3D point cloud driven organ semantic segmentation to assess maize structural responses along the planting-density...

Area of Science:

• Agricultural Science
• Plant Biology
• Digital Agriculture

Background:

• High-density planting boosts maize yield but requires plants to adapt structurally.
• Understanding plant architectural responses is crucial for optimizing dense planting strategies.

Purpose of the Study:

• To investigate maize structural response mechanisms under varying planting densities.
• To develop and validate a high-throughput 3D phenotyping system for complex field conditions.

Main Methods:

• Developed a multi-view 3D reconstruction system for high-precision point clouds.
• Utilized deep learning for semantic segmentation of plant organs (95.6% accuracy).
• Employed clustering for individual leaf separation (94.8% accuracy) and extracted 45 architectural traits.

Main Results:

• Increased planting density led to more compact plant forms and centralized ear height.
• Ear leaves showed high sensitivity to density, with significant changes in area, distribution, and inclination.
• Principal component analysis identified key traits driving structural differentiation under density stress.

Conclusions:

• The integrated 3D phenotyping pipeline enables robust structural analysis and trait extraction in maize.
• This approach supports intelligent breeding selection and optimization of dense planting strategies.
• The system offers valuable technical support for digital agriculture and crop improvement.