AppleQSM: Geometry-Based 3D Characterization of Apple Tree Architecture in Orchards.

Area of Science:

• Horticulture and Agricultural Engineering
• Computer Vision and 3D Reconstruction
• Forestry and Plant Science

Background:

• Apple tree architecture is fundamental for precision management and breeding, but traditional 2D imaging methods suffer from accuracy limitations due to occlusion and perspective issues.
• Accurate characterization of tree architectural traits, including height, trunk diameter, branch count, branch diameter, and branch angle, is essential for optimizing orchard productivity and advancing plant phenotyping.
• Terrestrial Laser Scanner (TLS) technology offers a potential solution for detailed 3D data acquisition of plant structures.

Purpose of the Study:

• To develop and validate a 3D geometry-based processing pipeline for segmenting apple tree structures and characterizing architectural traits using TLS point cloud data.
• To overcome the limitations of 2D imaging techniques in accurately assessing apple tree architecture.
• To provide a foundation for precision management in high-density apple orchards and enhance phenotyping in breeding programs.

Main Methods:

• A four-module processing pipeline was designed: data preprocessing, tree instance segmentation, tree structure segmentation, and architectural trait extraction.
• The pipeline utilized point cloud data acquired by a terrestrial laser scanner (TLS).
• The system was applied to analyze 84 apple trees from two distinct cultivars.

Main Results:

• The pipeline achieved high accuracy at the tree level, with R² values of 0.92 for tree height and 0.83 for trunk diameter, and corresponding Mean Absolute Errors (MAE) of 6.1 cm and 4.71 mm.
• At the branch level, the system demonstrated strong performance with R² values of 0.77 for branch diameter and 0.69 for branch angle, and MAEs of 6.86 mm and 7.48°, respectively.
• The study successfully characterized key architectural traits including tree height, trunk diameter, branch count, branch diameter, and branch angle.

Conclusions:

• The developed 3D geometry-based pipeline effectively segments apple tree structures and accurately extracts architectural traits from TLS data, surpassing traditional 2D methods.
• Accurate architectural trait measurements enable precision management in high-density orchards and support advanced phenotyping for apple breeding.
• The research identified general bottlenecks in 3D tree characterization, paving the way for future advancements in the field.