Show HN: A large format XY scanning hyperspectral camera | Hacker News Reader

Show HN: A large format XY scanning hyperspectral camera

anfractuosity.com

44 pointsanfractuosity10 comments

Well done, thanks for sharing.

May I ask a stupid question? In order to speed up this process you'd need to buy more of the broadcom spectrometers which are quite expensive.

So instead of buying more spectrometers, is there a way to use multiple fibre cables but make them different length so the single spectrometer takes longer to process the image, but the "exposure" of all fibre cables would be at the same time. Would it be possible to stack the signal in a way that the light of the first fibre cable arrives at the spectrometre, then there is a short pause, then the light of the second fibre cable arrives at the spectrometer, etc.

I imagine it like plumbing if you have two toilets attached to the same canalisation, and you flush both toilets at the same time but the length of each toilet's pipe to the main canalization pipe is done in a way that the water of the first toilet has fully passed before the water of the second toilet arrives.

To put it into numbers: toilet1 pipe length would be 1m and toilet2 pipe length would be 10m because before both pipes join and go into the canalisation. Water travels at a certain speed in the pipe, which would make this possible.

Can the same principle be used for fibre cables? If yes, it should be possible to construct a 100x100 fibre cable sensor matrix using large fibre cable lengths but only a single spectrometer.

Edit: I did some math and it seems the second fibre cable needs to be 40.000km long so that the singal of the first fibre cable reaches the sensor after the signal of the second fibre cable has passed the sensor (at 0.2s).

  speed of light      
  in vacuum 299.792.458,00 m/s    
  in fibre 66,00 % percent    
   197.863.022,28 m/s

  exposure per pixel 0,2 s

  exposure start time   distance traveled   
  fiber 1 0 s 0   
  fiber 2  0,2 s 39.572.604,46 meters 39.572,60 kilometers

Edit2: You can buy 1km fibre cable for 50€ , so only 2.000.000€ to test this

The much faster way is to move a slit in front of a diffraction grating and use a 2D sensor behind that, that way you can obtain multiple spectral patterns simultaneously, see the video I mention at the start - https://www.youtube.com/watch?v=L_u8NqmgElU

Thanks, sorry for asking if it was answered already. Would be very interesting to do some microscopy with this approach, as the video mentions medical use cases. There's been a lot of research on this I guess.

I wonder if hyperspectral cameras plus some ML could help discerning objects in microscopy without needing to stain them.

Yeah definitely, I've not looked at papers relating to microscopy with hyperspectral cameras. That might be especially interesting with samples that fluoresce.

I know there are several companies who have done object detection for microscopy images with - I assume - traditional rgb images. I wonder if someone trained their ML model with the spectral images yet.

Would be nice to detect parasites or parasite eggs just with hyperspectral cam without staining. Important step towards the non-chemical toilet-which-inspects-feces-for-parasites I've been talking about to my friends for more than ten years ;)

Most of fluorescence microscopy works by staining/modifying the sample organism with a fluorescent before examining it. Theoretically (tm) it should work just by using hyperspectral imaging but that might be nobel prize territory.

Super cool, have you considered using galvanometers rather than moving the fiber around?

It should allow for faster scan rate (assuming the XY is the limiting factor).

The scan rate really limits the usability of this. Do you have any leads on planar sensors?

A linear scanning mechanism would be a lot faster still than the XY approach.

I'm very curious how expensive sensors are though, that capture this data in a single shot, I've not seen any prices so far.

Each pixel of such a scanner would need to somehow scan the spectral content. For example, imagine an array of fibers each transporting light from image plane pixel coupled to a spectrometer (bulky, expensive). Slit (a.k.a. push broom) scanners take each pixel of its slit and disperses the light perpendicular to the slit onto a 2D sensor array (more compact, 1D mechanical scan required). I recall seeing spectral (color) filters made from dispersive materials sandwiched between rotating polarizers to filter (scan) the light entering camera (expensive, compact).

Applications for:

"Multispectral imaging through scattering media and around corners via spectral component separation" (2024) https://opg.optica.org/oe/fulltext.cfm?uri=oe-32-27-48786&id... .. https://news.ycombinator.com/item?id=42557904

"Multi-sensor characterization for an improved identification of polymers in WEEE recycling" (2024) [WEE: e-waste] https://news.ycombinator.com/item?id=42534637 ..

"Reversible optical data storage below the diffraction limit" (2023) https://news.ycombinator.com/item?id=42331986 :

>> [...] This is possible by multiplexing the storage in the spectral domain.

"Tongue Disease Prediction Based on Machine Learning Algorithms" (2024) https://www.mdpi.com/2227-7080/12/7/97 :

>> This study proposes a new imaging system to analyze and extract tongue color features at different color saturations and under different light conditions from five color space models (RGB, YcbCr, HSV, LAB, and YIQ).

"A self-healing multispectral transparent adhesive peptide glass" https://www.nature.com/articles/s41586-024-07408-x :

> Moreover, the supramolecular glass is an extremely strong adhesive yet it is transparent in a wide spectral range from visible to mid-infrared. This exceptional set of characteristics is observed in a simple bioorganic peptide glass composed of natural amino acids, presenting a multi-functional material

Further study:

/? hyperspectral, specra-: https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...

Related applications:

Was looking into phase imaging and the four Stokes parameters S0,S1,S2,S3 and so on in assessing applications for the Parallel Axis Theorem: 2x2 "superpixel" imaging can capture the ((-45, 45), (L Circuluar, R Circular)) polarization information necessary to infer phase (and non-quantum optical entanglement is polarization)

Is polarimetric imaging (per-pixel polarization information) hyperspectral or hyperspectropolarimetrical?

spectropolarimetry: https://www.google.com/search?client=firefox-b-1-d&q=spectro...

Spectropolarimetry -> Polarimetry: https://en.wikipedia.org/wiki/Polarimetry

For example a division-of-focal-plane (DoFP) camera / image sensor has 2x2 pixels.

Reply on news.ycombinator.com