[HN Gopher] Show HN: A large format XY scanning hyperspectral ca...
___________________________________________________________________
Show HN: A large format XY scanning hyperspectral camera
Author : anfractuosity
Score : 31 points
Date : 2025-10-11 07:30 UTC (6 days ago)
(HTM) web link (www.anfractuosity.com)
(TXT) w3m dump (www.anfractuosity.com)
| phubbard wrote:
| The scan rate really limits the usability of this. Do you have
| any leads on planar sensors?
| anfractuosity wrote:
| 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.
| DarkSucker wrote:
| 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).
| bflesch wrote:
| 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 50EUR , so only
| 2.000.000EUR to test this
| anfractuosity wrote:
| 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
| bflesch wrote:
| 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.
| anfractuosity wrote:
| Yeah definitely, I've not looked at papers relating to
| microscopy with hyperspectral cameras. That might be
| especially interesting with samples that fluoresce.
| bflesch wrote:
| 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.
| juancn wrote:
| 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).
| westurner wrote:
| 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.
___________________________________________________________________
(page generated 2025-10-17 23:01 UTC)