摘要
非法开采不仅危害国家资源,威胁国家财产安全,也存在重大安全隐患,寻找快速发现非法开采行迹的解决办法迫在眉睫。利用光学遥感影像进行人工解译,费时费力、效率极低;而传统的露天矿遥感自动解译方法,或基于像素,或基于面向对象,利用的图像特征简单且数量较少。将深度学习的全卷积神经网络算法引入露天矿自动提取中,充分从底层特征中挖掘大量高层抽象特征,实现露天矿智能高效解译。实验结果表明,该方法在一定程度上有效提高了露天矿识别的准确率,能够为及时发现非法露天矿开采提供基础的数据技术支持。
Illegal mining occurs frequently,which not only endangers national resources and threatens the security of national property,but also may bring major security risks.Therefore,it is extremely urgent to find a solution to early and quickly find the illegal mining track.The manual interpretation of optical remote sensing image is time-consuming and laborious,and the efficiency is extremely low.Traditional automatical open-pit mine interpretation in remote sensing method is either based on pixels,or object-oriented,which uses simple and limited image characteristics.In this paper we introduce the deep learning algorithm FCN into open-pit mine automatic extraction.This method can fully dig high-level features from the underlying ones,realizing the interpretation of open-pit mine intelligently and efficiently.The experimental results showed that,the method improves the accuracy of open-pit mine recognition in a certain extent,which will provide a fundenmental technical support for the timely detection of illegal open mining.
作者
刘发发
韩红太
张敏
麻连伟
Liu Fafa;Han Hongtai;Zhang Min;Ma Lianwei(Henan Institute of Geophysics and Spatial Information,Zhengzhou 450009,China;Henan Geological and Geophysical Exploration Engineering Technology Research Center,Zhengzhou 450009,China)
出处
《能源与环保》
2021年第6期82-85,262,共5页
CHINA ENERGY AND ENVIRONMENTAL PROTECTION
基金
河南省自然资源科研项目资助(豫财招标采购-2019-379-15)。
关键词
深度学习
全卷积神经网络
光学遥感影像
露天矿
自动提取
deep learning
fully convolution networks
optical remote sensing images
open-pit mine
automatic extraction