基于遥感影像的烟羽语义分割技术研究
首发时间:2024-04-26
摘要:化石燃料的燃烧是我国区域性污染天气形成的重要原因,其主要表现形式是燃煤电厂排出的烟羽。随着遥测和深度学习技术的蓬勃发展,基于遥感影像的语义分割技术也取得了长足进步。针对烟羽狭长的形态特征,本文设计了一个即插即用的多尺度多模板动态条形卷积模块,替代传统的方形卷积;此外,对语义分割模型中的下采样模块加入了空间和通道注意力约束,减少了信息损失。上述工作均已在公开数据集上进行了对比实验和消融实验,实验结果验证了方法的有效性。?????
For information in English, please click here
Research on Semantic Segmentation of Smoke Plumes Based on Remote Sensing Images
Abstract:The combustion of fossil fuels is a significant contributor to the formation of regional pollution weather in China, manifested primarily by smoke plumes emitted from coal-fired power plants. With the vigorous development of remote sensing and deep learning technologies, significant progress has been made in semantic segmentation techniques based on remote sensing images. In response to the elongated morphological characteristics of smoke plumes, an adaptable multi-scale multi-template dynamic strip convolution module is designed to replace traditional square convolutions. Additionally, spatial and channel attention constraints are integrated into the downsampling module of the semantic segmentation model to reduce information loss. The effectiveness of the proposed approach has been validated through comparative experiments and ablation studies on public datasets.
Keywords: Remote Sensing Images Semantic Segmentation Irregular Convolution
基金:
引用
No.****
同行评议
勘误表
基于遥感影像的烟羽语义分割技术研究
评论
全部评论0/1000