海洋测绘2023，Vol.43Issue(6)：66-69,82,5.DOI:10.3969/j.issn.1671-3044.2023.06.015
基于改进DeepLab v3+模型的养殖浮筏提取研究
Research on extraction of aquatic floating raft based on improved DeepLab v3+model
摘要
Abstract
In response to the lack of consideration of multi temporal and feature diversity issues in existing research on extracting aquaculture floating rafts,this paper proposes a method for extracting aquaculture floating rafts based on an improved DeepLab v3+model.A dataset of aquaculture floating raft images is constructed based on multi year Sentinel-2 remote sensing images;by charging the ASPP module structure in the model to DenseASPP and optimizing the expansion rate;replacing the original feature extraction network Xception with the lightweight network MobileNet v2;the addition of CBAM attention mechanism module effectively improves the extraction accuracy and efficiency of the model,taking into account multi temporal and feature diversity in breeding floating rafts.The experiment is conducted using Sentinel-2 remote sensing images covering Changhai County as the data source.The results show that the evaluation indicators of the improved DeepLab v3+model were higher than those of the DeepLab v3+model,with an accuracy rate of 89.64％and 88％,respectively.关键词
遥感影像/遥感智能解译/深度学习/DeepLab v3+模型/养殖浮筏Key words
remote sensing images/remote sensing intelligent interpretation/deep learning/deeplab v3+model/breeding floating rafts分类
天文与地球科学引用本文复制引用
刘远大,宋伟东,蓝歆玫,李佳,薛国坤..基于改进DeepLab v3+模型的养殖浮筏提取研究[J].海洋测绘,2023,43(6):66-69,82,5.基金项目
国家自然科学基金(42071343). （42071343）
