基于深度学习的森林资源变化图斑自动提取  

Automatic Extraction of Forest Resource Change Maps Based on Deep Learning

作　　者：张水花 秦琳 黄宁辉 刘新科 孟先进 陈鑫 薛亚东 胡圣元 ZHANG Shuihua;QIN Lin;HUANG Ninghui;LIU Xinke;MENG Xianjin;CHEN Xin;XUE Yadong;HU Shengyuan(Guangdong Forest Inventory and Planning Institute,Guangzhou 510520,Guangdong,China)

机构地区：[1]广东省林业调查规划院,广东广州510520

出　　处：《桉树科技》2023年第2期35-40,共6页Eucalypt Science & Technology

摘　　要：随着多学科交叉及智能技术的发展,一些高新技术方法应用到遥感分类中,如卷积神经网络算法和语义分割模型算法等,充实和完善了森林资源动态监测的方法。文章以广州市增城区为例,基于深度学习的全流程自动提取技术,探求森林资源变化图斑自动提取。研究发现,对模型进行多次迭代优化后,自动提取的准确率可达85%以上,为森林资源动态监测提供了新的技术手段。With the development of interdisciplinary and intelligent technology,some high-tech methods can be applied to remote sensing classification,such as convolution neural network algorithms and semantic segmentation model algorithms,to enrich and improve the methods of dynamic monitoring of forest resources.Taking Zengcheng District of Guangzhou as an example,this paper explores the automatic extraction of forest resource change map spots based on the whole process automatic extraction technology of deep learning.It was found that after repeated iterative optimization of the model,the accuracy of automatic extraction can reach more than 85%,which provides a certain technical support for the study of dynamic monitoring of forest resources.

关 键 词：人工智能 深度学习 变化图斑 森林资源动态监测 模型 

分 类 号：S771.8[农业科学—森林工程]