Related Experiment Video
Updated: Jan 8, 2026
Collecting and Processing Drone-based Remotely Sensed Data for Use in Forest Recovery Monitoring
Published on: October 24, 2025
A novel point cloud completion model for three-dimensional reconstruction of complex, dynamic population-level crop
Ziyue Guo1, Xin Yang1, Yutao Shen1
1State Key Laboratory for Vegetation Structure, Function and Construction (VegLab), College of Biosystems Engineering and Food Science, Zhejiang University, Hangzhou 310058, P.R. China; Key Laboratory of Spectroscopy Sensing, Ministry of Agriculture and Rural Affairs, Hangzhou 310058, P.R. China.
This study introduces a new framework using drone imagery and a point cloud completion model to accurately reconstruct dense crop canopies. This improves 3D canopy architecture analysis for better crop yield prediction.
Area of Science:
- Agricultural Engineering
- Computer Vision
- Plant Science
Background:
- Accurate 3D canopy architecture is crucial for crop yield prediction and ideotype design.
- Existing 3D reconstruction methods struggle with dense canopies due to occlusion.
Purpose of the Study:
- To develop a framework for 3D reconstruction of complex, dynamic crop population canopy architecture.
- To address occlusion issues in dense crop populations using a novel point cloud completion model.
Main Methods:
- Developed a complete point cloud generation pipeline for automated data annotation.
- Proposed the Crop Population Point Cloud Completion Network (CP-PCN) integrating MRDG, PPD, DGCFE, and GAN loss.
- Utilized unmanned aerial vehicle (UAV) multi-view imagery for rapeseed canopy reconstruction.
Main Results:
- CP-PCN achieved superior performance (Chamfer Distance 3.35-4.51 cm) compared to state-of-the-art methods.
- Reconstruction completeness by CP-PCN led to more accurate yield estimation.
- Demonstrated cross-crop generalizability by successfully reconstructing rice canopies.
Conclusions:
- The developed framework provides a scalable approach for quantitative analysis of complex canopy architectures.
- CP-PCN effectively addresses occlusion challenges in 3D crop reconstruction.
- Accurate 3D canopy characterization enhances crop modeling and breeding efforts.
