Advancing crop resilience and sustainability: the crucial role of realistic 3D canopy models in light interception analysis

Grasping the complex interplay between light and plant canopies is crucial for unlocking the secrets to enhanced crop yields and resilience. Researchers have traditionally used photosynthetically active radiation sensors and leaf area index calculations to measure light interception, but these methods struggle with the complex spatial arrangement of canopies. Recent advancements employ 3D models and optical simulations to conduct detailed analyses, yet obtaining realistic 3D canopy models (RCMs) in the field remains challenging due to technological limitations. The current focus is on improving these 3D reconstructions to precisely quantify light distribution and enhance our understanding of plant growth, with methods like Structure from Motion (SfM) offering promising avenues for more accurate field data acquisition.

In August 2023, Plant Phenomics published a research article entitled by “The Importance of Using Realistic 3D Canopy Models to Calculate Light Interception in the Field ”.

This study focuses on a comparative analysis of light interception between realistic 3D maize canopy models (RCM) and virtual canopy models (VCMs), aiming to substantially enhance the precision of light interception calculations. Detailed reconstruction of a large-area RCM was  realized employing an advanced unmanned aerial vehicle (UAV) with a cross-circling oblique (CCO) route, alongside a structure-from-motion multi-view stereo method. Three types of VCMs (VCM-1, VCM-4, and VCM-8) were then created by replicating 1, 4, and 8 individual realistic plants in the center of the RCM. The results indicated significant deviations in daily light interception per unit area (DLI) between the VCMs and RCM, with relative root mean square error (rRMSE) values of 20.22%, 17.38%, and 15.48% for VCM-1, VCM-4, and VCM-8 respectively. The deviation decreased as the number of plants used to replicate the virtual canopy increased, but even with eight plants, a significant discrepancy remained. The reconstructed 3D models provided detailed visualizations of the plant structure, showing high accuracy in estimating leaf dimensions, corroborated by R² and RMSE values. Comparing light interception at 48 and 70 days after sowing (DAS) revealed that the differences between RCMs and VCMs were smaller in the early stage than in the late stage, indicating a more pronounced variation in canopy structure and light interception as the plants matured. Hourly light capture comparisons also demonstrated a consistent trend, with the RCM capturing the intricate dynamics of light distribution more accurately than the VCMs, especially at later stages. 

Furthermore, the study explored the structural differences between RCM and VCMs, finding that as canopy density increased, the 1D phenotypic differences (like plant height and canopy cover) between the models diminished, while the 2D and 3D phenotypic differences (like plant area index and COV) increased. This signifies that the structural complexity of the canopy is better captured by RCM, particularly for denser canopies. The research confirmed that RCMs provide a more accurate representation of light interception in the field, particularly at later growth stages, and emphasized the importance of capturing realistic 3D canopy structures for precise light distribution analysis. Despite these advancements, the challenge of extracting precise leaf angle information persists, underscoring an urgent need for continued research and innovative methodologies to accurately segment individual plants and leaves from 3D point clouds.

In conclusion, the study not only validates the superiority of RCM over VCMs in analyzing light interception but also paves the way for groundbreaking advancements in agricultural research through accurate 3D reconstructions.

###

References

Authors

Shunfu  Xiao¹, Shuaipeng  Fei¹, Qing  Li¹, Bingyu  Zhang¹, Haochong  Chen¹, Demin  Xu¹, Zhibo  Cai¹, Kaiyi  Bi², Yan  Guo¹, Baoguo  Li¹, Zhen  Chen³,  and Yuntao  Ma^1*

Affiliations

¹College  of  Land  Science  and  Technology,  China  Agricultural  University,  Beijing,  China.  

²The  State  Key  Laboratory  of  Remote  Sensing  Science,  Aerospace  Information  Research  Institute,  Chinese  Academy  of  Sciences,  Beijing,  China.  

³Farmland  Irrigation  Research  Institute  of  Chinese  Academy  of  Agricultural  Sciences/Key Laboratory of Water-Saving Agriculture of Henan Province, Xinxiang, China.

About Yuntao Ma

She is a professor in the College  of  Land  Science  and  Technology at  China  Agricultural  University. Her research areas include:

1. 3D simulation and digital twinning of plants;
2. Data mining and application of plant growth information based on machine vision;
3.  Artificial Intelligence and Smart Agriculture;
4. Rapid investigation of large-scale breeding traits by drones;
5. R&D of multi-source sensors and application of digital agriculture;
6. Rapid Acquisition and Rendering of Large Scale 3D Scenes.