An Accurate Skeleton Extraction Approach From 3D Point Clouds of Maize Plants

Area of Science:

• Plant science
• Computer vision
• Agricultural engineering

Background:

• Accurate plant phenotyping is vital for crop improvement and research.
• Existing 3D point cloud methods struggle with accurate estimation of maize traits across growth stages.
• A robust method is needed to bridge the gap between 3D data and trait analysis.

Purpose of the Study:

• To develop an accurate skeleton extraction algorithm for 3D maize plant point clouds.
• To enable precise estimation of phenotyping parameters throughout the plant's life cycle.
• To improve the efficiency and accuracy of maize plant analysis.

Main Methods:

• Point cloud preprocessing: clustering and color-based denoising.
• Skeletonization: Laplacian contraction and adaptive sampling for key point selection.
• Calibration: local coordinate system adjustment for precise skeleton alignment.

Main Results:

• Successfully extracted accurate plant skeletons from 3D point clouds.
• Achieved high precision in estimating maize traits (NRMSE < 8.37%).
• Processing time per plant is under 100 seconds, demonstrating efficiency.

Conclusions:

• The proposed skeleton extraction method accurately represents maize plant structures from 3D point clouds.
• This technique significantly enhances the precision of phenotyping parameter estimation.
• The method supports advanced maize research, including genotype-to-phenotype studies and modeling.