Computer vision-based phenotyping for improvement of plant productivity: a machine learning perspective

Area of Science:

• Agricultural Science
• Computer Science
• Plant Biology

Background:

• Computer vision is crucial for identifying plant phenotypic changes.
• Machine learning, particularly convolutional neural networks, enhances image analysis for plant phenotyping.
• Multi-sensor data acquisition provides comprehensive insights into plant development and physiology.

Purpose of the Study:

• To review emerging aspects of computer vision for automated plant phenotyping.
• To highlight the role of advanced image analysis and machine learning in high-throughput phenotyping.
• To discuss the application of computer vision in both controlled environments and large-scale field settings.

Main Methods:

• Utilizing machine learning-based techniques, including convolutional neural networks, for image analysis.
• Employing combinatorial use of multiple sensors to acquire spectral data.
• Leveraging automated phenotyping platforms and remote sensing technologies (e.g., unmanned vehicles).

Main Results:

• Computer vision accelerates the elucidation of gene functions related to plant traits.
• Automated platforms enable high-throughput phenotyping under controlled conditions.
• Remote sensing facilitates large-scale field phenotyping for crop breeding and precision agriculture.

Conclusions:

• Computer vision-based plant phenotyping is essential for modern agriculture and plant science.
• This technology supports both nowcasting and forecasting of plant traits by modeling genotype/phenotype relationships.
• The integration of computer vision promises significant advancements in crop breeding and precision agriculture.