Multi-resolution Photovoltaic Panel Segmentation from Satellite and Aerial Imagery

A photovoltaic dataset from satellite and aerial imagery includes three groups of PV samples collected at spatial resolutions of 0.8m, 0.3m, and 0.1m. The dataset, authored by Jiang Hou of the Chinese Academy of Sciences, categorizes ground and rooftop samples by background land use and roof type. Data documentation is referenced in a preprint on the Copernicus ESSD platform.

Use Cases

• Train semantic segmentation models for photovoltaic panels based on multi-resolution imagery.
• Analyze the distribution of ground-mounted PV installations across different land use types like shrub land and cropland.
• Develop algorithms to distinguish rooftop PV panels on various roof materials such as flat concrete and steel tile.
• Benchmark model performance across different spatial resolutions from satellite and UAV sources.

Strengths

• Includes imagery at three distinct spatial resolutions: 0.8m, 0.3m, and 0.1m.
• Ground samples are categorized into five specific land use types: shrub land, grassland, cropland, saline-alkali, and water surface.
• Rooftop samples are categorized into three specific roof types: flat concrete, steel tile, and brick.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and dataset size are unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

paperswithcode, with data referenced from a Copernicus ESSD preprint.

Collection Method

Collected from Gaofen-2 and Beijing-2 satellite imagery, aerial photography, and UAV orthophotos.