测绘通报 ›› 2021, Vol. 0 ›› Issue (11): 96-100.doi: 10.13474/j.cnki.11-2246.2021.346

• 技术交流 • 上一篇    下一篇

基于注意力机制的高分遥感图像采石场识别

马林飞, 倪欢, 周子涵   

  1. 南京信息工程大学遥感与测绘工程学院, 江苏 南京 210044
  • 收稿日期:2021-07-19 出版日期:2021-11-25 发布日期:2021-12-02
  • 通讯作者: 倪欢。E-mail:nih@nuist.edu.cn
  • 作者简介:马林飞(2001-),男,主要研究方向为深度学习在遥感图像处理中的应用。E-mail:1402986996@qq.com
  • 基金资助:
    2021年高水平大学一省级大学生创新创业训练计划(202110300041Z);国家自然科学基金(41801384);江苏省自然科学基金(BK20180795);南京信息工程大学人才启动项目(2018r030)

Quarry recognition based on attention mechanism from high-resolution remote sensing imagery

MA Linfei, NI Huan, ZHOU Zihan   

  1. School of Remote Sensing & Geomatics Engineering, Nanjing University of Information Science & Technology, Nanjing 210044, China
  • Received:2021-07-19 Online:2021-11-25 Published:2021-12-02

摘要: 采石业对国民经济起到积极作用的同时,给生态、环境和人居安全带来了隐患。目前,石矿区生态修复成为重要的生态文明建设问题。解决该问题的前提是识别采石场且确定石矿区边界。高分遥感对地观测和深度学习技术为高精度识别采石场并确定矿区边界提供了有效的途径。本文基于残差网络,设计了注意力金字塔结构和细节增强模块,以少量参数和极低计算复杂度为代价,从无人机高分遥感图像中有效识别石矿区。试验采用福建省南安市高分遥感图像和人工实地考察获取的石矿区边界真值作为数据源,构建用于卷积神经网络训练的数据集,并对算法精度进行验证。试验结果表明,本文方法速度快、精度高,可为石矿区生态修复提供可靠的支撑材料。

关键词: 采石场, 高分遥感图像, 卷积神经网络, 注意力机制, 生态修复

Abstract: Quarrying industry plays a positive rule for national economy. At the same time, it arouses hidden danger to ecology, environment, and human security. Currently, quarry ecological restoration has become an important issue for ecological civilization construction. The precondition to solve this problem is quarry recognition and boundary determination. High-resolution remote sensing earth observation and deep learning techniques provide an efficient way for this task. In this paper, we design an attention pyramid and a detail enhancement module based on residual network to recognize quarry in UAV high-resolution images, using only a small number of parameters and extremely low computational complexity. The experiments employ the UAV images and human annotated ground truths in Nanan city of Fujian province as the data source. Then, we construct a dataset for training convolutional neural networks, and validate the performance of our proposed method. The experimental results show that our proposed method is with high inference speed and accuracy, and can be used to provide supporting material to quarry ecological restoration.

Key words: quarry, high-resolution images, convolutional neural networks, attention mechanism, ecological restoration

中图分类号: