In-field High Throughput Phenotyping and Cotton Plant Growth Analysis Using LiDAR

Area of Science:

• Agricultural Engineering
• Plant Science
• Remote Sensing

Background:

• In-field high-throughput phenotyping is crucial for advancing plant breeding and various plant science applications.
• Existing methods often lack the efficiency and precision required for large-scale field studies.

Purpose of the Study:

• To develop and validate a terrestrial LiDAR-based high-throughput phenotyping system for in-field plant analysis.
• To derive and analyze key morphological traits of cotton plants and their correlation with final yield.

Main Methods:

• A 2D LiDAR and RTK-GPS system was employed for overhead plant scanning and spatial coordinate acquisition.
• 3D plant models were reconstructed, with ground plane separation and noise removal (weeds) using RANSAC and Euclidean clustering algorithms.
• Morphological traits (canopy height, projected canopy area, plant volume) were derived using histogram analysis, point projection, and a Trapezoidal rule-based algorithm.

Main Results:

• High agreement was found between LiDAR-derived and manual measurements for canopy height (R²=0.97), projected canopy area (R²=0.97), and plant volume (R²=0.98).
• Growth trends and rates were established for cotton cultivars from 43 to 109 days after planting, with growth stabilizing around 88 days.
• Morphological traits showed strong correlations with final yield, particularly between 67 and 109 days after planting (max R² up to 0.88).

Conclusions:

• The developed terrestrial LiDAR system offers a high-throughput solution for in-field plant phenotyping.
• The system accurately quantifies cotton plant morphology and provides valuable insights into growth dynamics and yield prediction.
• This technology has significant potential to accelerate plant breeding programs and enhance plant science research.