Convolutional Neural Networks for Image-Based High-Throughput Plant Phenotyping: A Review

Area of Science:

• Agricultural Science
• Computer Science
• Plant Biology

Background:

• Plant phenotyping is crucial for crop improvement and understanding plant-environment interactions but remains a bottleneck.
• High-throughput imaging has emerged as a powerful tool for plant phenotyping, generating vast amounts of data.
• Developing rapid and accurate analytical tools for image-based phenotyping is essential.

Purpose of the Study:

• To provide a comprehensive review of deep convolutional neural networks (CNNs) in plant phenotyping.
• To examine CNN applications for plant stress evaluation, development monitoring, and postharvest quality assessment.
• To identify state-of-the-art solutions and future research directions in CNN-based plant phenotyping.

Main Methods:

• Systematic review of recent studies utilizing deep convolutional neural networks (CNNs) for plant phenotyping.
• Categorization of studies based on technical developments: image classification, object detection, and image segmentation.
• Analysis of CNN architectures applied to diverse phenotyping tasks.

Main Results:

• CNNs demonstrate significant potential for high-throughput plant phenotyping, addressing previous limitations.
• Various CNN architectures are effectively applied for accurate plant stress evaluation, development tracking, and quality assessment.
• Image classification, object detection, and segmentation techniques using CNNs offer robust solutions for phenotyping challenges.

Conclusions:

• Deep convolutional neural networks are transforming plant phenotyping by enabling rapid and accurate trait extraction from imaging data.
• CNNs offer advanced analytical capabilities for critical agricultural applications, including breeding and crop management.
• Further research into CNN architectures will continue to enhance the efficiency and scope of plant phenotyping.