DATASET_CARD.md

Dataset Card: Heat-Assisted Detection and Ranging (HADAR) Inputs

Overview

This bundle supports ground-based long-wave infrared (LWIR) hyperspectral imaging for heat-assisted detection and ranging. It combines five scenes from the DARPA IH Dataset with per-scene atmospheric products: a zenith (vertically downward) downwelling spectral radiance and a matching spectral attenuation coefficient.

The atmospheric files are not a shared U.S. Standard Atmosphere (USSA) snapshot. For each capture, attenuation and zenith downwelling were computed with libRadtran (uvspec, DISORT, thermal emission) from a layered temperature and gas-density profile at the nearest analysis time and nearest grid location to that scene, plus high-resolution HITRAN line-by-line optical depths (H₂O, CO₂, O₃, NO₂). Both products use the same 6–14 µm grid at 4 nm spacing.

The ENVI cubes are too large for GitHub. The image build fetches this same layout from the frozen Hugging Face dataset dccc2025/tmp_dataset_hadar and checks SHA256. That fetch is pinned to a commit (not a moving main branch) and is not a live scrape of Kitware.

Directory Layout

inputs/
├── DATASET_CARD.md
├── attenuation/
│   ├── IHTest_202104_Path15_Step2_LWHSI1_collect0_DistStA.npz
│   ├── IHTest_202104_Path17_Step1_LWHSI1_collect0_DistStA.npz
│   ├── IHTest_202104_Path17_Step3_LWHSI1_collect0_DistStA.npz
│   ├── IHTest_202104_Path26_Step2_LWHSI1_collect0_DistStA.npz
│   └── IHTest_202104_Path27_Step25_LWHSI1_collect0_DistStA.npz
├── downwelling_radiance/
│   ├── IHTest_202104_Path15_Step2_LWHSI1_collect0_DistStA.npz
│   ├── IHTest_202104_Path17_Step1_LWHSI1_collect0_DistStA.npz
│   ├── IHTest_202104_Path17_Step3_LWHSI1_collect0_DistStA.npz
│   ├── IHTest_202104_Path26_Step2_LWHSI1_collect0_DistStA.npz
│   └── IHTest_202104_Path27_Step25_LWHSI1_collect0_DistStA.npz
└── original_datasets/
    ├── IHTest_202104_Path15_Step2_LWHSI1_collect0_DistStA.{bsq,hdr}
    ├── IHTest_202104_Path17_Step1_LWHSI1_collect0_DistStA.{bsq,hdr}
    ├── IHTest_202104_Path17_Step3_LWHSI1_collect0_DistStA.{bsq,hdr}
    ├── IHTest_202104_Path26_Step2_LWHSI1_collect0_DistStA.{bsq,hdr}
    └── IHTest_202104_Path27_Step25_LWHSI1_collect0_DistStA.{bsq,hdr}

Each scene stem has one attenuation file and one downwelling file with the same name. Pair them by stem; do not mix products across scenes.

1. Hyperspectral Scenes (original_datasets/)

Five ENVI-format LWIR hyperspectral cubes (.bsq + .hdr) from the DARPA IH Dataset. Each pair is a band-sequential cube plus acquisition metadata. Headers record wavelength (µm) and a near-surface air temperature (kelvin). They do not record a radiance scale factor.

Scene stem	Acquisition time (UTC)	Site	Lines	Samples	Bands
IHTest_202104_Path15_Step2_LWHSI1_collect0_DistStA	2021-04-16T17:53:52	Fort A.P. Hill, VA	260	1600	256
IHTest_202104_Path17_Step1_LWHSI1_collect0_DistStA	2021-04-16T21:43:22	Fort A.P. Hill, VA	260	1600	256
IHTest_202104_Path17_Step3_LWHSI1_collect0_DistStA	2021-04-16T21:52:54	Fort A.P. Hill, VA	260	1600	256
IHTest_202104_Path26_Step2_LWHSI1_collect0_DistStA	2021-04-20T23:13:02	Fort A.P. Hill, VA	260	1600	256
IHTest_202104_Path27_Step25_LWHSI1_collect0_DistStA	2021-04-21T03:24:47	Fort A.P. Hill, VA	260	1600	256

• Format: ENVI BSQ, 32-bit float (data type = 4).
• Wavelengths: Stored in each .hdr file; units are micrometers (µm). All five cubes are offset by 120 nm toward longer wavelengths in the header calibration. Subtract 0.12 µm from every listed band center before use. Example: a header range of 8–13.17 µm corresponds to a corrected range of 7.88–13.05 µm. Apply this shift when aligning cubes to Planck functions or to the atmosphere files.
• Distribution: Distribution Statement A (see scene headers).

Cube radiance units

The .bsq arrays are not already in (\mathrm{W,m^{-2},sr^{-1},\mu m^{-1}}). They are stored a factor of (10^{2}) high relative to physical spectral radiance. Convert with

[ L_{\lambda};[\mathrm{W,m^{-2},sr^{-1},\mu m^{-1}}] = 10^{-2},L_{\mathrm{bsq}}. ]

Use this scaled (L_{\lambda}) in any comparison to Planck (B_\lambda(T)) or to the bundled zenith downwelling (which is already in physical units). Do not treat header temperature as the surface temperature map; it is local air temperature.

Source Citation

@InProceedings{Yellin_2024_WACV,
    author    = {Yellin, Florence and McCloskey, Scott and Hill, Cole and Smith, Eric and Clipp, Brian},
    title     = {Concurrent Band Selection and Traversability Estimation from Long-Wave Hyperspectral Imagery in Off-Road Settings},
    booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
    month     = {January},
    year      = {2024}
}

Dataset homepage: https://www.kitware.com/ihdataset/

2. How the atmosphere files were computed

For each scene, a vertical profile of temperature and molecular number density was taken at the nearest available analysis time and the nearest grid point to the capture site (Fort A.P. Hill, Virginia). Layer optical depths on 6–14 µm were built from HITRAN high-resolution line parameters for H₂O, CO₂, O₃, and NO₂. libRadtran then produced:

1. Spectral attenuation of the lowest (near-surface) layer, stored as α in dB m⁻¹.
2. Zenith downwelling spectral radiance only: DISORT thermal calculation, surface observer, viewing cosine umu = −1 (zenith angle 0°). There is no multi-angle downwelling basis.

Spectral grid (both products, all scenes): 6.000–14.000 µm, Δλ = 4 nm, 2001 samples. This grid is not the camera wavelength axis in the .hdr files.

3. Spectral Attenuation (attenuation/<scene_stem>.npz)

One file per scene. Use the file whose stem matches the cube.

Key	Shape	Description
attenuation	(2001, 1)	Spectral attenuation coefficient α, in dB m⁻¹
lambda	(1, 2001)	Wavelength grid, 6–14 µm at 4 nm

Approximate α ranges on this 4 nm grid (water-band peaks near 6–8 µm are much larger than the LWIR window):

Scene stem	α min (dB m⁻¹)	α max (dB m⁻¹)
Path15 Step2 LWHSI1	3.02×10⁻⁴	44.9
Path17 Step1 LWHSI1	3.49×10⁻⁴	50.3
Path17 Step3 LWHSI1	3.49×10⁻⁴	50.3
Path26 Step2 LWHSI1	5.59×10⁻⁴	71.0
Path27 Step25 LWHSI1	5.22×10⁻⁴	66.5

Load example:

import numpy as np

stem = "IHTest_202104_Path26_Step2_LWHSI1_collect0_DistStA"
data = np.load(f"attenuation/{stem}.npz")
alpha = data["attenuation"]   # (2001, 1), dB/m
lam_att = data["lambda"]      # (1, 2001), µm

Horizontal-path transmittance at range (d) (metres) is (\tau(d)=10^{-\alpha d/10}). Resample alpha onto the corrected scene wavelength grid before applying this formula.

4. Zenith Downwelling Radiance (downwelling_radiance/<scene_stem>.npz)

One file per scene, already in physical radiance units. A single column: zenith only (vertical downward / zenith angle 0°). This is not an angle-resolved sky basis.

Key	Shape	Description
lambda	(1, 2001)	Same 6–14 µm, 4 nm grid as the matching attenuation file
downwelling_res	(2001, 1)	Zenith downwelling spectral radiance, W m⁻² sr⁻¹ µm⁻¹
zenith_angle_deg	(1,)	[0]

Load example:

import numpy as np

stem = "IHTest_202104_Path26_Step2_LWHSI1_collect0_DistStA"
data = np.load(f"downwelling_radiance/{stem}.npz")
lam = data["lambda"]                    # (1, 2001), µm
I_down = data["downwelling_res"]        # (2001, 1), W·m⁻²·sr⁻¹·µm⁻¹
zenith = data["zenith_angle_deg"]       # [0]

Units and Conventions

Quantity	Symbol / key	Units
Wavelength	lambda	µm
Attenuation coefficient	attenuation (α)	dB m⁻¹
Zenith downwelling spectral radiance	downwelling_res	W m⁻² sr⁻¹ µm⁻¹
Hyperspectral cube	ENVI .bsq	stored (10^{2}) high; (L=10^{-2}L_{\mathrm{bsq}}) in W m⁻² sr⁻¹ µm⁻¹
Header temperature	air temperature	K

Notes for Use

1. Header wavelength offset. Every cube’s .hdr wavelengths are 120 nm too long. Correct with (\lambda_{\mathrm{true}}=\lambda_{\mathrm{hdr}}-0.12,\mu\mathrm{m}) before any spectral work. If a header lists 8–13.17 µm, the physical band is 7.88–13.05 µm.
2. Wavelength alignment. After that 120 nm correction, all five LWHSI1 cubes have 256 bands. The atmosphere lambda grid is 6–14 µm at 4 nm and does not match the camera bands. Interpolate or band-select attenuation and downwelling onto the corrected scene wavelength grid before combining them with a cube.
3. Paired products. Each scene has its own matching attenuation and zenith-downwelling files, computed from the nearest analysis time and nearest grid location to that capture. Load both .npz files from the same scene stem; they share one lambda axis. This per-scene pairing is more accurate than a shared standard atmosphere.
4. Zenith only. downwelling_res has shape (2001, 1). There are no additional zenith-angle columns. This vector is not the delivered single-band texture map X.