基于改进卷积神经网络的地表覆盖分类方法
Land Cover Classification Method Based on Improved Convolutional Neural Network
张荞 1任星辰1
作者信息
- 1. 西南石油大学地球科学与技术学院,四川成都 610500
- 折叠
摘要
卷积神经网络在高分辨率影像分类中对全局上下文信息的精炼能力不足,导致分类精度较低.针对此问题,文章提出基于改进卷积神经网络的地表覆盖分类方法.该方法采用基于自注意力模块的骨干网络增强聚合全局上下文信息的能力,在土地覆盖数据集中取得优于其他对比网络的分割效果.
Abstract
Convolutional neural networks have insufficient refining ability for global contextual information in high-resolution image classification,resulting in lower classification accuracy.In response to this issue,the article proposes a land cover classification method based on an improved convolutional neural network.This method utilizes a backbone network based on self attention modules to enhance the ability to aggregate global contextual information,achieving better segmentation results than other comparative networks in land cover datasets.
关键词
改进卷积神经网络/地表覆盖分类/遥感影像Key words
improve convolutional neural networks/classification of surface cover/remote sensing images引用本文复制引用
出版年
2024
NETL
NSTL
万方数据