel Core i3" laptop running at 2.4GHz.  And that is to process a mere 32k pixels at 8 or 9 frames/sec.  With 9.5 times as many pixels and 3x the frame rate, I would need 120%-150% of the CPU on my laptop!  Couldn't run the thing even with 4 threads.  Of course I expect real programmers to do batter than I can... 

I run the Seek on a Raspberry Pi, but I would not even be able to run the Seek Pro on a Pi at its full frame rate since the ordinary Seek consumes half a core on the Pi.

[Ultrapurple]

EncryptedEddy did some work on getting the Therm-App running on a Raspberry Pi some time ago. It required all the resources of a Pi 3, though he suggested this may be down to non-optimal code. However, it did successfully display a Therm-App thermal image (8.7Hz) on the Pi screen. I suspect, as it was using VLC, it may also be possible to stream the video via Wi-Fi. There was some discussion over whether it would be possible to do something like that with a drone. However, given that a relatively lightweight mobile phone running the stock driver or ThermApp Plus can natively stream video, I'd probably choose to mount a phone on my drone (if I had one). I figure that would be lighter than a Pi 3 with CPU heatsink, LiPo battery pack, voltage regulator board and all the necessary cables - and much neater, too.

While we're on the subject of the Raspberry Pi, I used a Therm-App camera to film a Pi CPU at work - you can see different parts of it 'light up' as various tasks are executed.

[IwuzBornanerd]

Interesting story on your flikr thread.  I have seen versions of Raspbian consuming 100% CPU while running nothing, though, so I wonder if it might have been the OS & not the ThermApp doing it, but I probably should not belabor that issue on this thread.

That CPU video is cool; I'll have to try that before I put a heatsink on my Pi3.