文献类型：期刊文献

英文题名：A novel method for forest spatial structure heterogeneity evaluation of plantation utilizing point-wise vector network and neighborhood index

作者：Qiu, Hanqing[1,2] Zhang, Huaiqing[1,2] Lei, Kexin[1,2] Wang, Jiansen[1,2] Zhang, Huacong[1,2,3] Yu, Longhua[1,2,3]

第一作者：Qiu, Hanqing

通信作者：Zhang, HQ[1];Zhang, HQ[2]

机构：[1]Chinese Acad Forestry, Inst Forest Resource Informat Tech, Beijing 100091, Peoples R China;[2]NFGA, Key Lab Forest Management & Growth Modeling, Beijing 100091, Peoples R China;[3]Chinese Acad Forestry, Expt Ctr Subtrop Forestry, Fenyi 336600, Peoples R China

年份：2025

卷号：229

外文期刊名：COMPUTERS AND ELECTRONICS IN AGRICULTURE

收录：;EI(收录号：20245017523345);Scopus(收录号：2-s2.0-85211589704);WOS:【SCI-EXPANDED(收录号:WOS:001386008100001)】；

基金：Acknowledgments This work was supported by the National Natural Science Foundation of China (No. 32271877) , the National Key Research and Development Program of China (No. 2022YFE0128100) , the National Natural Science Foundation of China (No. 32071681)

语种：英文

外文关键词：Forest structure heterogeneity; Point-wise deep learning; Neighborhood index; Individual tree extraction; Forest management practices

摘要：Existing forest structure heterogeneity studies focus on single-perspective analysis and field investigation, lacking top-down perspective and automated forest heterogeneity quantification methods. Quantifying and elucidating the stand structural heterogeneity differences is challenging in uneven-aged forests. This paper addresses such challenges by presenting a point-wise stand spatial structure heterogeneity evaluation approach through deep learning and multi-dimensional structure index. This method aims to fill this gap of micro-region forest structure heterogeneity evaluation, and optimize the priority region selection of forest management. This research contributes to decreasing the costs of labor-intensive forest management. The approach was applied in a subtropical plantation in Southern China. Results showed that point clouds exhibit great potential in depicting fine-scale stand spatial structure heterogeneity. The precision and the F-score of the individual tree extraction were more than 95 % and 92 % in different forest types, respectively. The analysis of the tree- and plot-level stand structure indicated that the spatial heterogeneity of tree height and crown are at a relatively low level. Mixed forests with pronounced vertical layering are considered to have a more heterogeneous structural composition. Large-scale structure mapping showed that spatial distribution pattern is predominantly characterized by uniform distribution, and the overall degree of competition is relatively low. Priority forest management area were found to be in the northwest and southeast with a relatively high stem density distribution. The top-down 3D perspective of UAV-LiDAR makes up for the limitations of the 2D perspective of UAV images and traditional filed measured-based structure calculation. Our method provides a novel and feasible solution for micro area stand structure heterogeneity mapping across various forest types.