Automatic Measurement of Morphological Traits of Typical Leaf Samples

Area of Science:

• Agricultural Engineering
• Computer Vision
• Plant Science

Background:

• Automatic plant measurement is crucial for agricultural research and development.
• Existing methods face challenges with occluded plants and require manual intervention.
• Accurate plant phenotyping is essential for high-throughput screening and crop improvement.

Purpose of the Study:

• To propose a novel method for automatic plant measurement using a handheld 3D laser scanner.
• To enable automatic selection of typical leaf samples and estimation of morphological traits from occluded plants.
• To assess the accuracy, efficiency, and robustness of the proposed method.

Main Methods:

• Data acquisition using a handheld 3D laser scanner to create high-precision 3D plant models.
• Data processing involving multi-level region growing segmentation and two leaf shape models for leaf extraction.
• Estimation of four scale-related and six scale-invariant morphological traits.

Main Results:

• Successfully scanned 94.02% of typical leaf samples across different canopy occlusions.
• Achieved 87.61% accuracy in automatic extraction of typical leaf samples.
• Demonstrated high correlation (EF > 0.8919 for scale-related, EF > 0.7434 for scale-invariant traits) with manual measurements.
• Completed plant measurement in an average of 196.37 seconds.

Conclusions:

• The proposed method provides a robust and efficient solution for automatic plant measurement.
• It shows significant potential for real-time plant measurement and high-throughput phenotyping, even with canopy occlusion.
• The low time cost and high accuracy make it suitable for practical applications in plant science.