Individual TLS Tree Clouds from Alto Tajo and Cuellar, Spain

36 plots across two Spanish sites were scanned with a Leica HDS6200 scanner at 3.2mm resolution. Individual trees were segmented, zero-centered, and downsampled to 5 cm for species classification using deep learning. Harry Owen from the University of Cambridge created this dataset, which includes species IDs for five tree types and metadata on single or multi-stem status.

Use Cases

• Classifying tree species (Quercus faginea, Quercus ilex, Pinus nigra, Pinus sylvestris, Pinus pinaster) based on TLS-derived point clouds.
• Training deep learning models for automated tree segmentation and feature extraction from 3D scans.
• Analyzing forest structure by distinguishing between single-stem and multi-stem trees as indicated in the metadata.

Strengths

• Data originates from 36 distinct 30x30 meter plots, providing a structured sampling framework.
• Individual trees were processed with a combination of automated (treeseg) and manual methods, suggesting careful refinement.
• Each tree point cloud is zero-centered and downsampled to a consistent 5 cm resolution, which may aid model training.

Limitations

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

Provenance

Source

University of Cambridge (Harry Owen), via paperswithcode.

Collection Method

Collected using a Leica HDS6200 terrestrial laser scanner following a grid pattern; trees were co-registered, segmented, and manually refined.

Geography

Alto Tajo and Cuellar, Spain.