Tutorial - Basics (with GUI)

Discover Main Features

The are 4 main features included in the application. These modules are not completely independent, using them sequentially is recommended for first usages.

Dataset Analysis (only with GUI): for analyzing multi- or hyper-spectral datasets. It includes vizualization of data slices, spectrum analysis, and dataset labeling.
Optical Design: for evaluating and comparing the performances of various optical systems.
Coded Aperture: for generating various patterns and corresponding filtering cubes.
Acquisition: for simulating the acquisition process of coded images

Feature A : Dataset Analysis

The Dataset analysis tab is used to load & display datasets characteristics.

1. Settings

Located on the left side of the application window.

Includes:

datasets directorypath to the datasets directory. All datasets be stored here.
ATTENTION : click on the reload datasets button if you change the datasets directory path
dataset name : a ComboBox displaying the datasets available in the selected directory
loaded dataset dimensionsThese values are displayed once the dataset is loaded
- dimension along X : dimension of the dataset in the X direction (main spectral dispersion direction)
- dimension along Y : dimension of the dataset in the Y direction (perpendicular to spectral dispersion direction)
- number of spectral bands : number of spectral bands in the loaded dataset
- minimum wavelength : minimum wavelength, usually corresponds to the spectral band n°0
- maximum wavelength : maximum wavelength, usually corresponds to the last spectral band

2. Load dataset button

By clicking on this button, the dataset selected in the dataset name ComboBox is loaded by the application.

3. Display windows

Located on the right side of the application window.

Once a dataset is loaded, one can inspect the spatial and spectral content of the dataset.

Hyperspectral cube

By moving the slider, you choose the spectral plane to be displayed.

Compare Spectra

Labelisation map

Labelisation Histogram

Feature B : Optical Design

The Optical Design tab is used for quick evaluation of the optical system characteristics (spectral dispersion & distortions).

1. System Settings

Located on the left side of the application window.

Includes:

infos:
- system name: name of the studied system
system architecture: All parameters that define the optical system and thus the spatial/spectral filtering
- system type
- propagation type : model used for evaluating the spatial/spectral filtering
- focal lens F [in micrometers]
- dispersive element:
  
  type: Prism or Grating
  
  A (only when prism is selected): apex angle of the prism [in degrees]
  
  m (only when grating is selected): considered order of diffraction [no units]
  
  G (only when grating is selected): grating lines density [lines/mm]
  
  delta alpha c [in degrees]
  
  delta beta c [in degrees]
  
  wavelength center [in nm]
detector: parameters that define the detector grid
SLM: parameters that define the mask grid
spectral range: the spectral boundaries of the system and the number of spectral bands to consider

2. Run Simulation button

For each considered wavelength, the mask grid points coordinates is propagated onto the detector.

3. Display

Located on the right side of the application window. It can be used to analyse the mask grid object and its images in the detector plane.

Coded aperture grid

Propagated coded aperture grid

Spectral images of the input coded aperture grid for the minimum, maximum, and center wavelength.

ATTENTION: center wavelength (605 nm on the given example) is different from the system architecture center wavelength

Distortion maps

Get qualitative and quantitative distortion data:

Feature C : Pattern generation

The Coded Aperture tab is used for designing patterns and generating associated filtering cube.

1. Patterns Settings

Located on the left side of the application window.

The patterns characteristics depend on the chosen pattern type.

Available patterns:

slit:only one column of the coded aperture is open (perpendicular to the spectral dispersion), thus generating a spectral gradient type filter.
- slit position: relative to the center column between -100 and 100 coded aperture elements
- slit width: between 1 and 30 coded aperture elements.
random: random noise pattern with a normal law
blue noise: random noise pattern with boosted high frequencies
custon h5 pattern: custom pattern that should be a h5 file with a container named “pattern”. Once loaded, the pattern is cropped to fit SLM dimensions

2. Generate pattern

By clicking on this button, a 2D array representing a coded aperture pattern is generated through pattern generation functions contained in the functions_patterns_generation.py file.

3. Generate Filtering Cube button

By clicking on this button, a CassiSystem instance is creating the filtering cube corresponding to the detector dimensions along X and Y and the number of spectral bands.

Each slice of the filtering contains the projection of the coded aperture pattern on the detector grid.

ATTENTION : The spectral sampling of the filtering cube is not the same as the dataset’s sampling. It is defined in the spectral range section of the Optical Design tab. The wavelengths are equally spaced between “minimum wavelength” and “maximum wavelength”.

4. Display Pattern and Filtering Cube

Located on the right side of the application window.

Pattern

Shows the generated (or loaded) pattern:

Filtering Cube, slice by slice

Shows the corresponding filtering cube. By moving the slider, one can inspect the filtering cube slice by slice:

Feature D : Acquisition

The Acquisition tab is used to generate compressed measurements given: a dataset and a filtering cube.

Note that the dataset is:

cropped in the spatial dimensions to fit the filtering cube sampling (detector dimensions).
interpolated in the spectral dimension according to the filtering cube sampling.

1. Settings

For now, the GUI only includes one mode: single acquisition.

A Point-spread-function (PSF) can be added for more realism. For now, each slice of the filtered scene is convolved by the same kernel. A wavelength-dependent PSF will be added in the future.

2. Run Acquisition button

By clicking on this button:

First, the dataset cube is cropped in the spatial dimensions and interpolated in the spectral dimension.
Second, a point by point multiplication is performed between the filtering cube and the reinterpolated scene.

3. Display measurements

compressed measurements

The image as measured by the detector.

Spectral images

Each slice of the filtered scene.

Panchromatic image

No spatial/spectral filtering, the interpolated scene is simply summed along its spectral dimension.