杨知, 欧文浩, 刘晓燕, 李闯, 费香泽, 赵斌滨, 刘龙, 马潇. 基于LinkNet卷积神经网络的高分辨率遥感影像水体信息提取[J]. 云南大学学报(自然科学版), 2019, 41(5): 932-938. doi: 10.7540/j.ynu.20180782
引用本文: 杨知, 欧文浩, 刘晓燕, 李闯, 费香泽, 赵斌滨, 刘龙, 马潇. 基于LinkNet卷积神经网络的高分辨率遥感影像水体信息提取[J]. 云南大学学报(自然科学版), 2019, 41(5): 932-938. doi: 10.7540/j.ynu.20180782
YANG Zhi, OU Wen-hao, LIU Xiao-yan, LI Chuang, FEI Xiang-ze, ZHAO Bin-bin, LIU Long, MA Xiao. Water information extraction for high resolution remote sensing image based on LinkNet convolutional neural network[J]. Journal of Yunnan University: Natural Sciences Edition, 2019, 41(5): 932-938. DOI: 10.7540/j.ynu.20180782
Citation: YANG Zhi, OU Wen-hao, LIU Xiao-yan, LI Chuang, FEI Xiang-ze, ZHAO Bin-bin, LIU Long, MA Xiao. Water information extraction for high resolution remote sensing image based on LinkNet convolutional neural network[J]. Journal of Yunnan University: Natural Sciences Edition, 2019, 41(5): 932-938. DOI: 10.7540/j.ynu.20180782

基于LinkNet卷积神经网络的高分辨率遥感影像水体信息提取

Water information extraction for high resolution remote sensing image based on LinkNet convolutional neural network

  • 摘要: 针对高空间分辨率(简称高分辨率)遥感影像在水体信息提取时准确度不高的问题,采用LinkNet卷积神经网络模型对高分辨率遥感影像进行水体信息提取,验证该网络模型对于水体信息提取的可行性及有效性. 首先选取包含水体的影像数据作为训练数据,然后构建LinkNet卷积神经网络模型,并利用构建好的网络模型训练得到水体信息的先验模型,最后采用该模型对真实高分辨率遥感影像进行水体信息提取,并与经典算法比较. 实验结果表明:LinkNet卷积神经网络模型能够实现高分辨率遥感影像水体信息的高精度提取,且优于经典算法.

     

    Abstract: In order to solve the problem of low accuracy of high spatial resolution (abbreviated as high resolution) remote sensing image in water information extraction, this paper tries to use LinkNet Convolutional Neural Network (CNN) model to extract water information from high-resolution remote sensing image, and to verify the feasibility and effectiveness of this network model in water information extraction. Firstly, the image of water is selected as training data, and then the LinkNet CNN model is constructed. Then, the prior model of water information is trained by the LinkNet CNN model. Finally, the model is used to extract the water information from the real high resolution remote sensing image and is compared with the existing algorithm. Experimental results show that the LinkNet CNN is capable of achieving high precision extraction of water information for high resolution remote sensing image, and is superior to the contrast algorithm.

     

/

返回文章
返回