Automated non-destructive phenotyping of Camellia oleifera seedlings based on 3D point clouds
Phenotypic characterization of Camellia oleifera seedlings is crucial for cultivation management, variety breeding, and germplasm conservation. However, existing point clouds segmentation studies on this species are predominantly focused on canopy and fruit segmentation, which has constrained in-dep...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-12-01
|
| Series: | Smart Agricultural Technology |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2772375525004538 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Phenotypic characterization of Camellia oleifera seedlings is crucial for cultivation management, variety breeding, and germplasm conservation. However, existing point clouds segmentation studies on this species are predominantly focused on canopy and fruit segmentation, which has constrained in-depth phenotypic analysis. To bridge this gap, an automated, non-destructive algorithm was developed for extracting phenotypic parameters of C. oleifera seedlings with heights ranging from 35 cm to 60 cm. The proposed hierarchical segmentation algorithm was integrated with skeletonization and implemented a cyclic segmentation strategy, whereby seedling point clouds were partitioned into multiple regions. In each iteration, independent morphological analysis and skeletonization were conducted on a specific region, thereby enhancing the traditional clustering algorithm's sensitivity to local morphological feature variations. By extracting only the main stem or specific branches in each iteration, the integrity of the overall segmentation result was maintained, enabling the transition from canopy segmentation to precise stem-leaf segmentation. When compared to canopy segmentation that point clouds clustering, canopy height model, and layer stack fitting methods, the algorithm demonstrated improvements of 2.5 %, 7.5 %, and 11.5 % in segmentation accuracy, respectively. Subsequently, five key phenotypic parameters—plant height, stem diameter, leaf width, leaf length, and leaf area—were quantified using bounding boxes, improved Delaunay triangulation, and a novel slice projection algorithm. The measurement accuracies were determined to be 96.7 %, 93.4 %, 93.2 %, 90.1 %, and 88.4 % for each parameter, respectively. These results are indicative of a substantial advancement in non-destructive phenotyping methodologies for C. oleifera seedlings. |
|---|---|
| ISSN: | 2772-3755 |