A Comprehensive Review of High Throughput Phenotyping and Machine Learning for Plant Stress Phenotyping

Area of Science:

• Agricultural Science
• Plant Biology
• Data Science

Background:

• Rapid adoption of ground and aerial platforms with multiple sensors for crop phenotyping.
• High throughput phenotyping (HTP) aims to reduce bottlenecks in breeding programs and accelerate genetic gain.
• Neglected areas like root phenotyping are being addressed with new tools.

Purpose of the Study:

• To review recent findings on machine learning (ML) and deep learning (DL) applications in plant stress phenotyping using HTP data.
• To provide an overview of available ML and DL tools, including their advantages and disadvantages.
• To discuss conceptual challenges and future perspectives in managing HTP data for plant breeding and pathology.

Main Methods:

• Review of literature on HTP platforms (ground-based, aerial, remote sensing) for plant stress phenotyping.
• Analysis of ML and DL approaches for extracting information from large HTP datasets.
• Discussion of techniques including feature extraction, classification, and prediction.

Main Results:

• HTP technologies generate big data, necessitating advanced analytical methods like ML and DL.
• ML and DL offer powerful tools for analyzing complex HTP data to derive meaningful conclusions.
• Various ML and DL tools are available, each with specific benefits and limitations for plant stress phenotyping.

Conclusions:

• ML and DL are essential for unlocking the potential of HTP data in plant breeding and pathology.
• Addressing conceptual challenges is key to advancing the use of HTP and ML/DL in agriculture.
• Future research should focus on integrating HTP platforms with sophisticated data analysis techniques for improved crop management and breeding.