Related Experiment Video
Updated: Apr 28, 2026
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
High-throughput estimation of sugarcane phenotypic traits using UAV multispectral data under high-density planting
Jinghao Li1, Yaning Li1, Lei Xu1
1State Key Laboratory for Conservation and Utilization of Subtropical Agro-bioresources, Guangxi Key Laboratory of Sugarcane Biology, Province and Ministry Cosponsored Collaborative Innovation Center of Canesugar Industry, College of Agriculture, Guangxi University, Nanning, Guangxi 530004, China.
Unmanned aerial vehicle (UAV) phenotyping enhances sugarcane breeding by accurately predicting traits like plant height using machine learning and clustering. This integrated approach improves trait prediction efficiency in large-scale breeding programs.
Area of Science:
- Agricultural Science
- Plant Breeding
- Remote Sensing Technology
Background:
- High-throughput phenotyping is crucial for efficient sugarcane breeding.
- Unmanned aerial vehicle (UAV)-based multispectral imagery offers a powerful tool for capturing detailed crop data.
- Predicting key phenotypic traits in sugarcane varieties is essential for genetic improvement.
Purpose of the Study:
- To develop robust predictive models for sugarcane phenotypic traits using UAV imagery.
- To evaluate the effectiveness of various feature processing and machine learning strategies for trait prediction.
- To establish an integrated framework for enhanced trait prediction in sugarcane breeding.
Main Methods:
- Utilized UAV-mounted multispectral sensors to image 652 sugarcane varieties.
- Extracted vegetation indices, texture indices, and canopy height parameters.
- Implemented feature selection (COR, SWR, PCA) and machine learning algorithms (LASSO, Ridge, SVM, RF, GBR), including ensemble methods (BMA, Stacking).
- Applied K-means clustering to partition sugarcane varieties into distinct groups.
Main Results:
- The Stacking ensemble method achieved the highest prediction accuracy (R² = 0.77 for plant height).
- K-means clustering identified two distinct sugarcane variety clusters.
- Cluster-specific models using PCA-processed features showed exceptional validation accuracy (e.g., R² = 0.94 for plant height).
Conclusions:
- An integrated framework combining feature processing, clustering, and ensemble learning significantly enhances trait prediction accuracy in UAV-based sugarcane phenotyping.
- This approach provides a valuable tool for accelerating sugarcane breeding programs.
- Optimized feature selection and cluster-specific modeling are key to maximizing predictive performance.
