Microscope image based fully automated stomata detection and pore measurement method for grapevines.

Area of Science:

• Plant physiology
• Computational biology
• Agricultural science

Background:

• Stomatal behavior is a key indicator of grapevine water stress and plant health.
• Microscope imaging is crucial for analyzing stomatal behavior.
• Current methods for stomatal analysis are largely manual and time-consuming.

Purpose of the Study:

• To develop a fully automated method for stomata detection and pore measurement in grapevines using microscope images.
• To improve upon existing manual, semi-automatic, and automatic methods for stomatal feature analysis.

Main Methods:

• A cascade object detection learning algorithm was employed for identifying multiple stomata in microscopic images.
• Image processing techniques, including segmentation and skeletonization, were used for estimating stomatal pore dimensions.
• The developed method was compared against template matching and maximum stable extremal regions approaches.

Main Results:

• The automated stomata detection achieved a precision of 91.68% and an F1-score of 0.85, outperforming existing methods.
• The segmentation and skeletonization approach accurately estimated stomatal pore dimensions, even for partially visible pores, with accuracies up to 99.43% for eccentricity.
• The method correctly identified stoma openings in 86.27% of test images.

Conclusions:

• The fully automated solution offers superior performance for stomata detection and measurement compared to current methods.
• The system accurately measures stomatal features from incomplete images and minimizes false positives.
• This automation accelerates plant health analysis, reducing the need for manual measurements and aiding in water stress assessment.