Answers are grounded in this paper and cite their sources.
• This study proposed a 3D voxel-based LiDAR method to estimate Camellia oleifera canopy parameters. • The quadratic model outperformed the single-feature models in the yield estimation. • The multi-feature models led to more accurate estimates of yield than the single-feature models. • The LiDAR point clouds data derived estimates exhibited a high correlation with the measured ones. • The canopy relevant parameters had more influence on the yield than the height percentiles. Accurately and rapidly estimating the yield of Camellia oleifera trees is crucial for large-scale cultivation management. Individual tree parameters show direct yield relationships and require model integration. LiDAR captures individual tree point clouds, enabling scalable parameter and yield estimation for Camellia oleifera . However, such studies area are still lacking. This study constructed 3D voxel models from Terrestrial Laser Scanning (TLS) derived point clouds for 84 Camellia oleifera trees to estimate crown projection area ( CA ), crown volume ( CV ), and leaf area index ( LAI ). Individual tree characteristics, including north-south crown width ( C NS ), east-west crown width ( C EW ), tree ground diameter ( GD ), tree height ( H ), and height percentiles, were simultaneously acquired. Both linear and nonlinear models (power, exponential, logarithmic, and quadratic functions) were constructed to estimate Camellia oleifera yield using single feature variables and to examine their predictive relationships. Furthermore, multiple linear regression (MLR), support vector machines (SVM), random forests (RF), and Artificial Neural Network (ANN), were utilized for multivariate yield estimation. The results demonstrated that LiDAR-estimated parameters of Camellia oleifera trees exhibited strong agreement with field measurements, with high correlation coefficients (H: r = 0.94; C EW : r = 0.89; GD: r = 0.82), validating the efficacy of TLS for structural parameter estimation. All features showed significant yield correlations ( P <0.01), with quadratic models demonstrating optimal fit. The multivariate yield models demonstrated superior estimation accuracy compared to the single-feature models, with the ANN model performing best ( R ² = 0.8937, RMSE = 0.9429 kg, RRMSE =13.20 %). The incorporation of multiple features significantly improved yield estimation precision.
No takes yet. Share an insight, caveat, or question.
Tang et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: