Dynamic Monitoring of Crop Canopies Using Time-Series Point Clouds: Insights into Phenotypic Variation and Leaf-Level Photosynthetic Performance

Jiaren Zhou, Mengqi Zhang, Shulin Sun, Man Zhang, Minjuan Wang; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, 2025, pp. 7150-7160

Time-series point clouds provide a powerful means for continuous, high-resolution crop monitoring and quantitative growth analysis. This study constructed a spatiotemporal point cloud dataset spanning four species and eleven varieties, enabling organ-level instance segmentation, phenotypic trait extraction, growth quantification, and canopy photosynthesis assessment. We propose a skeleton-based method for organ segmentation and trait extraction, demonstrating robust performance across all crops. To leverage the temporal dimension, we introduce a novel organ matching algorithm, achieving over 0.823 accuracy across species. Combined with trait extraction, this yields time-series phenotypic data and the phenotypic variation rate, a metric for quantifying dynamic growth. These data were integrated into a canopy photosynthesis model to derive time-resolved metrics, including photosynthetic rate, light absorption, light use efficiency, and organ-level contributions. Our approach enables fine-grained analysis of crop morphology and physiology, offering new insights into growth dynamics and photosynthetic strategies. The datasets and code supporting this study are publicly available on GitHub at https://github.com/JiarenZhou/Time-Series_Crop_Photosynthesis.git.