An automated phenotyping method for Chinese Cymbidium seedlings based on 3D point cloud

Area of Science:

• Plant Science
• Computer Vision
• Agricultural Technology

Background:

• Manual determination of Cymbidium seedling phenotypic parameters is inefficient and costly.
• Accurate phenotypic data is crucial for plant breeding and research.

Purpose of the Study:

• To develop a fully automated measurement scheme for Cymbidium seedling phenotypic parameters.
• To address the challenge of segmenting individual tillers based on their morphology.

Main Methods:

• A novel point cloud segmentation method was designed, involving two rounds to separate overlapping tiller parts.
• Segmentation was based on edge point clouds and horizontal slicing weighted by tiller structure.
• The algorithm aligns with tiller growth direction for accurate skeleton point extraction.

Main Results:

• The automated system accurately calculated five phenotypic parameters: plant height, leaf number, leaf length, leaf width, and leaf area.
• Achieved accuracies of 98.6%, 100%, 92.2%, 89.1%, and 82.3% for the respective parameters.
• The method significantly improved the prediction accuracy of phenotypic parameters compared to artificial approaches.

Conclusions:

• The proposed automated scheme offers an efficient and cost-effective solution for measuring Cymbidium seedling phenotypes.
• The segmentation algorithm's accuracy in capturing tiller morphology ensures reliable phenotypic data for research and applications.