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](https://www.kitware.com/ihdataset/) with a shared U.S. Standard Atmosphere (USSA) spectral pair: horizontal-path attenuation coefficients and a ten-angle downwelling-radiance basis.

The atmospheric products are a frozen copy of the `data/` directory from [Ozone-Cues (LWIR absorption-based ranging)](https://github.com/unaydorken/Ozone-Cues-Mitigate-Reflected-Downwelling-Radiance-in-LWIR-Absorption-Based-Ranging/tree/main/data). All five IH captures are daylight, near-surface scenes. For clear-sky conditions with similar air temperatures, recent work on this problem finds that site-to-site differences in downwelling spectral features are small compared with other inversion errors, so the five cubes share one USSA atmosphere rather than a per-scene sounding.

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](https://huggingface.co/datasets/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

```text
inputs/
├── DATASET_CARD.md
├── attenuation.npz
├── I_downwelling_res.npz
└── original_datasets/
    ├── IHTest_202104_Path26_Step2_LWHSI1_collect0_DistStA.{bsq,hdr}
    ├── IHTest_202104_Path27_Step25_LWHSI1_collect0_DistStA.{bsq,hdr}
    ├── IHTest_202204_Path2_Step1_LWHSI2_DistStA.{bsq,hdr}
    ├── IHTest_202204_Path13_Step9_LWHSI2_DistStA.{bsq,hdr}
    └── IHTest_202204_Path13_Step13_LWHSI2_DistStA.{bsq,hdr}
```

## 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) | Lines | Samples | Bands |
|------------|------------------------|------:|--------:|------:|
| `IHTest_202104_Path26_Step2_LWHSI1_collect0_DistStA` | 2021-04-20T23:13:02 | 260 | 1600 | 256 |
| `IHTest_202104_Path27_Step25_LWHSI1_collect0_DistStA` | 2021-04-21T03:24:47 | 260 | 1600 | 256 |
| `IHTest_202204_Path2_Step1_LWHSI2_DistStA` | 2022-04-20T21:14:29 | 480 | 1500 | 250 |
| `IHTest_202204_Path13_Step9_LWHSI2_DistStA` | 2022-04-22T23:00:35 | 480 | 1500 | 250 |
| `IHTest_202204_Path13_Step13_LWHSI2_DistStA` | 2022-04-22T23:09:03 | 480 | 1500 | 250 |

- **Format:** ENVI BSQ, 32-bit float (`data type = 4`).
- **Wavelengths:** Stored in each `.hdr` file; units are micrometers (µm).
- **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 downwelling basis (which is already in physical units).
Do not treat header `temperature` as the surface temperature map; it is
local air temperature.

### Source Citation

```bibtex
@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. Spectral Attenuation (`attenuation.npz`)

Frozen USSA horizontal-path spectral attenuation, copied from the Ozone-Cues `data/` snapshot above. Not recomputed per scene.

| Key | Shape | Description |
|-----|-------|-------------|
| `attenuation` | `(256, 1)` | Spectral attenuation coefficient α, in **dB m⁻¹** |
| `lambda` | `(1, 256)` | Wavelength grid paired with `attenuation`, in **µm** |

**Value range (α):** approximately 2.8×10⁻⁴ to 1.4×10⁻² dB m⁻¹ over 7.98–13.05 µm.

Load example:

```python
import numpy as np

data = np.load("attenuation.npz")
alpha = data["attenuation"]   # (256, 1), dB/m
lam_att = data["lambda"]      # (1, 256), µm
```

## 3. Downwelling-Radiance Basis (`I_downwelling_res.npz`)

Ten resampled USSA downwelling spectral radiance profiles, one per zenith angle, from the same Ozone-Cues `data/` snapshot. These columns are a fixed angular basis for reflected hemispherical downwelling. They are already in physical radiance units.

| Key | Shape | Description |
|-----|-------|-------------|
| `lambda` | `(1, 256)` | Wavelength grid, in **µm** (approximately 8.10–13.17 µm) |
| `downwelling_res` | `(256, 10)` | Downwelling spectral radiance at each angle, in **W m⁻² sr⁻¹ µm⁻¹** |

**Zenith angles (column order, degrees):**

`0, 30, 60, 70, 80, 82, 84, 86, 88, 89`

Column `k` of `downwelling_res` corresponds to zenith angle `k` in the list above.

Load example:

```python
import numpy as np

data = np.load("I_downwelling_res.npz")
lam = data["lambda"]                    # (1, 256), µm
I_down = data["downwelling_res"]        # (256, 10), W·m⁻²·sr⁻¹·µm⁻¹
```

## Units and Conventions

| Quantity | Symbol / key | Units |
|----------|--------------|-------|
| Wavelength | `lambda` | µm |
| Attenuation coefficient | `attenuation` (α) | dB m⁻¹ |
| 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. **Wavelength alignment.** LWHSI1 cubes have 256 bands starting near 8.06 µm. LWHSI2 cubes have 250 bands starting near 6.84 µm. The downwelling `lambda` grid is a 256-channel LWIR window and does **not** automatically match every scene. Interpolate or band-select every spectral product onto the scene wavelength grid before combining them.
2. **Separate attenuation grid.** `attenuation.npz` has its own `lambda` axis, offset from the downwelling grid. Resample `attenuation` onto the scene grid before using \(\tau(d)=10^{-\alpha d/10}\).
3. **Shared atmosphere.** One USSA attenuation cube and one ten-angle downwelling basis are used for all five scenes. This is a documented approximation for clear-sky, similar-temperature collections, not a claim that the real soundings were identical.
4. **Downwelling basis.** Treat the ten columns of `downwelling_res` as a fixed, ordered angular basis; do not permute them. They are not the delivered single-band texture map `X`.
5. **Scene selection.** `IHTest_202104_Path27_Step25_LWHSI1_collect0_DistStA` replaces the previous 202108 Path9 cube. Path9 had too little thermal-texture and range contrast (nearly flat `X` and `d`) to invert stably. Path27 is the same LWHSI1 camera family (260×1600×256) with more usable geometric texture and ranging contrast. This Path27 header does not include an air `temperature` field; use the solver fallback (default 290 K) unless you set path temperature another way.