PypeTree: a tool for reconstructing tree perennial tissues from point clouds

Area of Science:

• Botany
• Computer Science
• Forestry

Background:

• Terrestrial LiDAR scanning (TLS) generates point clouds crucial for studying tree development.
• Accurate tree reconstruction from point clouds is vital for ecological modeling.
• Existing skeletal extraction methods require adaptation for complex perennial tissue structures.

Purpose of the Study:

• To develop an efficient and accurate method for reconstructing tree perennial tissues from TLS point clouds.
• To introduce PypeTree, an open-source tool integrating enhanced skeletal extraction and semi-supervised adjustment.
• To improve tree modeling and developmental studies through precise structural reconstruction.

Main Methods:

• Extensive modifications to the Verroust and Lazarus skeletal extraction method (2000).
• Development of PypeTree, a visual modeling environment with user-friendly interfaces.
• Implementation of semi-supervised adjustment tools to handle imperfect point cloud data.
• Validation using synthetic models and real-world tree data.

Main Results:

• Automatic reconstruction accuracy varied, particularly for small branches (<3.5 cm).
• Mean reconstruction error for cumulated skeleton length was 5.1% (automatic) and 1.8% (semi-supervised).
• Semi-supervised tools significantly improved reconstruction accuracy, enabling perfect perennial tissue reconstruction in some cases.

Conclusions:

• PypeTree offers an efficient and adaptable solution for tree perennial tissue reconstruction from TLS data.
• Semi-supervised approaches enhance accuracy, addressing challenges posed by imperfect point clouds.
• The developed method represents a significant advancement for tree study and developmental modeling.