摘要
针对林权确权中图斑边界精度不足与作业效率低的问题,研究提出了基于遥感影像的图斑边界自动提取方法。通过融合归一化植被指数(Normalized Difference Vegetation Index,NDVI)、边缘梯度与归一化建筑指数(Normalized Difference Built-up Index,NDBI)构建多通道输入数据,设计边界增强网络并引入差异尺度卷积结构,实现复杂边界的高精度提取。实验结果表明,城市林地边界平均偏移小于2米,边界一致性优于传统方法,具备在确权前期识别中的推广价值。
To address the challenges of low boundary precision and limited operational efficiency in forest tenure verification,the study proposes an automatic plot boundary extraction method based on remote sensing imagery.By integrating normalized difference vegetation index(NDVI),edge gradient,and normalized difference built-up index(NDBI)to construct multi-channel input data,a boundary-enhanced network is designed with the incorporation of a differential scale convolutional structure to enable accurate extraction of complex boundaries.Experimental results show that the average boundary deviation for urban forest areas is less than 2 meters,and boundary consistency outperforms traditional methods.The approach demonstrates practical value and applicability in the early-stage identification of forest tenure verification.
作者
林楚
LIN Chu(Guangzhou Urban Planning&Design Survey Research Institute Co.,Ltd.,Guangzhou Guangdong 510060,China)
出处
《信息与电脑》
2025年第18期38-40,共3页
Information & Computer
关键词
林权图斑
遥感影像
边界提取
确权作业
forest right map spot
remote sensing imagery
boundary extraction
forest tenure verification work
