Multi-class tree crown detection in unmanned aerial vehicle (UAV) imagery is essential for forest monitoring but remains challenged by severe target-scale variation, feature ambiguity caused by crown overlap and background interference, and unstable bounding box regression in crowded scenes. To address these coupled errors, we propose Density-Guided YOLO with Frequency Decoupling and Large Selective Kernels (DFL-YOLO), a YOLOv11-based framework that combines scale-context construction, density-conditioned feature disambiguation, and improved bounding box regression within a progressive detection process. First, structurally pruned Large Selective Kernel (LSK) modules combine cascaded depthwise convolutions with spatial selection to enrich local and long-range context along selected backbone and neck paths. Second, our Density-Guided Frequency Decoupling Branch (DFDB) predicts a Gaussian-supervised density mask from the resulting context-enhanced P4 feature and applies complementary modulation to its high- and low-frequency components, preserving boundary-sensitive responses while suppressing background interference. The modulated P4 feature is then fused with the P3 and P5 paths and passed to the multiscale detection head. Finally, Multi-Strategy Weighted Intersection over Union (MW-IoU) supervises bounding box regression by combining MPDIoU geometric constraints, the non-monotonic focusing mechanism of Wise-IoU, and task-oriented soft weights. On the Mixed NEON dataset constructed from NEON and Roboflow imagery, DFL-YOLO achieved 82.2% mAP@50 and 54.7% mAP@50:95, representing absolute improvements of 2.6% and 4.2% over the YOLOv11l baseline, respectively. After dataset-specific retraining on VisDrone2019, DFL-YOLO achieved 54.5% mAP@50 and 33.7% mAP@50:95, showing that the architecture can be trained for another dense aerial-imaging domain. In general, DFL-YOLO is a reliable method for detecting multi-class tree crowns in intricate forest environments.
IntroductionThis study examines the relationship between household participation in the underforest economy and multidimensional child welfare in rural China.MethodsWe use a purposive survey of 786 households with children aged 7–15 in five mountainous counties of Hunan Province, conducted in 2023. We construct a welfare index covering education, health, mental well-being, care, and development opportunities. The analysis uses regression models with village-clustered standard errors, alternative welfare measures and specifications, propensity score matching, inverse-probability-weighted regression adjustment, and instrumental-variable estimation based on historical forestry-station accessibility.ResultsUnderforest participation is associated with a 0.141-unit increase in the welfare index, equivalent to 0.383 standard deviations. Positive associations are found for education, health, mental well-being, and care; the estimate for development opportunities is positive but statistically imprecise. The baseline association remains stable across robustness checks. Historical forestry-station accessibility predicts participation, and the 2SLS estimates are positive under both county and township fixed effects. Participating households report higher income and education spending, greater parental local employment and companionship, and lower parental migration. Children spend 0.576 additional hours per week on household work/help.DiscussionThe findings are consistent with resource-enhancing and care-retention pathways linking underforest participation to child welfare, alongside a time-allocation trade-off. Forest livelihood programs should consider children’s schooling, care, and work demands when assessing household benefits.
IntroductionTo assess forest ecosystem service value (FESV) in China and support ecological conservation and high-quality forest ecosystem development, this study examined FESV from 2000 to 2023 from a geographical perspective.MethodsThe equivalent factor method was applied with coefficient adjustments at the provincial and county levels. County-level FESV estimates were aggregated to obtain provincial and national estimates for each year, and the center-of-gravity model, standard deviational ellipse, and Geodetector were used for analysis.ResultsThe results showed that: (1) FESV in China followed an overall upward trend with fluctuations, increasing by 12% from 2000 to 2023. Regulating services accounted for the largest share, exceeding 66% in every year. (2) Total FESV was highest in western China, which accounted for more than 46% of the national total on average over the study period. Central and northeastern China occupied intermediate positions, while eastern China had the lowest total FESV. FESV per unit forest ecosystem area was highest in eastern China, while central China generally ranked second, with relatively lower values in northeastern and western China. Provinces in the highest FESV class (Level 5) were mainly located in western and northeastern China, and their number increased from two to four. The distribution of low-value provinces remained relatively stable. (3) The spatial distribution of FESV exhibited an overall northeast–southwest orientation, with a relatively stable distributional shape. The center of gravity remained within Henan Province and shifted predominantly northward, with a slight eastward component. Its locations in 2000 and 2023 were 88.42 km apart. (4) Among the natural factors selected in this study, slope and elevation had the greatest explanatory power for the spatial variation in FESV within the Geodetector framework. All pairwise factor interactions exhibited either bivariate enhancement or nonlinear enhancement in each representative year.DiscussionThese findings can inform the spatial allocation of forest resources and the formulation of ecological conservation and management policies.
