• 图纸下载
  • 专业文献
  • 行业资料
  • 教育专区
  • 应用文书
  • 生活休闲
  • 杂文文章
  • 范文大全
  • 作文大全
  • 达达文库
  • 文档下载
  • 音乐视听
  • 创业致富
  • 体裁范文
  • 当前位置: 达达文档网 > 图纸下载 > 正文

    基于深度学习的高分辨率遥感图像海陆分割方法

    时间:2021-01-14 08:01:40 来源:达达文档网 本文已影响 达达文档网手机站

    摘 要:将高分辨率遥感图像进行像素级海陆分割是遥感应用领域的一项基础性工作,对海岸线提取和海洋近岸目标检测具有重要意义,但传统阈值方法往往由于高分辨率遥感图像覆盖范围广、地物纹理复杂等特点而难以取得预期效果。为了提升高分辨率遥感影像海陆分割精度,改善传统阈值方法的不足,基于深度神经网络模型利用编码器—解码器架构,并在编码层中引入残差块,以更好地对特征图进行高级语义信息提取,通过解码层将编码层生成的特征图还原成与输入尺寸相同的特征图,最后通过Sigmoid层对图像进行像素级海陆分割。在高分辨率遥感图像数据集上的实验结果表明,该网络模型取得良好了分割效果,准确率和Kappa系数分别达到了94.3%和93.7%。与传统方法相比,海陆分割精确度得到了有效提升。

    关键词:深度学习;高分辨率遥感图像;海陆分割;深度神经网络;编码—解码架构

    DOI:10. 11907/rjdk. 192771

    中图分类号:TP301   文献标识码:A                文章编号:1672-7800(2020)003-0095-04

    Land and Sea Segmentation Method of High-Resolution Remote Sensing Image Based on Deep Learning

    CUI Hao

    (School of Computer Science & Engineering,Shandong University of Science & Technology,Qingdao 266590,China)

    Abstract:
    Pixel-level sea-land segmentation of high-resolution remote-sensing images is a basic work in remote sensing applications. It is of great significance for coastline extraction and marine near-shore target detection. However, However, the traditional threshold method is often difficult to obtain the expected results due to the wide coverage of high-resolution remote sensing images and the complex texture of the ground features. In order to improve the accuracy of sea land segmentation of high-resolution remote sensing image and improve the shortcomings of traditional threshold methods, based on the depth neural network model, the encoder decoder architecture is used, and residual blocks are introduced into the coding layer to better extract the high-level semantic information of the feature map. Through the decoding layer, the feature map generated by the coding layer is restored to the feature map with the same size as the input. Finally through the Sigmoid layer, the sea land segmentation of images is made at the pixel level. The experimental results on the high-resolution remote sensing image dataset show that the network model achieves good segmentation results, and the accuracy rate and Kappa coefficient reach 94.3% and 93.7%, respectively. Compared with the existing traditional methods, this method improves the accuracy of land and sea segmentation.

    Key Words:
    deep learning;high-resolution remote sensing image;sea and land segmentation;deep neural network; encoding-decoding architecture

    0 引言

    近年来,随着我国遥感卫星技术的快速发展,高分辨率遥感图像在海洋开发应用与权益保护监督等方面获得广泛应用。高分辨率遥感图像具有覆盖范围广、地物纹理信息丰富、成像光谱波段多、重访时间短等多种特征。高分辨率遥感图像解译也是数字图像分析的重要组成部分,已廣泛应用于土地测绘、环境监测、城市建设等领域。其中,语义分割在遥感图像解译中扮演重要角色,是低高层遥感图像处理及分析的重要衔接[1]。对高分辨率遥感图像进行海陆分割的目的是将遥感近岸图像准确地分割成海洋和陆地区域。提升高分辨率遥感图像海陆分割精确度,有利于对近海区域目标进行检测,并且获取的海岸线等信息对海岸演化分析[2] 、潮间带性质和分布信息提取[3]等具有重要意义。

    2.3 实验评价

    为了定量评估所提模型在高分辨率遥感影像中进行海陆分割的效果,引入3个指标,分别为准确率(P)、召回率(R)和F1分数(F1)。计算形式如下:

    其中,TP代表样本为正,预测结果为正;FP代表样本为负,预测结果为正;FN代表样本为正,预测结果为负。

    2.4 实验结果分析

    为了验证本文构建的深度学习模型在高分辨率遥感图像海陆分割任务中的有效性,将实验结果与传统LATM方法以及图像语义分割领域经典的深度学习网络模型FCN、PSPNet进行高分辨率遥感图像海陆分割效果对比,实验中使用相同的训练样本和验证样本,结果如图3、表1所示。

    由图3可知,传统方法LATM对于纹理和强度变化复杂的高分辨率遥感图像,不仅容易对陆地地物进行错误分类,而且在海陆分割处呈现明显不规则性,分割效果较差;FCN网络更容易对土地像素进行错误分类;海陆分割在PSPNet得到了较好结果,但仍然在港口处存着错误分类。与这些方法相比,本文方法(Ours)可以获得更一致的空间结果,海陆分割效果最好。

    由表1可知,传统方法LATM和FCN网络对高分辨率遥感图像海陆分割能力相对较弱,PSPNet能取得较好结果,本文网络模型在测试数据集上各项指标均表现出最优结果。

    3 结语

    为了更好地实现高分辨率遥感图像海陆分割,本文利用编码器—解码器架构,在编码层中引入残差块构建了一个新型网络,实现了高分辨率遥感图像端到端的海陆分类。为了验证网络架构的有效性,手动标记高分辨率遥感图像真值图,并在此数据集上与PSPNet等方法进行比较。

    实验结果表明,本文提出的网络结构取得了最好结果。但其海陆分割结果精度还有待进一步提升,尤其是海陆交界边缘部分,仍有一定误差,整个网络模型尚有改进空间。未来将重点对架构进行优化,以进一步提高分割准确性。

    参考文献:

    [1]苏健民,杨岚心,景维鹏. 基于U-Net的高分辨率遥感图像语义分割方法[J]. 计算机工程与应用,2019,55(7):207-213.

    [2]ZHANG S P,ZHANG C T.Image analysis for wave swash using color feature extraction[C]. Proceedings of the 2nd International Congress on Image and Signal Processing (CISP),2009:1-4.

    [3]QIN P.Waterline information extraction from radial sand ridge of south Yellow Sea[C]. Proceedings of the 6th International Congresson Imageand Signal Processing(CISP),2013:459:463.

    [4]LIU H,JEZEK K C.Automated extraction of coastline from satellite imagery by integrating canny edge detection and locally adaptive thresholding methods[J]. International Journal of Remote Sensing,2004,25(5):037-958.

    [5]ZHANG H W,ZHANG B M,GUO H T, etal.An automatic coastline extraction method based on active contour model[C]. Proceedings of the 21st International Conference on Geoinformatics,2013:111-115.

    [6]NIEDERMEIER A,ROMANEESSEN E,LEHNER S.Detection of coastlines in SAR images using wavelet methods [J]. IEEE Transactions on Geoscience and Remote Sensing,2000,38(5):2270-2281.

    [7]吳一全,刘忠林. 遥感影像的海岸线自动提取方法研究进展[J]. 遥感学报,2019,23(4):582-602.

    [8]MCFEETERS S K.The use of the normalized difference water index (NDWI) in the delineation of open water features[J].  International Journal of Remote Sensing,1996,17(7):1425-1432.

    [9]LIU G,ET AL.A new method on inshore ship detection in high-resolution satellite images using shape and context information[J]. Geoscience and Remote Letter,IEEE,2014.11(3):617-621.

    [10]CAI S,WU H M.Study on change detection of ship target based on sea-land segmentation[J]. Video Engineering,2010,34(5):109 -112.

    [11]LIU H, JEZEK K C.Automated extraction of coastline from satellite imagery by integrating Canny edge detection and locally adaptive thresholding methods[J]. International Journal of Remote Sensing,2004,25(5):937-958.

    [12]HU F,XIA G S,HU J W,et al.Transferring deep convolutional neural networks for the scene classification of high-resolution remote sensing imagery[J]. Remote Sensing,2015,7(11):14680-14707.

    [13]LONG J,SHELHAMER E,DARRELL T.Fully convolutional networks for semantic segmentation[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence,2014,39(4):640-651.

    [14]RONNEBERGER O,FISCHER P,BROX T.U-Net:Convolutional networks for biomedical image segmentation[J]. Medical Image Computing and Computer-Assisted Intervention(MICCAI),2015,9351:234-241.

    [15]BADRINARAYANAN V,KENDALL A,CIPOLLA R. SegNet:a deep convolutional encoder-decoder architecture for scene segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(12):2481-2495.

    [16]ZHAO H,SHI J,QI X,et al. Pyramid scene parsing net-work[DB/OL]. https://arxiv.org/abs/1612.01105,2016.

    (責任编辑:孙 娟)

    收稿日期:2020-01-03

    作者简介:崔昊(1993-),男,山东科技大学计算机科学与工程学院硕士研究生,研究方向为智能信息处理、机器学习。本文通讯作者:崔昊。

    相关热词搜索: 海陆 遥感 分割

    • 生活居家
    • 情感人生
    • 社会财经
    • 文化
    • 职场
    • 教育
    • 电脑上网