DBH

DBH estimates diameter at breast height, tree height, and biomass in plantation forests using UAV-mounted hyperspectral imaging and machine learning to assess cultivar growth rates.

Key Features:

• Integrated Pipeline Methodology: Integrates UAV acquisition with hyperspectral imaging to enable biomass measurement in dense-crown plantations such as Cunninghamia lanceolata.
• Hyperspectral Imaging: Captures hyperspectral images with UAV-mounted cameras and orthorectifies them using HiSpectral Stitcher.
• Vegetation Indices and Feature Extraction: Extracts features including 19 vegetation indices from hyperspectral data, producing a dataset of approximately 12,000 pixels for analysis.
• Machine Learning Algorithms: Models vegetation indices in Python using decision trees, random forests, support vector machines (SVM), and eXtreme Gradient Boosting (XGBoost).
• Classification and Validation: Classifies trees into fast, median, and normal growth categories using manually measured DBH from 2,880 samples with cross-validation, with random forests achieving 75% prediction accuracy.

Scientific Applications:

• Cultivar Screening: Quantifies DBH and height to screen cultivar growth performance, including applications to Cunninghamia lanceolata.
• Biomass Estimation: Supports biomass estimation and carbon stock assessment in high-density plantation forests using hyperspectral-derived indices.
• Scalability and Adaptability: Applied to Cunninghamia lanceolata plantations in Fujian, China and adaptable to other tree species and plantation contexts.

Methodology:

Acquire UAV-mounted hyperspectral images; orthorectify with HiSpectral Stitcher; extract features including 19 vegetation indices (~12,000 pixels); model indices in Python using decision trees, random forests, SVM, and XGBoost; classify growth categories using 2,880 manually measured DBH samples with cross-validation (random forest accuracy 75%).