SeekOFix - new windows software for SeekThermal

[frenky]

I think It would be better to have this new software in a new thread...

So this new windows app was based on the work of forum members: sgstair, miguelvp, jadew. I just added a bit of my own touch. 

The goal of the app was to get clean images out of very noisy Seek thermal sensor.

Sample photos and attached SeekOFix v0.1 
Raspberry Pi model B:
100mm ZnSe lens:


50mm ZnSe lens:


50mm ZnSe -> Smal PCB of USB SD card reader (1" width):


Room corner with a 3*C temp range:


How to use SeekOFix:
Run it...
If you click on AutoSave checkbox the image will be saved in the "c:\seek\export" folder after every shutter calibration.
If you want to use external calibration:
- place an isothermal plane in front of seeks lens
- wait until you hear the click of shutter calibration
- press the "Calibration: Use frame" button
- remove the isothermal plane

If you wish to save this frames then you should check AutoSave checkbox before doing external calibration.
After you have done that all frames between this external calibration and next shutter calibration will be saved in the  "c:\seek\export" folder.
You can repeat that as many times you want...

P.S.
After I took this RPI images It was probably the first time that I thought at my self "perhaps Seek is not total crap after all".

[miguelvp]

Cool, thank you very much for doing this.

[frenky]

Well I used a lot of your code to, so thanks to you.

I was kind of forced to make a new app for Seek because I'm saving for a house and my wife would kill me if I would spend 1000€ (with VAT) to buy i3 ThermalExpert.

[frenky]

New version with:
- added display of the first image after calibration (it has the least noise because pixels drift with time)
- autorange based on histogram (no visible histogram in this version)
- dynamic sliders (for fine tuning the desired range)

[efahrenholz]

It looks like you really got the image quality right where it needs to be to be useful. If I could suggest, how about an image sharpness filter? Even a basic unsharp mask or a regular sharpening filter would really put the final touches on determining what can be pulled out of the sensor. If you manage to implement a sharpening filter, I just might go back out and buy a new Seek Compact. I never thought i'd go back, but damn frenky, you really got something good here. Seek should take some advice from all the work here to improve the app on the phone. You guys are making the image look way better than it ever originally did before.

[Fraser]

Excellent work. Impressive results. 👍

[frenky]

If I could suggest, how about an image sharpness filter? Even a basic unsharp mask or a regular sharpening filter would really put the final touches on determining what can be pulled out of the sensor. If you manage to implement a sharpening filter

Probably it could be done... But I don't know what kind of result would you like?
Could you take above images and apply some filters on them so I can see the desired result?

I did quick unsharpen mask in Irfanview but to me this looks worse because the noise really pops out:
(Perhaps some other filter would be better?)

Left original, right unsharpen mask:

[frenky]

Excellent work. Impressive results. 👍

Thank you. 

[efahrenholz]

Frenky,

That unsharp mask filter is just about perfect. It's not too much and look at how it brought out the faint circuit traces. Noise aside, I'd say it's a fair trade-off.

[frenky]

I have found this: http://www.aforgenet.com/framework/features/convolution_filters.html
Check out the results of the "Sharpen" and "Gaussian Sharpen" filters.

[frenky]

New version v0.3. 

- changed dynamic sliders a bit so it's more user friendly.
- saved files are now in 412x312 resolution
- added sharpening with AForge.NET: http://www.aforgenet.com/framework/

New libraries are in AForge_libs folder. But It should work out of the box...

Original & with sharpening enabled:

[rodpp]

Amazing! Thanks for sharing.

Enviado de meu XT1572 usando Tapatalk

[frenky]

Just for fun:

[Fraser]

Now if the same can be achieved for the FLIR One Gen 2 I will be a very happy bunny

[Bill W]

Noise appearing on colour palettes is a brain-eye interface problem.  It is one reason that so many imagers tend to offer monochrome as the default, the brain hates chroma noise but happily averages luminance noise.  You can do a lot by trying hard to smooth the rate of change of each R, G and B channel, which is where 'Iron' and similar schemes have been / can be optimised.  For a known scene type like a PCB you could do a better job than the manufacturer by optimising for that specific outcome.  You know there is a lot of ambient, nothing much cold and are not too bothered about distinguishing hot and very hot

ImageJ is really good for trying out palettes on corrected raw* images for those that have not tried already.  With access to sensor ADC data you can also replicate the whole factory calibration and 1-NUC process in it, or export the 320x240 to a spreadsheet for processing.

* by corrected raw I mean 16 bit with a known data value - temperature relationship, as opposed to 'sensor ADC data'.

[frenky]

Interesting...
I don't really like sharpening effect on the Iron palette but on grayscale it's not that bad:

[Speeder]

Nice work!

Will it also be possible to analyse stored pictures with your software?

[frenky]

Yes for sure...

My current to-do list is:
1. Histogram (mostly done)
2. Display temp values instead of sensor values *C/*F (I need one afternoon to calibrate seek with handheld IR gun thermometer...)
3. Implement various color palettes and enable users to add their own (mostly done)
4. Capture video (if AForge.NET won't cause any problems)

After that I intended to start working on analytical part of an app...
But now that I'm thinking about it, the best solution would be to just export data in FLIRs thermal image format, so we could use Flir tools to inspect Seek's files.

[santasemilla]

Hi frenky and everyone. Sorry if this is a completely stupid question, I'm a noob in this field. Got my Seek XR last November for Energy Efficiency inspections and I'm quite disappointed with the original soft as all of you are. I've been following the progress of the different topics in this forum and I'm impress with the capabilities you got out of this gadget! Is there any chance to port kind of this soft to android? I know that you are using specific libraries that are only windows compatible. Just a question. Thanks for the hard work, to you and all the team!

[Fraser]

Very nice work Frenky.

I may have to buy a cheap SEEK camera and use it with your software

Fraser

[frenky]

Is there any chance to port kind of this soft to android? I know that you are using specific libraries that are only windows compatible.

I don't think I'll be porting it to android. Mainly because I don't have the knowledge to make android apps. 

I will go into a different (hardware) way: RPI2 + 5" TFT + 5000mAh powerbank + nice handle + Seek with external shutter

[frenky]

I may have to buy a cheap SEEK camera and use it with your software

Why on earth would you do something like that? (With my software it's just a little less crappy...)
If you want one I'll trade mine for one of yours 320x240 cameras... 

[Fraser]

Frenky,

I know the limitations of the SEEK hardware well I did own one.

I am just hopefull that someone clever like you can 'polish the turd' a little  with better software I can get my hands on an unwanted SEEK for around £50 so it may still be worth having it in the collection

I really want SEEK to succeed and that maybe why I do not want to completely abandon them ?

Fraser

[frenky]

I really want SEEK to succeed and that maybe why I do not want to completely abandon them ?

Ok I see your point.
I also wish them to succeed because more competition is better for customers. I just wish they would improve on their design of the sensor and shutter...

[santasemilla]

Is there any chance to port kind of this soft to android? I know that you are using specific libraries that are only windows compatible.

I don't think I'll be porting it to android. Mainly because I don't have the knowledge to make android apps. 

I will go into a different (hardware) way: RPI2 + 5" TFT + 5000mAh powerbank + nice handle + Seek with external shutter

 

I was just curious, there was a guy working with the seek and a Teclast X98 tablet (a pretty nice dual boot one). He said everything was working great but as far as I saw, the mini USB port was in the wrong possition for the Seek camera... I'm thinking on experimenting with a cheap Windows tablet to try the same, just making sure the port is oriented as it should!

Thanks again for your work frenky