Yet another cheap thermal imager incoming.. Seek Thermal

[eneuro]

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.

Examined many different calibration Seek frames and it looks like that range of unresponsive pixels visible from application on other usb side in Seek dongle rather easy fits into this 0.1%-0.2% of bad pixels, however when we take into account those hexagon useless "patent" pixels 2.1k on 32.1k 260x156 sensor which is  6.5% blanked hexagon pixels +/- 0.1%-0.2%  dead pixels it is not what can be seen on Flir calibration frames if not hidden somehow 

Tests show this range 0.1%-0.2% bad pixels (in the case of @frenky data provided here which everybody can examine oneself also made cross veryfication by combination of difference of each 14 calibration frames-total 91 comparisions):

pat.cynafab_temp_300-30_rawCal.png  bad pattern pixels: 23/32136  (0.072 %)
pat.marshallh.raw.000000007.0x1.png  bad pattern pixels: 28/32136  (0.087 %)
pat.frenky_osc_motor.raw.000000027.0x1.png  bad pattern pixels: 79/32136  (0.246 %)

However, when we exclude those hexagon useless pixels (from application point of view-no image thermal information there they have to guessed by image postprocessing)-than 206x156 pixels Seek matrix has not 32136 "good pixels" with thermal image data, but  2.1k (2143) less, so those numbers above will change, but not too much:

23/(32136-2143): 0.077 %

Still this huge amount of those blanked hexagon "patent" pixels (6.5%) is much bigger then those bad pixels count and we can speculate if it is some kind of trade secret in Seek Thermal device-just preparing simulations how it affects output thermal image quality.

Hopefully, it looks like circular motion can help "see through" those hexagon 2.1k pixels in some Seek custom DIY applications and even this simple @frenky experiment with stepper motor and horizontal motion showed it might help improve output thermal image quality 

[Stoffel Janssens]

Deal all

I was wondering what the minimal working distance of the Seek Thermal device is.
The minimal working distance is the minimal distance from the lens to an object with the requirement that the object is in focus.
I'm performing some research on liquid droplets (2 < 3 mm) which are several degrees hotter than the environment and I just need a thermal image of them (relative temperature is OK) without spending thousands of dollars of tax money. There are some cheap lenses available (ZnSe) which can help a lot with this experiment but it would be nice if there are some people who can provide me some values about working distances with and without these lenses.
It would also be great if someone just deposits a small droplet which is heated a bit on a surface, just to test.

Another question is about the possibility of connecting the Seek Thermal device to a PC (windows) with some kind of trick and still running the standard app provided for a device with Android as an operating system.

You guys are doing great work!

[-jeffB]

I was wondering what the minimal working distance of the Seek Thermal device is.
The minimal working distance is the minimal distance from the lens to an object with the requirement that the object is in focus.
I'm performing some research on liquid droplets (2 < 3 mm) which are several degrees hotter than the environment and I just need a thermal image of them (relative temperature is OK) without spending thousands of dollars of tax money. There are some cheap lenses available (ZnSe) which can help a lot with this experiment but it would be nice if there are some people who can provide me some values about working distances with and without these lenses.
It would also be great if someone just deposits a small droplet which is heated a bit on a surface, just to test.

Hey, I can do that!

My procedure:

Spread out some wax paper (non-wetting, non-absorbent) on a surface at ambient temp (~20C -- the Seek was indicating ~77F, but I think it's off considerably).

Get my hand wet with some tepid water. Splatter it onto the wax paper. Image it, and realize that the water's pretty close to ambient temperature, too.

Blow across the paper. Image again. Yep, the droplets have cooled evaporatively.

Now, get some HOT water (~40C) on my hand, and splatter it. Image again. See the nice blue-black cold droplets, and the nice yellow hot droplets.

Now, run and find a tape measure with a metric scale (that was embarrassingly hard), and get a visual picture with it in the frame.

I was holding the camera at maybe 15cm distant for the thermal image. I tried other images at 5, 10, and 15cm, but it was a bit hard to tell anything definitive about resolution -- it's just plain fuzzy at most distances. With these optics, I'm thinking the effective "as good as it's going to get" range is somewhere from 5cm to infinity. Others have posted much more detailed imagery here to address that.

Things I'd do given unlimited time:

1. Set up a rig to hold the camera at fixed, calibrated distances from the target.
2. Use oil droplets instead of water to eliminate evaporative cooling.
3. Use a liquid-crystal mat to check the ACTUAL temperature of the droplets? I know I've got a liquid-crystal mouse pad somewhere around here...

...but, alas, none of this is likely to happen very soon. Busy week. If you've got suggestions for other quick things to try, maybe I can help, but probably not tonight -- getting close to bedtime!

[Stoffel Janssens]

Thank you very much -jeffB!!!!

Your experiments help me a lot! No need to conduct more experiments! Please take a good rest!
The most important question you solved is if the resolution of the camera is high enough to see small drops with slightly different temperatures than the environment. And without any layering of images, ... the resolution is already good enough for my experiments. To obtain actual temperatures could be a pain in the *** but the camera can probably qualitatively prove my theory and therefore it is interesting buying the camera.

When I obtain some data, I'll surely share it on this forum. Now lets see if the camera can work on a PC ...

Cheers and again, thanks a lot!

[barehill]

Thanks Uho for the great closeups!

FLIR mentions their modules are "export compliant, <7 HZ." Some sensitive software and electronics are export restricted by the US State Dept to hobble features. Bolometers have an inherent time constant so we don't know the useful frame rate yet.

[igor88]

hi

will it be a FAQ in the first post about that device?

its nice for that price... but gradient....

[Uho]

I have a big request. Need to check the program. I translated into Russian version 1.3. This is a trial version. Still need editing. You need to check if it works. My phone is not available. I can not check myself. I need to verify the correctness of the signature. Who can send me an email. I will send you to an email. Large file. Share on site fails. Thanks in advance.

[eneuro]

Another question is about the possibility of connecting the Seek Thermal device to a PC (windows) with some kind of trick and still running the standard app provided for a device with Android as an operating system.

Do you need to monitor experiment, so Android app and PC apps running at the same time or there are other requirements there in this project?

Probably in my setup small Linux distro on pendrive or DVD or even diskless machine with network boot from old PC could be fine and running small web server could enable sending via network or wifi image from this dongle on Android device easy without need to mess with Window$ and this oryginal Seek app while probably more advanced filtering than those implemented in Seek app can be needed to deconvolve thermal images of droplets taken from this Seek dongle 

It looks like Seek app might use kalman filter and bilateral with median in 1.4.0.2 version of their java native library: /com.tyriansystems.SeekThermal_1.4.0.2-20141113/lib/armeabi/libSeekware.so

Code: [Select]

kalman_deinit
kalman_init
kalman_process
_Z10med5_fixedPi
_Z10med9_fixedPi
_Z11median_compPiS_
_Z13kalman_updateiP14kalman_private
_Z14histogram_clipPjj
_Z15create_transferPjPtj
_Z15median5_processPtS_ii
_Z15median9_processPtS_ii
_Z16bilateral_filterPtS_iiff
_Z16histogram_createPtjjjPjS_
_Z16median5x_processPtS_ii
_Z16median9f_processPfS_ii
_Z7aeg_abss
_Z7filter_PtPij
_Z8opt_med3Pt
_Z8opt_med4Pt
_Z8opt_med5Pt
_Z8opt_med9Pt
_Z9opt_med9fPf
However those Seek android app thermal images and video looks very bad, so looking for ways to use more advanced filtering including hardware circular motion support to "see through" those blanked hexagon patent pixels.

BTW: In the more tricky setup but still quite easy to setup, running virtualbox on Windows$ with Linux Ubunto as guest to provide diskless boot of dedicated small Seek server PC networked machine could be advantage in laboratory while no need to setup PC with disk, messing with Window$ install-just let it boot via network and let send captured imaginery from Seek device or provide live WWW content to any other operating system including latest 42" wi-fi and web browsers enabled TV sets, etc for live presentation for viewers anywhere in the world 

[eneuro]

its nice for that price... but gradient....

It is quite easy make automatic correction of linear surface component of this gradient on Seek thermal output, while nonlinear part is more tricky and not fully automated for the moment, but early Hubble Space Telescope images were distorted by a flawed mirror and could be sharpened by deconvolution 

If you provided also visual light photo of this interesting scene, where there is small difference between maximum and minimum temperature, so those color gradients will be more visible, people watching thermal image like this could have better idea what is in the background, while those Seek output images for the moment are very difficult to understand for viewers watching only IR-probably also only 8bit thermal LUTs are used in Seek app so not so easy to see anything but the noise there 

BTW: Did you tried save this image in 100% jpeg quality? -then by applying reverse LUT viewers could try to do futher image processing and try to deblur this image but probably is too noisy to do this...

[Stoffel Janssens]

Do you need to monitor experiment, so Android app and PC apps running at the same time or there are other requirements there in this project?

Dear Eneuro

Thanks for the fast and rigorous reply! I will take your ideas in serious consideration.
The hardware circular motion is a nice way of capturing good images. You need to do something like that to solve the blanked pixel problem.

Now I will buy the device and start playing with it. Probably will use some of the nice programs written by forum members. I'm not a hard core electronic engineer like most of the guys on this forum but I'm very happy I can get lots of help over here.

Cheers and many thanks!!

[eneuro]

The hardware circular motion is a nice way of capturing good images. You need to do something like that to solve the blanked pixel problem.

Maybe inserting a few NOPs to Seek firmware will... disable this intenional blanking of those hexagon anti patent pixels, so another reason after effective Seek java app disassembly to look around in those Seek firmware files and try to guess if pixels are blanked in hardware during sensor manufacturing or simply in this firmware? 
./app/com.tyriansystems.SeekThermal_1.4.0.2-20141113/com/tyriansystems/Seekware/SUBI_LPC43XX_LPCOpen_1.3.0.0.bin
./app/com.tyriansystems.SeekThermal_1.4.0.2-20141113/com/tyriansystems/Seekware/SUBI_LPC43XX_LPCOpen_1.2.0.0.bin

It was on Seek facebook a while ago:
"

Dark Pixels. No great mystery here. Every 15th pixel is intentionally blanked to avoid a potential patent infringement. Seek has an updated design for future product that eliminate the need for this measure. With the effective blur length of a 12 micron pixel resolving 8-13 micron Radiation, loss of single isolated pixels does not (in itself) degrade image quality."

What do they mean, which kind of measure is hidden in those hexagon pattern pixels-sensor temperature detection? 

Seek has an updated design for future product that eliminate the need for this measure

Any ideas what patent infringement could be if they didn't blanked somehow those 6.5% thermal image pixels?

This is huge amount of useless pixels and when we substract from only one image calibration frame and add 207th column average those hexagon pixels looks really bad before filtering and lets see how much they destroy output thermal image quality-it will be much more significant if this Seek cam were mounted on stationary tripod, while keeping it moving might help "see through" them.

[-jeffB]

What do they mean, which kind of measure is hidden in those hexagon pattern pixels-sensor temperature detection? 

Seek has an updated design for future product that eliminate the need for this measure

This is a different definition of "measure".

You're thinking of "measuring something" as determining its quantity or value.

"Taking measures to accomplish something" means "doing things to accomplish something".

So, "eliminating the need for this measure" means "finding another way to avoid patent infringement so we don't need to blank out pixels".

[JcDenton]

Does anyone know if it is possible to use a lens to focus to infinity or distant objects with the seek?

[Fraser]

The SEEK uses a classic reverse telescope fixed focus lens. It is capable of focusing a usable image at both close range and infinity. It is a compromise as focus is never as sharp as a manual focus lens.

If your question is whether a telescope auxiliary lens may be used with the SEEK, then the answer is yes BUT first you need to find such a telescope lens at reasonable cost. The lens you need is designed to be non inverting and to sit in front of the cameras objective lens. Such telescopes are very expensive and the lack of suitable cheap ZnSe lenses makes a DIY telescope somewhat hard to achieve.

As a price guide, my large FLIR X2 telescopes cost me approx $400 each via ebay..... that was a bargain price I might add as they cost several thousand Dollars normally. My Inframetrics X3 telescope lenses cost around $160 each and are sadly an image inverting design so not 'plug & play'. These were also bargains as they originally cost even more than the FLIR telescopes. Modern new telescope lenses are very expensive and make little sense when the SEEK camera is so cheap. $200 for the camera, $3000+ for the lens   It is a pity that affordable thermal camera auxiliary lenses are not more common.

This thread on the FLIR E4 contains information that may be of interest to SEEK owners as well. I discuss focus points and auxiliary lens options.

https://www.eevblog.com/forum/reviews/flir-e4-the-useful-information-thread/

Aurora

[JcDenton]

Thanks for the response aurora, I will take a look at the thread.  I did order a cheap ZnSe lens off ebay for close focus, not even sure if it will work.  I find that I can't really see anything at long distances with the seek which is why I was curious, but I guess you can't expect too much from a low cost imager like this.

[-jeffB]

The lens you need is designed to be non inverting and to sit in front of the cameras objective lens.

...or you could just buy a 6-inch Micro USB extension cable, like those of us who are stuck with reverse-oriented USB ports, and rotate the Seek 180 degrees. 

[eneuro]

I find that I can't really see anything at long distances with the seek which is why I was curious, but I guess you can't expect too much from a low cost imager like this.

Trying to figureout and calculate if those lenses mentioned in this thread

Has any1 tryed to isntall this lens  http://therm-app.com/product/19mm-lens/  on to a flir e4 cammera since it is fairly cheap 19mm narow fov lens that could give a longer detection range (at least i hope so ) there would be some housing modifications since i dont think the threads on the lenses are the same...

Optical specs below:

Maybe with custom lens holder it could be possible mount those lenses and add also custom shutter with cicular motion plate inside similar to this from Flir E4 and still be in the range below $500 with Seek Thermal "enhanced" sensor (firmware hacked).

@Aurora
Any good tutorials how to choose correct lenses for given thermal sensor unit parameters?

[jaybeez]

I find that I can't really see anything at long distances with the seek which is why I was curious, but I guess you can't expect too much from a low cost imager like this.

Trying to figureout and calculate if those lenses mentioned in this thread

Has any1 tryed to isntall this lens  http://therm-app.com/product/19mm-lens/  on to a flir e4 cammera since it is fairly cheap 19mm narow fov lens that could give a longer detection range (at least i hope so ) there would be some housing modifications since i dont think the threads on the lenses are the same...

Optical specs below:

Maybe with custom lens holder it could be possible mount those lenses and add also custom shutter with cicular motion plate inside similar to this from Flir E4 and still be in the range below $500 with Seek Thermal "enhanced" sensor (firmware hacked).

@Aurora
Any good tutorials how to choose correct lenses for given thermal sensor unit parameters?

these are a slightly cheaper option than the therm-app lenses. they work on the therm-app, and there is a larger selection.
http://eom.umicore.com/en/infrared-optics/product-range/35-mm-f-1.1/

[Uho]

Application lenses 4 and 7 degrees. As you can see, everything is possible.

[eneuro]

these are a slightly cheaper option than the therm-app lenses. they work on the therm-app, and there is a larger selection.
http://eom.umicore.com/en/infrared-optics/product-range/35-mm-f-1.1/

Thx for this link 
It looks like those lenses could work with Seek 12 um sensor while they claim that it might support this kind of next-genartion microbolometers as well as lower than VGA resolutions 

So, lets try to see their technical specs and try to design custom lens holder based on Seek oryginal one but with shutter closed inside.

Now it is time to review maybe available shutter manufacturers to find similar to this quite nitty compact design used in Flir E4:

Do you know any shutters modules like this above available to buy for custom projects?

[Uho]

Fixing the electronic board is difficult to make. Manufacturer minimizes costs. Even the case is not dismountable. Need a good alignment. I can not securely attach the lens. Think. The lens of heavy. Glue will not hold.

[marshallh]

Everybody please check your credit card statements.
I received a "REFUND CONFIRMATION" email from Seek for $199 for the first order I placed. Completely out of the blue.

Then I checked my bank statement and they had debited $199 from my account.
Hoping I can get this straightened out.

[nkw]

I received my Seek Thermal from Amazon a few days ago and have been playing around with it. A few observations:

1) Software (v.1.4.02) seems really unstable on Android 5.0 Lollipop running on a stock Nexus 5. Usually it will immediately crash the first time I run it, but once relaunched it will stay running.

2) When first plugging in the camera and starting up my camera seems to get stuck in some sort of mode where I can hear the shutter click once per second or so continuously and the image appears to be either blank or what might be a calibration image. It looks noisy and indicates two nearly adjacent points at 32 deg F and 626 def F simultaneously. I've attached one of the images. Eventually I have been able to usually get the camera to start working by unplugging and replugging a few times along with stopping and restarting the software. For the first few minutes out of the box I was really disappointed thinking my unit was completely borked. I don't know if this is a software issue or a hardware issue with my unit, so far it has exhibited this behavior each day when I initially connect the unit.

3) The gradient issue is pretty bad. Looking at some of the previous posts in this thread I didn't think it would be that bad, but it is really annoying and on mine seems to make about half the display unusable (probably a quarter of it just fades away depending on the mode). I'm seeing about a 12 deg F difference between opposite corners. Attached is a picture of a uniform surface.

4) Cable - Buy this one: http://amzn.com/B00HAOKCE8 not this one: http://amzn.com/B00B5HSCW0 - I thought I was buying an OTG compatible cable when I was checking out, but apparently I can't read. I have now experimentally determined the first one works and the second one does not.

I'm hoping all these issues can be fixed via software updates by the Seek Thermal folks. I'm going to wait until they push the next update to determine whether or not #2 is a hardware problem specific to my camera I might need to contact support to resolve.

Overall, it is a pretty nifty device for the price. If they have success with a software fix for the gradient issue I think it is definitely a winner. If not, it still has a bit of a gee-wiz factor, but I don't think it will be useful to me for much of anything. I would probably go ahead and up the money for a Flir handheld -- I've been spending waaay too much time reading through the Flir E4 thread.

[eneuro]

3) The gradient issue is pretty bad. 
....
I'm seeing about a 12 deg F difference between opposite corners. Attached is a picture of a uniform surface.

What was  distance from Seek dongle to this surface and what kind of surface it was?
Publishing only Seek thermal fuzzy image which can be created for example very close to table surface or with other reflected IR sources in bad artificial lightting conditions might easy create gradient.
How do you know this surface had uniform temperature? Did you made measurements of temperature in a few points to ensure this surface temperature is uniform?

If there were applied thermal LUT not from its very cold to maximum color ranges than this "gradient" sometimes simply can exist in real world too, so while there is no visual light image, there is very difficult to say there shouldn't be a temperature gradient.
Another thing how long Seek  was  running when those images were made and what was storage temperature of this device and operating air room temperature.

We are dealing with thermal events which are not visible at daily basis, so first we need to ensure that this what we think is true, so more measurements needed, not just turn thermal camera on, make some screenshoot, but environment in which it were used very carefully examined as well as thermal phenomena taken into account...

On some Seek cams I see not such a huge linear gradient when its calibration frames are examined, but this is on device probably running longer-not just switched ON to record gradient.

./sts_analysis: Calibration image surface:  z0: 8003.694395287  zx: 8008.635499644  zy: 8002.172119205  zmin: 7997.231014848  zmax: 8010.157775727  dz: 12.926760879

Its difficult to say that this small gradient on calibration surface makes such a huge difference in temperatures (6*C) on expected uniform thermal object surface.

BTW: There are people which will never buy any Flir equipment for very simple reason-Flir issued lawsuit to delay thermal imaging at lower costs and payed $36M for this stupid move, but they lost much more-their image and people who know it won't buy their products, because of money are not everything...
Anyway be carefull when reading posts of any forum members with only a few posts created when Seek launched, while Flir's "agents" will try to spam Seek threads to fight for his lost "piece of cake" ...

[maex]

A new Seek Thermal app is available on the playstore, old 1.4.0.2(76) new 1.4.0.2(81).