吉林大学学报(工学版) ›› 2023, Vol. 53 ›› Issue (1): 248-254.doi: 10.13229/j.cnki.jdxbgxb20211389

• 计算机科学与技术 • 上一篇    

基于改进SegNet的遥感图像建筑物分割方法

朱冰1(),李紫薇2,李奇1,3()   

  1. 1.长春理工大学 计算机科学技术学院，长春 130022
    2.吉林师范大学 计算机学院，吉林 四平 136000
    3.长春理工大学 中山研究院，广东 中山 528437
  • 收稿日期:2021-12-16 出版日期:2023-01-01 发布日期:2023-07-23
  • 通讯作者: 李奇 E-mail:baulina@163.com;liqi@cust.edu.cn
  • 作者简介:朱冰（1992-），女，博士研究生. 研究方向：仿脑智能与脑信息学. E-mail： baulina@163.com
  • 基金资助:
    国家自然科学基金青年基金项目(81600923);深圳海外创新团队项目(KQTD20180413181834876);吉林省自然科学基金项目(20210101273JC);吉林省卫生与健康技术创新项目(2020J052);吉林大学白求恩计划项目(2020B47)

Building segmentation method of remote sensing image based on improved SegNet

Bing ZHU1(),Zi-wei LI2,Qi LI1,3()   

  1. 1.College of Computer Science and Technology，Changchun University of Science and Technology，Changchun 130022，China
    2.College of Computer，Jilin Normal University，Siping 136000，China
    3.Zhongshan Institute，Changchun University of Science and Technology，Zhongshan 528437，China
  • Received:2021-12-16 Online:2023-01-01 Published:2023-07-23
  • Contact: Qi LI E-mail:baulina@163.com;liqi@cust.edu.cn

RICH HTML

 12   

PDF (PC)

196

摘要：

针对高分辨率遥感图像建筑物分割精度低以及边缘模糊问题，在SegNet网络的基础上提出一种改进的全卷积神经网络。首先，选择在深度学习任务中表现良好的GELU作为激活函数，避免神经元失活；其次，在编码网络中使用改进的残差瓶颈结构提取更多的建筑物特征；然后，利用跳跃连接融合图像的低级与高级语义特征，辅助图像重构；最后，在解码网络末端连接改进的边缘修正模块进一步修正建筑物边缘细节，提升建筑物的边缘完整度。在Massachusetts Buildings Dataset数据集上进行实验，其精确率、召回率和F1值分别达到93.5%、79.3%和81.9%，综合评价指标F1值相比于基础网络提升约5%。

关键词: 计算机应用技术, 遥感图像, 建筑物分割, SegNet, 全卷积神经网络

Abstract:

Aiming at the low accuracy of building segmentation and boundary blur in high resolution remote sensing images， an improved full convolution network is proposed based on SegNet network. Firstly， GELU， which performs well in deep learning task， is selected as the activation function to avoid neuron inactivation； Secondly， the improved residual bottleneck structure is used to extract more building features in the encoding network； Then， the skip connection is used to fuse the low-level and high-level semantic features of the image to assist image reconstruction； Finally， an improved boundary refinement module is connected at the end of the decoding network to further correct the boundary details of the building and improve the boundary integrity of the building. The experiment is carried out on the Massachusetts Buildings Dataset. The accuracy， recall and F1 score reach 93.5%， 79.3% and 81.9% respectively. The F1 score of the comprehensive evaluation index is about 5% higher than that of the basic network.

Key words: computer application technology, remote sensing images, building segmentation, SegNet, full convolution network

中图分类号: 

  • TP389.1

图1

RsBR-SegNet模型结构图"

图2

GELU激活函数图像"

图3

改进的残差瓶颈结构图"

图4

边缘修正模块对比图"

图5

Satellite dataset I （global cities）分割结果对比图"

表1

Satellite dataset I （global cities）评价"

网络名称精确率召回率F1值交并比
FCN0.863 150.692 520.725 430.575 86
SegNet0.839 970.620 440.666 170.506 69
U?Net0.866 160.731 940.731 960.598 67
RsBR?SegNet0.875 260.754 580.759 070.618 34

图6

Massachusetts Buildings Dataset分割结果对比图"

表2

Massachusetts Buildings Dataset评价"

网络名称精确率召回率F1值交并比
FCN0.917 890.738 960.771 180.630 75
SegNet0.918 440.731 450.770 000.629 98
U?Net0.923 450.758 020.793 030.674 29
RsBR?SegNet0.935 030.792 880.819 150.697 46
1 余帅, 汪西莉. 含多级通道注意力机制的CGAN遥感图像建筑物分割[J]. 中国图象图形学报, 2021, 26(3): 686-699.
Yu Shuai, Wang Xi-li. Remote sensing building segmentation by CGAN with multilevel channel attention mechanism[J]. Journal of Image and Graphics, 2021, 26(3): 686-699.
2 齐滨, 赵春晖, 王玉磊. 基于支持向量机与相关向量机的高光谱图像分类[J]. 吉林大学学报: 工学版, 2013, 43(S1): 143-147.
Qi Bin, Zhao Chun-hui, Wang Yu-lei. Hyperspectral imagery classification based on SVM and RVM[J]. Journal of Jilin University(Engineering and Technology Edition), 2013, 43(S1): 143-147.
3 Long J, Shelhamer E, Darrell T. Fully convolutional networks for semantic segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 39(4): 640-651.
4 Ronneberger O, Fischer P, Brox T. U-Net: convolutional networks for biomedical image segmentation[C]∥International Conference on Medical image computing and computer-assisted intervention. Cham: Springer International Publishing, 2015: 234-241.
5 Badrinarayanan V, Kendall A, Cipolla R. SegNet: A Deep Convolutional Encoder-Decoder Architecture for Scene Segmentation[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017, 39(12): 2481-2495.
6 何青, 孟洋洋, 李华智. 多层次编码—解码网络遥感图像建筑物分割[J]. 计算机应用研究, 2021, 38(8): 2510-2514.
He Qing, Meng Yang-yang, Li Hua-zhi. Multi-level encoding and decoding network remote sensing image building segmentation[J]. Application Research of Computers, 2021, 38(8): 2510-2514.
7 李传林, 黄风华, 胡威, 等. 基于Res_AttentionUnet的高分辨率遥感影像建筑物提取方法[J]. 地球信息科学学报, 2021, 23(12): 2232-2243.
Li Chuan-lin, Huang Feng-hua, Hu Wei, et al. Building extraction from high-resolution remote sensing image based on Res_AttentionUnet[J]. Journal of Geo-information Science, 2021, 23(12): 2232-2243.
8 杨潇宇, 汪西莉. 结合多尺度注意力和边缘监督的遥感图像建筑物分割模型[J]. 激光与光电子学进展, 2022, 59(22): 335-344.
Yang Xiao-yu, Wang Xi-li. Building segmentation model of remote sensing image combining multi-scale attention and edge supervision[J]. Laser & Optoelectronics Progress, 2022, 59(22): 335-344.
9 王舒洋, 慕晓冬, 杨东方, 等. 融合高阶信息的遥感影像建筑物自动提取[J]. 光学精密工程, 2019, 27(11): 2474-2483.
Wang Shu-yang, Mu Xiao-dong, Yang Dong-fang, et al. High-order statistics integration method for automatic building extraction of remote sensing images[J]. Optics and Precision Engineering, 2019, 27(11): 2474-2483.
10 Abdollahi A, Pradhan B, Alamri A M. An ensemble architecture of deep convolutional segnet and unet networks for building semantic segmentation from high-resolution aerial images[J]. Geocarto International, 2020, (3): 1-16.
11 Liu Y, Gross L, Li Z, et al. Automatic building extraction on high-resolution remote sensing imagery using deep convolutional encoder-decoder with spatial pyramid pooling[J]. IEEE Access, 2019, 7(1): 128774-128786.
12 Shunping J, Shiqing W, Meng L. Fully convolutional networks for multisource building extraction from an open aerial and satellite imagery data set[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(1): 574-586.
13 Mnih V. Machine Learning for Aerial Image Labeling[D]. Toronto: University of Toronto, 2013.
14 Hendrycks D, Gimpel K. Gaussian error linear units (gelus)[J/OL]. arXiv preprint arXiv: 1606. 08415, 2016.
15 He K, Zhang X, Ren S, et al. Deep Residual Learning for Image Recognition[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, 2016: 770-780.
16 Sandler M, Howard A, Zhu M, et al. Mobilenetv2: inverted residuals and linear bottlenecks[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Piscataway, USA, 2018: 4510-4520.
17 Peng C, Zhang X, Yu G, et al. Large kernel matters-improve semantic segmentation by global convolutional network[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Hawaii, USA, 2017: 4353-4361.
[1] 薛珊,张亚亮,吕琼莹,曹国华. 复杂背景下的反无人机系统目标检测算法[J]. 吉林大学学报(工学版), 2023, 53(3): 891-901.
[2] 时小虎,吴佳琦,吴春国,程石,翁小辉,常志勇. 基于残差网络的弯道增强车道线检测方法[J]. 吉林大学学报(工学版), 2023, 53(2): 584-592.
[3] 马为駽,张䶮,马传香,朱飒. 不同光照条件下含噪遥感图像边缘检测算法[J]. 吉林大学学报(工学版), 2023, 53(1): 241-247.
[4] 王俊杰,农元君,张立特,翟佩臣. 基于施工场景的视觉关系检测方法[J]. 吉林大学学报(工学版), 2023, 53(1): 226-233.
[5] 秦贵和,黄俊锋,孙铭会. 基于双手键盘的虚拟现实文本输入[J]. 吉林大学学报(工学版), 2022, 52(8): 1881-1888.
[6] 曲福恒,丁天雨,陆洋,杨勇,胡雅婷. 基于邻域相似性的图像码字快速搜索算法[J]. 吉林大学学报(工学版), 2022, 52(8): 1865-1871.
[7] 白天,徐明蔚,刘思铭,张佶安,王喆. 基于深度神经网络的诉辩文本争议焦点识别[J]. 吉林大学学报(工学版), 2022, 52(8): 1872-1880.
[8] 刘铭,杨雨航,邹松霖,肖志成,张永刚. 增强边缘检测图像算法在多书识别中的应用[J]. 吉林大学学报(工学版), 2022, 52(4): 891-896.
[9] 方世敏. 基于频繁模式树的多来源数据选择性集成算法[J]. 吉林大学学报(工学版), 2022, 52(4): 885-890.
[10] 王生生,李晨旭,王翔宇,姚志林,刘一申,吴佳倩,杨晴然. 基于改进残差胶囊网络和麻雀搜索的脑瘤图像分类[J]. 吉林大学学报(工学版), 2022, 52(11): 2653-2661.
[11] 车翔玖,陈赫元. 基于改进YOLOv4的多目标光盘检测算法[J]. 吉林大学学报(工学版), 2022, 52(11): 2662-2668.
[12] 王生生,陈境宇,卢奕南. 基于联邦学习和区块链的新冠肺炎胸部CT图像分割[J]. 吉林大学学报(工学版), 2021, 51(6): 2164-2173.
[13] 赵宏伟,张子健,李蛟,张媛,胡黄水,臧雪柏. 基于查询树的双向分段防碰撞算法[J]. 吉林大学学报(工学版), 2021, 51(5): 1830-1837.
[14] 曹洁,屈雪,李晓旭. 基于滑动特征向量的小样本图像分类方法[J]. 吉林大学学报(工学版), 2021, 51(5): 1785-1791.
[15] 李雄飞,吴佳婧,张小利,王泽宇,冯云丛. 基于相对总变差结构提取的遥感图像融合[J]. 吉林大学学报(工学版), 2021, 51(5): 1775-1784.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
版权所有 © 《吉林大学学报(工学版)》编辑部
地址：长春市人民大街5988号　电话：+86-431-85095297　传真：+86-431-85094074　E-mail: xbgxb@jlu.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发