Joint Multi-Leaf Segmentation, Alignment, and Tracking for Fluorescence Plant Videos.

Area of Science:

• Plant biology
• Image processing
• Computational biology

Background:

• Leaf-level photosynthetic analysis is crucial for plant research.
• Accurate segmentation, structural estimation, and temporal tracking of leaves are prerequisites.
• Existing methods may not adequately address the joint problem of multi-leaf segmentation, alignment, and tracking.

Purpose of the Study:

• To propose a novel framework for fluorescence plant video processing.
• To address the joint problem of multi-leaf segmentation, alignment, and tracking.
• To develop a quantitative evaluation framework and predictive models for algorithm guidance.

Main Methods:

• A two-stage approach: leaf segmentation and alignment on the last frame, followed by leaf tracking on preceding frames.
• Formulation of two optimization problems with shared objective function terms for alignment and tracking.
• Development of models to predict alignment accuracy and detect tracking failures.

Main Results:

• The proposed framework effectively segments, aligns, and tracks multiple leaves in fluorescence plant videos.
• Quantitative evaluation using four metrics demonstrates the algorithm's performance.
• Learned models provide guidance for subsequent plant biology analysis by predicting alignment accuracy and tracking failures.

Conclusions:

• The developed framework is effective, efficient, and robust for fluorescence plant video processing.
• The method facilitates detailed leaf-level photosynthetic analysis.
• The study contributes a valuable tool for advancing plant science research.