SWIR Images - InGaAs Linescan Camera

[Fandyus]

Very interesting, thanks for posting.

[Andy Perrin]

I see you got the portraits working! Did you end up doing what we discussed and starting the camera moving, pose for pic, then crop the edges after?

 

The portraits came out very nicely too.

[Daniel Csati]

40 minutes ago, Andy Perrin said:

I see you got the portraits working! Did you end up doing what we discussed and starting the camera moving, pose for pic, then crop the edges after?

 

The portraits came out very nicely too.

I got some help but I had to crop the edges anyway :)

[Andy Perrin]

You might want to invest in the Topaz Gigapixel and Neat Image denoising plugins. Between those, you can upscale by a factor of two.

 

Here's how the portrait looks with a combo of those plugins:

 

 

compared to the original:

[dabateman]

Very nice.

Looks like no wayer in your hair, thus nearly white. 

You could try spraying your head with a water bottle or adding drops of water to your hair to get black spots. 

[Daniel Csati]

29 minutes ago, Andy Perrin said:

You might want to invest in the Topaz Gigapixel and Neat Image denoising plugins. Between those, you can upscale by a factor of two.

I heard of the Gigapixel before and the denoising did a great improvement too. I haven't tried to do any changes to the images yet, I wonder how free alternatives (for instance IrfanView tools) compare

[Andy Perrin]

Daniel, I don’t think anything beats that particular combo. Gigapixel is the best upscaling I have ever seen, and Neat Image gives the most control over the denoising. You won’t find a free alternative with the same quality. 

[Daniel Csati]

I noticed today an another interesting thing.

 

Looking at the first image I uploaded here, the reflection of the sun is visible on the cars:

Some of the pixels are obviously saturated and there is some blooming on the sensor or ghosts from the lens in vertical direction. However, the most intense areas seem to have a black(!) spot in the middle. Has anyone seen this effect before? 

[Andy Perrin]

That looks like overflow error maybe. Sometimes on computers if you go over yhe size of your datatype it wraps back around to the other end. 

[Stefano]

Very SWIR-style portraits, I like them.

 

Is there a way to fix the overflow error by setting it to be 100% white?

[Andy Perrin]

I mean, it would be pretty easy to select them and just make them white in Photoshop or similar programs. 

[Daniel Csati]

Thanks for the tip. I did yesterday some experiments with the bulb and during the focus/exposure setup I had the bulb with full power and high exposure time on the image. No such effect was visible.. But it definitely looks like an overflow error. This calls for more experiments! Until then here is an another landscape photo, this time with the 1575nm bandpass filter: 

 

Visible for reference (far right part of the SWIR panorama): 

[Fandyus]

Interesting how the leaves are suddenly dark again. I suspect it's due to their water content?

[Stefano]

I found this: https://www.google.com/imgres?imgurl=https%3A%2F%2Fwww.researchgate.net%2Fprofile%2FXiaojin-Qian%2Fpublication%2F331068558%2Ffigure%2Ffig1%2FAS%3A736556390301696%401552620136054%2FReflectance-spectra-for-three-types-of-leaves-with-minimum-and-maximum-chl-a-b-content.png&imgrefurl=https%3A%2F%2Fwww.researchgate.net%2Ffigure%2FReflectance-spectra-for-three-types-of-leaves-with-minimum-and-maximum-chl-a-b-content_fig1_331068558&tbnid=mE7otbbqpT9iLM&vet=12ahUKEwj2tpqizJL5AhWMh_0HHS-SCJUQMygAegUIARCrAQ..i&docid=bRdfXj7O22nW7M&w=850&h=564&q=foliage reflection spectrum&client=ms-android-samsung-ss&ved=2ahUKEwj2tpqizJL5AhWMh_0HHS-SCJUQMygAegUIARCrAQ

 

the notches do line up with water's 1450 nm and 1950 nm peaks.

[Fandyus]

1 hour ago, Stefano said:

I found this: https://www.google.com/imgres?imgurl=https%3A%2F%2Fwww.researchgate.net%2Fprofile%2FXiaojin-Qian%2Fpublication%2F331068558%2Ffigure%2Ffig1%2FAS%3A736556390301696%401552620136054%2FReflectance-spectra-for-three-types-of-leaves-with-minimum-and-maximum-chl-a-b-content.png&imgrefurl=https%3A%2F%2Fwww.researchgate.net%2Ffigure%2FReflectance-spectra-for-three-types-of-leaves-with-minimum-and-maximum-chl-a-b-content_fig1_331068558&tbnid=mE7otbbqpT9iLM&vet=12ahUKEwj2tpqizJL5AhWMh_0HHS-SCJUQMygAegUIARCrAQ..i&docid=bRdfXj7O22nW7M&w=850&h=564&q=foliage reflection spectrum&client=ms-android-samsung-ss&ved=2ahUKEwj2tpqizJL5AhWMh_0HHS-SCJUQMygAegUIARCrAQ

 

the notches do line up with water's 1450 nm and 1950 nm peaks.

Interesting that it's still technically brighter than the trees' reflectance of green light.

Edit: Also fascinating how the graph slopes up towards 200nm. Does that mean leaves get brighter again? Maybe dabateman has something on that.

[dabateman]

Yes leaves are reflecting in UVC. So much more than flowers. Thats how I used to find the location of my spot mercury bulb.

[Andy Perrin]

I would be very very cautious about believing in data from close to either end (UV-C or near IR) of a spectrum based on everything I’ve ever seen presented on this forum. Those are the spots where the data is least reliable unless counter-measures have been taken. David’s experiences notwithstanding (I mean, that just tells you they are bright compared to the flowers.)

[Daniel Csati]

I had not much time recently to play with my camera, but here is a recent photo. Boron doped Si wafer being almost completely transparent at 1560-1580nm. Light source: Lucky Reptile 35W halogen, Exposure time 52ms. Not sure if the boron makes any difference.

I played a bit with pyrite too, but it wasn't transparent at all. I expected it to be.

 

This is the spectrum of my light source from 200nm to 1100nm. I don't expect it to have too much light around 1500. 

 

[colinbm]

Pretty spooky, The heatsink looks like it is in front not behind ?

[Daniel Csati]

It does look like it, a bit. And the visible photo has a reflection cascade, a lucky accident :D

[dabateman]

That spectrum of your hallogen light isn't acurate.

Was it recorded with a usb spectrometer,  that uses a silicon sensor,  which isn't sensitive to muc light above 1000nm?

See this for the Hallogen spectrum at various temperatures:

https://zeiss-campus.magnet.fsu.edu/articles/lightsources/tungstenhalogen.html

 

You will have lots of light at 1500nm.

 

This is an example of needing to know the limits of your equipment.  If the Hallogen was correctly measured, it wouldn't be dropping after 700nm, but would be increasing.  The sensitivity of your spectrometer is what you have plotted here, not the spectrum of the Hallogen light.

[Stefano]

Nice. You could calibrate your images on a uniform background to get rid of the striping.

 

Try some water next. Or a camera sensor, which should be transparent. There's lots of things to image in SWIR.

[Andy Perrin]

David is correct, halogens have loads of light at 1500nm. I actually use one as my preferred light source in SWIR. 

[Daniel Csati]

You are probably right about the spectrometer, it's a Si sensor so max 5% QE above 1000nm. I can feel the bulb being hot so it should have a lot of light at 1500nm. It's just still much less than with scattered sunlight - outdoor I can do images usually with 4-5 times shorter exposure times. 

 

I know I should make some effort because of the strips.  It's much worst if I increase the gain or go with the exposure time further up.  It would be either a Matlab code or just using the noise filter App from Andy's recommendation. Maybe cooling the sensor more would help it too, but I can't find any commands to change the TEC setpoint.