(HRI/CRI) Building Damage Classification — File Specification
The building damage classification map attributes damage classification to individual buildings by overlaying building footprints onto the damage raster and aggregating impact at a structural level.
The file provides building centroids with a binary flag classifying the structure as damaged/undamaged, based on the number of raster pixels within the footprint that exceed the damage threshold. This file allows for a structure-by-structure assessment of damage.
To determine whether a building is considered damaged or not, ICEYE considers both the magnitude of change detected within a building's footprint (maximum damage intensity) and the total area of the footprint (above a fixed threshold) that has been affected. The exact thresholds used to determine damage is subject to regional variation, and are specified within the file.
To aid damage triage, ICEYE also provides a classification of the damage as low, medium or high, based on the fraction of a building footprint that exceeds the change detection threshold. Note that since the damage severity and damage fields are calculated using independent criteria, there may be minor discrepancies between these fields.
Individual structures can be cross-referenced with exposure files (via PID or G-NAF ID), allowing for seamless integration into workflows.
Attributes
Both
| Attribute | Description |
|---|---|
id |
ICEYE unique identifier for the building record within the analysis. |
damage |
Binary flag classifying the structure as damaged (1) or undamaged (0). |
damage_frac |
Fraction of pixels within the building footprint above the damage threshold. |
damage_threshold |
The intensity threshold used to determine whether a pixel is considered as damaged. |
max_val |
Maximum intensity detected within the building footprint. |
pix_dmg |
Number of pixels within the building footprint exceeding the damage threshold. |
pix_valid_total |
Total number of pixels contained within the building footprint. |
US only
| Attribute | Description |
|---|---|
vendor_building_footprint_id |
Persistent identifier for a building, mapping to BLDGID in the Precisely dataset. |
vendor_address_id |
Unique addressable location, mapping to Precisely's PBKEY. |
vendor_address_detail_id |
An address identifier mapping to Precisely's ADDR_ID field. |
damage_severity |
An estimate of damage severity based on the proportion of the building footprint exceeding the damage threshold: no_damage: <5% of pixels exceed damage threshold; low: ≥5-25%; medium: ≥25-75%; high: ≥75%. |
Australia only
| Attribute | Description |
|---|---|
address_detail_pid |
Persistent identifier for an address record in the Geocoded National Address File in Australia. |
address_pid |
Linked to address_detail_pid, the address persistent identifier that is related to the building feature. The persistent identifier is unique to the address feature this record relates to. |
Factors leading to false detections
As the heatmap is an estimate of change, it is likely that false positives unrelated to damage are included in the scene. These detections may not be relevant to the user attempting to judge damage, so they are good to be aware of.
- Highly angled roofs may produce spurious detections
- Small vertical roof structures (e.g. chimneys, A/C exhausts) may produce spurious detections
File Format: Geopackage (.gpkg)
CRS: EPSG:4326