生物灾害科学2025，Vol.48Issue(2)：311-318,8.DOI:10.3969/j.issn.2095-3704.2025.02.40
基于北京二号遥感影像和深度学习的薇甘菊发生点识别
Identification of Mikania micrantha outbreak points based on Beijing No.2 remote sensing and deep learning
摘要
Abstract
[Objective]This study aimed to grasp the distribution and occurrence of Mikania micrantha and improve the efficiency and accuracy of large-scale monitoring.[Method]This study used the deep learning tool set in ArcGIS Pro to extract the outbreak point of Mikania micrantha in Baihua Town,Huizhou City,based on the high-resolution remote sensing images of Beijing No.2.Firstly,the actual ground object category of the sample area was obtained through drone aerial photography,then the sample was labeled and classified tiles were created in the satellite image data,and the model training and pixel classification were processed.[Result]The recall rate of the model based on DeepLab V3 was 58.43%,precision rate was 77.46%and the F1 score was 66.61%,being combined with the field object categories obtained by the field investigation,and the producer's accuracy and user's accuracy of the pixel classification results were 80.0%and 85.4%.[Conclusion]The results showed that the method based on the remote sensing images of Beijing No.2 Satellite and deep learning had high accuracy in the classification of Mikania micrantha,and the classification model had good performance.The method can provide basis and data support for the monitoring of Mikania micrantha on a large scale,and have important significance for the control of Mikania micrantha.关键词
高分遥感影像/深度学习/薇甘菊/外来入侵物种/监测Key words
high-resolution remote sensing images/deep learning/Mikania micrantha/alien invasive species/monitoring分类
农业科技引用本文复制引用
郭茂涛,陈湛昊,刘春燕,许少嫦,韦美满,杨振意,孙思..基于北京二号遥感影像和深度学习的薇甘菊发生点识别[J].生物灾害科学,2025,48(2):311-318,8.基金项目
广东省林业科技创新项目(2024KJCX010) （2024KJCX010）
