Project 1: GUI for Pushbroom Hyperspectral Imaging Demonstrator

Introduction and Motivation

Project introduction on UTAT SS Optics x Data Processing on personal Notion page.

This is largely inspired by the Hinalea App, which I demoed to you earlier this year.

Detailed Description of GUI functionality and features

Qualitative requirements for a successful product:

desktop application that runs on Windows 10/11 machines
GUI that is clear and functional. It does not need to be “professional”, but good-looking enough that we can “wow” people when doing a live demo.
Clear & navigable file folder storage structure with thumbnail and metadata included with each datacube
runtime speed that is not necessarily instantaneous, but still acceptable for a live demo (i.e., the whole processing & display pipeline should not take more than 1 minute of waiting).

Data acquisition pipeline:

establish connection to two cameras (Basler & Thorlabs), verify connection is active w/ “indicator LED” on the GUI
Data-cube acquisition: prompt user through a sequence of frame captures (instruction to move target by 5mm, then click button to capture image)
Save data to a folder immediately on completion of capture. Prompt user to enter meta data, save together.
Stitch data from thorlabs hyperspectral linescan camera into datacube.
- option to perform de-striping, etc.
- will use pre-loaded wavelength-to-pixel-column mapping provided from optics subsystem
- optics will also provide an approximate distance-to-pixel-row mapping
Stitch data from basler 2D RGB camera and convert into quasi-georeferenced data
- GIVEN there is (e.g.) a 5mm distance between adjacent snapshots PLUS taking an image of a calibration target with known dimensions WITH THE BASLER CAMERA, (as well as any supplemental information, such as focal length of the lens, and the distance between the lens and target, size of pixels, etc)
- determine the distance-to-pixel mapping for the BASLER camera (this is like your geospatial dataset, can be done a-priori for the imaging target).
Perform data fusion between basler 2D camera and thorlabs hyperspectral linescan camera
- Examine the hyperspectral dataset and (re)stitch it by recognizing features in the Basler camera image, to produce a datacube with calibrated wavelength (in um) on the spectral axis and a real calibrated dimension (in centimeters) for the spatial axes
- Can be classical and/or ML enhanced, etc. - whichever algorithm you plan on using for the FINCH dataset

Processing Data: