Yet another cheap thermal imager incoming.. Seek Thermal

[eneuro]

Offset of each frame is 0.255 DEG which is approx 1.35 pixel.

It looks like you assumed 36 DEG FOV of this Seek dongle so 1 pixel is 0.173 DEG which means about 1.3 pixel in your setup 
However, in the case of circular motion with only +/-  1 pixel in vertical and horizontal position we can get easy 8 frames and output enhanced thermal image at rate 1Hz and we have smooth camera movement (no need to stop camera to change it direction right-left, yhis way save its kinetic energy)
1 pixel 1 meter away from Seek camera has dimension 3mm.
It will depend on use case and sometimes planar circular movement when distance is known and it is close to Seek dongle it can have higher precision.

Edit:

If I would go into making circular motion I would do something like this: https://static.hackaday.io/images/9207121405936029200.jpeg

It is too big and too heavy to move such small Seek dongle-we have I think better idea how to keep this thing moving at close to constant angle around in  a few steps

Maybe we do not need move camera to get circular motion thermal image on Seek sensor 

However they put this thing between sensor and lenses 

What do you think-is it possible tweak Seek dongle in similar way (of course we need something transparent to IR) to get circular motion images on Seek sensor, but by putting this rotating part in front of Seek lens, while there is no chance to do it inside like they did in this prototype camera?

BTW: @efahrenholz Computers understand my language very well, and difference between me and you is I  things quickly and have working solution and happy customers, while you are only talking 

[Fraser]

Can we please see a cessation of hostilities in this thread.

There is much good work going on but some members are determined to ruffle feathers. There is little point in this and I have respect for you all, except when asinine comments are made.

Be excellent to eachother !

Party on Dudes !

(Bill & Ted's Excellent adventure)

Aurora

[frenky]

What do you think-is it possible tweak Seek dongle in similar way to get circular motion images on Seek sensor, but by putting this rotating part in front of Seek lens, while there is no chance to do it inside like they did in this prototype camera?

I've paid 350$ for Seek so I won't be removing lens or lens holder any time soon.
But making a pano head with 2 stepper motors is a realistic option for me...

[frenky]

@Aurora:
I agree with you that if we work together instead of insulting each other much more could be accomplished...

We've come a long way. Just look at the photos below: original photo from android app VS stacked photos on PC.
I have now figured out that 1 step 0.225 fits exactly into 1 pixel. So perhaps lens angle is not 36 degrees but closer to 52. (I will measure it.)

If we had one brilliant android programmer on forum he could use collected knowledge to make an alternative app that would use camera angle & position to stack multiple photos and do it without need for pc and stepper motor. Only by using built in accelometer, compas and gyro...
That would be really an accomplishment.

Photo from android app:

Stacked photos got from pc:


Take a look at the detail on IC pads on bottom left. This module has a huge potential.

[frenky]

I have measured lens angle and it actually is 36 DEG.
That must mean that my stepper motor is making 0.1747 DEG (sub)steps and not 0.225 as I thought.

[eneuro]

I have now figured out that 1 step 0.225 fits exactly into 1 pixel. So perhaps lens angle is not 36 degrees but closer to 52. (I will measure it.)

Seek thermal website shop http://obtain.thermal.com/product-p/uw-aaa.htm says it has 36 DEG FOV 
I don't remember if this ONLY awailable Seek spec was shown in this thread so it is there:

Anyway they says about 206x156 32.1k thermal pixels, but as we can see on raw sensor data there is about 2.1k useless hexagon pattern pixels send from this Seek cam via USB so it is NOT true or those "patent" pixels are created in Seek firmware on PCB to degrade output image quality for some reason-who knows while end customer have to believe in 32.1k thermal pixels but those 2.1k hexagon doesn't have thermal image information, so why they are talking about 32.1k in this spec? It is 206x156 but thanks to this thread we know what is send from this dongle to customer Android 
Even such short technical description raises many questions, so it is worth try to verify FOV of this dongle and other claimed spects as well 

Did anyone verified capability to detect -40*C thermal object temperature using this dongle?
SKF TKTI 21 has this lower limit -17*C for example...no chance for me to test this 

[mikeselectricstuff]

Did anyone verified capability to detect -40*C thermal object temperature using this dongle?

It says detection, not measurement, so possibly not a very meaningful figure.

[jaybeez]

@Aurora:
I agree with you that if we work together instead of insulting each other much more could be accomplished...

We've come a long way. Just look at the photos below: original photo from android app VS stacked photos on PC.
I have now figured out that 1 step 0.225 fits exactly into 1 pixel. So perhaps lens angle is not 36 degrees but closer to 52. (I will measure it.)

If we had one brilliant android programmer on forum he could use collected knowledge to make an alternative app that would use camera angle & position to stack multiple photos and do it without need for pc and stepper motor. Only by using built in accelometer, compas and gyro...
That would be really an accomplishment.

Photo from android app:

Stacked photos got from pc:


Take a look at the detail on IC pads on bottom left. This module has a huge potential.

post a review of the seek over on the xda-developer's forum, this is their accessories sub forum
http://forum.xda-developers.com/general/accessories

gather enough interest and you/we may be able to solicit some development help

[cynfab]

For your viewing pleasure:
rawData.png
rawCal.png
Diff.png
CImage.png

I took some canned air (most likely not air) and used it as freeze spray to freeze the sensor of an indoor/outdoor digital thermometer.
It got down to -30C more or less.
I then took a picture of it and my hot soldering gun which gets to about 300C.
Processing --> rawData + 2000 - rawCal = Diff  --> saved as 16bit Diff.png

CImage is the colorized/denoised and up-scaled image.

Looking at the pixel values for Diff the lowest corresponds to the coldest temp and is (x=130,y=33) 975 the hottest is (x=157,y=45) 8658.
The corresponding values for the rawData are 6625 and 15766.

   ...ken...

btw, the CImage is rotated 180 from the others...
And, I make no claim that the lowest temp at x=130,y=33 actually corresponds to -30C or that the hottest at x=157,y=45 corresponds to 300C,
it was several seconds after I froze the sensor and the temp registered ~-30C that I took the picture. Emmisivity of the surface is a total crap shoot since there was most likely condensation on a black plastic sensor housing and it took several seconds after freezing for the thermometer to drop down to ~-30C.
So, folks beware of drawing any conclusions from my images.
Yep the probe was cold, and the solder gun was hot.

[efahrenholz]

@eneuro,

I'm not trying to offend you, but I'd appreciate it if you refrained from calling me a failure. The fact that I have a hard time figuring out what you are trying to say wasn't intended to make you sound ignorant. In fact I even tried to defend you by proposing it's a language barrier. It's a misunderstanding, let's leave it at that.

[eneuro]

The corresponding values for the rawData are 6625 and 15766.

It was too interesting data @cynfab to do not to try to use sixth sense and try to predict absolute temperature in [*C]  based on this additional Seek raw data

Code: [Select]

Seek Thermal sensor: raw: 6384 temp: -40.0 [*C]
Seek Thermal sensor: raw: 6625 temp: -31.1 [*C]
Seek Thermal sensor: raw: 7903 temp: 16.2 [*C]
Seek Thermal sensor: raw: 15766 temp: 307.1 [*C]
Seek Thermal sensor: raw: 16384 temp: 330.0 [*C]
Using this simply formula cold part (-31.1*C) and hot iron (307.1*C) looks good, as well as... Seek claimed thermal object detection temperatures -40*C .. 330*C 

But it is too simple formula to be true and it is based not on exact values which you showed, by my guess what this approximation function could be.
Data which you showed fits very nice into this approximation function which is very easy and estimate temperatures are fine, however we could expect nonlinear relationships... 

Is it possible, it could be around 16 *C only inside this Seek dongle when you made this experiment? 

BTW: Attached your senssor bad dots pattern which looks very similar to this calibration frames from your earlier experiments with hot iron, so probably it is only 23 such dots-more frames needed to make additional difference change comparisions, but in the case of @frenky data additional check lead to the same amount of those dead looking pixels while they were the only not changing on other 14 calibration frames he provided-excluded perfect hexagon pattern and 207 & 208 column of course 

[eneuro]

That must mean that my stepper motor is making 0.1747 DEG (sub)steps and not 0.225 as I thought.

36 DEG/206 is exactly: 0.174757282 ~  0.175 DEG, so very close to those (sub)steps.

I've divided by mistake by 208 in previous post and got slightly lower value, so this is correct one 

[eneuro]

I think your method to find unresponsive pixels is flawed because mine does have plenty of them.

Could you @miguelvp reveal your algorithm for finding such huge amount of dead pixels on Seek frames?

I've made all permutations (91 differences) of those 14 @frenky calibration frames and in this simple way:

Code: [Select]

...
for(int li=0; li<14; li++ ) {
 for(int lj= li+1; lj<14; lj++ ) {
  lcount++;
   calprvf= calsf[li];
   calcurf= calsf[lj];

   caldiff= calcurf- calprvf;
 ...
and after marking dots that  changed at least once (difference was not zero) I've got the same result as with simple <128 threshold before, because of when we add to this changed pixels map:

Code: [Select]

../sts_analysis: Calibration difference: #11 vs #13  number: 90  mean: -2.580  deviation: 141.947
../sts_analysis: Calibration difference: #12 vs #13  number: 91  mean: -0.440  deviation: 441.066
../sts_analysis: Calibration frame change pattern of #91 differences written to 8bit gray file: calchgpat.pngcompletly dead pixels map created from ONLY one calibration frame we can see that that it looks like this dead dots map is fine and no chance to find any other unresponsive thermal image pixels 

[Fraser]

FLIR have just sent me an email newsletter on the LEPTON. May be of intereset to SEEK readers as it is the only competition at the moment.

http://cvs.flir.com/webmail/6132/582294369/392afc20347a08c88115a301559fa4ae

FLIR are making their LEPTONs available to hobbyists via an 'agent' who runs group buy campaigns.

The resolution is undeniably lower but the LEPTON is still competition to the SEEK in terms of image processing.

Aurora

[nabla]

Hey Folks,

please apologize my stupid question, but I don't know where to ask that except on this thread.

I'm about to buy one of those cams. But unfortunately I live in Europe and those cams aren't sold here at the moment. I cannot download the app via the Play Store as well.
Nevertheless I want to use this cam (and I think I got some sources to get it shipped to Europe) .
Since some of you posted the *.apk here I'm thinking about if that really works with the cam. I read about a login at the App...so I'd really appreciate if somebody can help me with that.

Thank you!!!

[frenky]

You will need a phone or tablet with USB OTG support. Just check for your model on gsmarena. If you see "USB OTG" or "USB Host" you'll be fine.
You will also need to enable "Unknown sources" on your device. After that just download apk from: http://apk-dl.com/download/android/com.tyriansystems.SeekThermal/Seek-Thermal1.4.0.2-apk/
and install it on your phone. No need to register, you can skip that if you don't want to do it. App will work just fine without.

[nabla]

Oh well yeah forgot to tell you, I got all the other required stuff at home 

Just wasn't sure about the app cause the Play Store says "incompatible with all of your devices" ^^

So thank you very much!!!

[bhallier]

Is there anyone on the forum working on the firmware side of the device? I remember awhile back some talk about the sensor fps possibly being limited in the firmware. Do we know if that is true?

[eneuro]

I remember awhile back some talk about the sensor fps possibly being limited in the firmware. Do we know if that is true?

Didn't check it yet, but since we want to use circular motion to enhance Seek thermal output quality it could be nice to hack this <9Hz limit 

For the moment the main concern might be to be carefull when conecting this Seek dongle to Andorid devices with different app versions, while it looks like older app CAN TRY DOWNGRADE firmware ON PCB in your Seek Thermal just to match with its own which comes with apk 

It is in their website there Seek Thermal Camera Firmware Updates FAQ .
Probably I could imagine that when inserting my Seek to friend Android with old app version, when this app detects newer firmware in this dongle WIILL RATHER ask  to upgrade itself and do NOT OVERWRITE FIRMWARE in  my Seek dongle

[Uho]

Dismantled body imager. He was sealed with glue. The lens is very thick. Sawed off 3.5 mm. Thickness more. Lens material sawed needle files. Gradient disappears if you use an external shutter.

[Fraser]

Regarding the lens. It will be a single element reverse telescope type and these are normally quite thick lenses by design. I will add a picture of such a lens type when I find it.

On the topic of dead pixels, a commercial microbolometer will generally have one of two dead pixel specifications: Type 1 = 99.9% or better operational pixels, Type 2 = 99.8% operational pixels. Only 0.1 % between them and this is the normal range found in a manufacturers product range.

Either the SEEK is using a seriously flawed microbolometer fabrication process (unlikely) or the dead pixel maps that eneuro is seeing are not a true reflection of the non functional pixels on the array.  Eneuro, are you able to get a total number for those dead pixels ? I ask because sometimes dead pixel maps do look awful but the FPA still passes the 99.8% operational pixel requirement.

The dead pixel map lives inside the ROIC pixel processing and that is where it should be extracted from and not the USB output of a device. All manner of changes can take place inside the ROIC processing stages.

Aurora

[-jeffB]

Is there anyone on the forum working on the firmware side of the device? I remember awhile back some talk about the sensor fps possibly being limited in the firmware. Do we know if that is true?

I'd love to see a faster frame-rate; I'm sure most of us would.

But if it turns out that you can hack the device to increase its frame rate, might that expose Seek to legal action? I'd hate to see a hack that gave us lucky adopters more FPS, but caused all new shipments to be blocked.

The fact that it's apparently easy to overwrite the camera's firmware should be of great interest. I'm reminded of the infamous firmware hack that let researchers disable the "hard-wired" activity light on MacBook cameras. If there's a path to get at firmware, all sorts of possibilities open up...

[mikeselectricstuff]

For the moment the main concern might be to be carefull when conecting this Seek dongle to Andorid devices with different app versions, while it looks like older app CAN TRY DOWNGRADE firmware ON PCB in your Seek Thermal just to match with its own which comes with apk 

It is in their website there Seek Thermal Camera Firmware Updates FAQ .
Probably I could imagine that when inserting my Seek to friend Android with old app version, when this app detects newer firmware in this dongle WIILL RATHER ask  to upgrade itself and do NOT OVERWRITE FIRMWARE in  my Seek dongle

What's the problem with that ? As long as they don't do anything that breaks the FW update process, it will just re-update it if connected to a device with an app that needs a later version.

I've not read the MCU datasheet for a while but I have a feeling that the normal mode of operaton is for it to load the firmware from the external flash into RAM and then execute it, in which case you should be able to load new FW via JTAG/SWD for experimentation without affecting the flash version.
I know that later serial flash devices support  read modes fast enough to execute direct from them - a quick probe around with a scope would tell you if this is the case or not.

I observed what appeared to be a framerate of around 30FPS from the sensor to the MCU, so higher rates are probably possible, however the image will be even noisier than with the standard lens as it wouldn't have the multi-frame filtering, so may only be of use with a bigger, more expensive lens.


   

[eneuro]

I observed what appeared to be a framerate of around 30FPS from the sensor to the MCU, so higher rates are probably possible, however the image will be even noisier than with the standard lens as it wouldn't have the multi-frame filtering, so may only be of use with a bigger, more expensive lens.

It is a good news and maybe even some of the operation modes already allow select different one with higher frame rate, but probabbly not while there is this <9Hz requirement for those thermal cameras, so it have to be hidden inside MCU firmware and not easy to change just by guessing this code and setting this mode as we saw in Seek jave source code.

Anyway with circular motion we have multi-frame filtering and more- we can "see through" those useless hexagon pattern "patent" 2.1k pixels, because turning Seek camera right,up,down,left in eliptical motion will allow see the same thermal scene shifted 1,2, pixels and we have real theraml pixel in place of this hexagon "patent" or bad pixel, so in some applications with higher framerate  we can get much better output thermal image quality, while output frequency can still be <9Hz (even 1Hz)  but stacking much more frames per second in circular motion approach is advantage   


BTW: One of the reasons why I do not like Seek app to downgrade firmware version because of it is common that software which detects newer version of firmware asks me if I want upgrade software to latest version, but for the moment it is not issue while there is only a few versions of Android app and probably all have the same firmware *.bin files.
When someone wans to play with this device using his own app, then of course firmware change could be issue if not detected, so downgrade to older firmware version which can happen with Seek device used in many different future app versions on different android devices could be  surprise, so it is rather warning for people trying to use own code to talk to this Seek device and also using oryginal Seek app 

[feral2]

I've been following this thread for a while and very much appreciate all the work and thought going into this. One quick contribution I wanted to make that I verified with my own Seek -- for those looking for a cabling solution to hook into a standard USB port (at least on a PC), this is one relatively inexpensive and convenient option:

This 3' micro USB extension cable
http://www.amazon.com/YCS-Basics-Micro-female-extension/dp/B00HAOKCE8/ref=sr_1_1?ie=UTF8&qid=1418051899&sr=8-1&keywords=B00HAOKCE8

Combined with this little micro USB -to- USB A adapter:
http://www.amazon.com/Micro-Female-Adapter-Worldwide-shiping/dp/B009AWA3VK/ref=sr_1_1?s=electronics&ie=UTF8&qid=1418051918&sr=1-1&keywords=B009AWA3VK]

Best regards,
feral2