A TIF Format Image for the Paper "An Automated Skeleton Extraction Method for 3D Point-Cloud Phenotyping of Schima Superba Seedlings"

This study proposes an automated, non-invasive method for acquiring phenotypic parameters of Schima superba seedlings using 3D point cloud data. Schima superba seedlings were scanned from six angles, and their 3D models were reconstructed using 4PCS and ICP algorithms. A "density-weighted voxel centroid method" was developed to extract stem skeletons, which integrates a minimum spanning tree and principal component analysis for accurate identification. The stem and leaves were effectively separated using Dijkstra's algorithm combined with spatial search and geometric analysis. Leaf segmentation was refined through locally preserved convexity clustering and the K-means++ algorithm. Key phenotypic traits, including plant height, stem length, stem diameter, and leaf area, were calculated with correlation coefficients of 0.994, 0.992, 0.938, and 0.873, respectively, significantly enhancing the accuracy and efficiency of managing Schima superba seedlings.