3d recostrucrion of thermal data

[joe-c]

...There is a way for use the same strech for every CSV in the folder. I think that all CSV must be processed for looking for a min and max temperature before restart the process and create the jpg. So after mosaic all of image i can create a color legend...

Done in the new V1.0.3.0
It still processes the images like before. but now a second run for the folder generates Images (without Scale and with the same "stretch", just the Image in the Color scale you have chosen).
They all in the new folder "Scaled_JPG".

I hope it helps.

[tuffi]

Hi Joe...I make some test and...yes it works (in the next week i will make a test with a drone for thermal mapping).

But I have a couple questions for you.
Using a 8 bit jpeg i have only 256 different gray value, if the delta T is very large the resolution decrease until 0.5°C. Do you think that is possible export data in 16 bit TIFF?

And: with some dataset the scaled output is ok, but in the streached preview folder all of the exported image is the last image of the CSV folder and during conversion Thermovision give me an error in the bottom bar...but the scaled outpud is ok!! It's strange? Do you think that is a my mistake?

Thank you

[joe-c]

Using a 8 bit jpeg i have only 256 different gray value, if the delta T is very large the resolution decrease until 0.5°C. Do you think that is possible export data in 16 bit TIFF?

yes, 16bit tiff sounds better, i will deeper look on that next time.

And: with some dataset the scaled output is ok, but in the streached preview folder all of the exported image is the last image of the CSV folder and during conversion Thermovision give me an error in the bottom bar...but the scaled outpud is ok!! It's strange? Do you think that is a my mistake?

hmm... sounds like my mistake. sadly the images you have provided has very little differences, so i can't see a problem like that.
whats the error message in the bottom bar?
(you can make screenshots with doubble mouse click on "Screen" in the bottom left corner)

[tuffi]

In attach the screenshot (the image is the lastest of dataset) and the log file....the strange is I have error but I have also right scaled jpeg 

[joe-c]

Try the new 1.0.5.0 Version, it supports 16bit tif export and should now work correctly.

[tuffi]

Thank you so much!
Now i'am in Saudi Arabia for work.
As i came back in Italy i try the new version and give you me feedbeck..


Thanks

[Trancept]

Hello,
Thanks for all your informations, it's great !
I'm also using an Optris.

@tuffi : which software do you use for photogrammetry ?
AgiSoft Photoscan does not handle 16bit and Pix4D know only Flir file. I'm looking to Micmac but it's quite hard to use...

Well, I'm thinking a good option will be to convert file from Optris to Flir format to be able to use Pix4D, is that possible ?

[joe-c]

...Pix4D know only Flir file...

Pix4D knows *.jpg and *.tiff so the "flir format" with embedded radiometric info will not be supported i think.

if you have too a folder of CSV images, you can use the Thermovision Software (http://joe-c.de/pages/posts/programm_thermovision_joe-c_184.php), then select File->Autoselect and change to "Drop -> Optris Pi400 Mode".
then just drop the Folder with your CSV images. in Devices->Optris the "16bit .tif" is autoselected.

I think the main problem is to get the right geographical coordinates for the images... are they stored somewhere? 

[Trancept]

Thanks Joe-C,

Pix4D support radiometric information for Flir and Tau2, more information here : https://support.pix4d.com/hc/en-us/articles/210140783-Which-Thermal-Cameras-are-supported-by-Pix4Dmapper-#gsc.tab=0 and there : https://support.pix4d.com/hc/en-us/articles/208592186#label2

Geographical coordinates is not mandatory but it speed up the calculation. You could georeference the resulting map later.
If you have a dataset and want to try there is a step by step guide : https://support.pix4d.com/hc/en-us/articles/210140893-How-to-process-Thermal-Datasets-in-Pix4Dmapper#gsc.tab=0
EDIT : I saw you upload only 3 pictures, you need many more to make it work. Because picture have to overlap at more than 70% and cover every face of the object.

I will also give a try.

Tuffi tells me he is using Photoscan but I didn't know his workflow.
I've tested with Photoscan, who now support 16bit, and Optris images (in Tiff thanks to Tomas123 command), setting 25 ?m pixels size give some results but it does not seems to have an option to keep radiometric information.
I have to look to different pixels values to see if I can make reliable matching with temperature.

[joe-c]

Pix4D support radiometric information for Flir and Tau2, more information here...

sorry, i missed this info. but i don't tried much.

whats your goal?
a Panorama with usable thermal information, maybe a big measurement?
or just a thermal mapping of a field, ground, city or some other geographical?

[Trancept]

My goal is to map a field with UAV.

In fact, Pix4D need geographical coordinates to process thermal image. Maybe because it is too hard without them as the picture is small with few detail.

[gio123]

Hi; I know this thread is old, and there might not be replies to my question..but I'll try anyways 
@ Joe-c: how do you convert the 16 bit tiff generated by ThermoVision back to actual temperaures in degree C?
Thanks a lot: your software is excellent.

[Vipitis]

this thread is old - but I am attempting the same.

My issue is that I have a 80x60 sensor with an effective resolution of 160x120 using the Superresolution feature in the Thermal Camera+ app. With the most recent update the app supprots the option to save an image without the palette and the min/max number imprinted. But even when I use greyscale and lock the palette to a min and max temperature(if you set min and max to enable you get a blue/red overflow, that can be avoided by enabling the limits and then locking the palette and disabling the limits, and using gamma of .50 for best coverage) - VisualSFM can't find enough matches for a recontruction.

I might need to look for an really really easy example - with no reflection and no background - wait for a green for each capture and just hope for no noise. I can't take captures automatically yet(technically I can take a rubberband and force the burst shutter - but it is a bug right now - and auto tuning becomes an issue) but I would attempt aerial by just taping my phone under a drone and do a easy zigzag route.

[joe-c]

how do you convert the 16 bit tiff generated by ThermoVision back to actual temperaures in degree C?

i only know something about the way how FLIR dit it. I use this "Planck" calibration:
http://joe-c.de/pages/posts/planck_calibration_209.php

i guess it can be used for all types of thermal cameras, if they work between 7-14 µm, but i am not really sure.

@Vipitis
if you use a flir one to capture with a phone, you can use the "fone Ultimate" App in combination with Wifi-Direct to get a live image to a notebook. here you can capture images or make a fixed setup.
See here:
http://joe-c.de/pages/posts/waermebildkamera_flir_one_gen2_211.php

and if not, you can extract the raw 16bit png from stored *.jpg files. they can be used for panorama stitching... i don't  try it, but some other guys here did 

[Vipitis]

@joe-c

I use CAT S60 and Fone Ultimate does not support CAT S60... as I wrote in the answer to your pm - which somehow got lost on my end. I still haven't learned the .php and all the tricks you can do with radiometric jpgs. My best option right now is Thermal Camera+ with the most recent update - I can capture images with superresolution on medium and nearest Neighbor zooming to get 160x120 pixels of data presented in a 680x907 23bit .png which I convert into .jpg with imagemagick and then use them in VisualSFM for photogrammetry reconstruction. I am emailing George to output .pngs in 160x120 when I disable data burning and use no blurred upscaling.

[Vipitis]

So i am digging this thread up again. I don't want to start more threads and the information within here is still useful.

I want to attempt thermal photogrammetry reconstruction.

What I got is a tripod setup and turntable. The idea is to set minimum and maximum temperature in the Thermal Camera+ app and then lock the palette. Disable the min and max range afterwards do you end up with black and white instead of blue and red, which looks horrible with superresolution.

But I now need ideas for targets. Simply geometry, and high thermal dynamic range. Should be warmer then room temperature becuase my backdrop is just paper. Non reflective. Rather small, like max 150mmx150mmx150mm.

A cup of tee is an option, but is loses its temperature over time, so I need to be fast with my capture. As I plan to have 3 passes and turn the table by about 30° each step.

Soldering iron is too hot I think; It will just be blown out on my camera, and it's reflective.
I got a stage lamp that heats up, it's painted black so it's less reflective, maybe with some emissivity tweaks it can look good.

Please give me some suggestions. I mainly plan to take images with the custom render and a greyscale palette, but will also save radiometric jpegs and use Palette generator to set top and bottom ends and scale the histogram linear to export .pngs.

For software I will use VisualSFM and Autodesk ReCap Photo, cloud based. 

[MK_cruiser]

Hello every body..
thank you for expensive information.
I need a dataset of thermal Images to reconstruct a 3d image.
can any one suggest any thing.
best regards.
thank you.
 

[Vipitis]

Hey,

Are you looking for something specific? Like arial shots from a drone. Buildings or close up pcb.
The most difficult part is to have the data normalized, even for a static target.

[Hyper_Spectral]

FWIW I've been looking into this for the past few years as I extensively work with UAS and various payloads.

I have, for the most part, given up on the production of thermal orthomosaics with current thermal cores and Pix4D's current algorithms. You can make it work with a metric ton of overlap 90/90 or 90/80, an entirely flat elevation with unique features, and a ton of manual tie points. The result is still meh and needs touch-up but it sort of works. The efficiency and effectiveness of the whole process essentially nulls the point of it, for all use cases I've experienced anyways.

Once we get above 1 megapixel resolution cores I expect to see more viable use cases. There's a reason why you don't see beautiful thermal orthomosaics plastered all over the place, cost is only one factor.