Remote Sensing Change Detection Based on Deep Belief Networks Optimized by Domain Knowledge
-
摘要: 为提高高分辨率遥感影像变化检测精度，提出一种以领域知识为优化策略的深度学习变化检测方法。利用改进的变化矢量分析和灰度共生矩阵算法获取影像的光谱和纹理变化，设定合理阈值获得变化区域待选样本；引入领域知识中图斑形状特征指数与光谱知识，筛选得到高质量的训练样本；构建并训练了深度置信网络模型，使用优化策略对深度学习变化检测结果进行优化，以减少“椒盐”噪声和伪变化区对检测精度的影响。通过高分二号与IKONOS影像的变化检测实验表明，该方法较优化前准确率与召回率最大增幅分别为7.58%和14.69%（高分二号）、17.08%和23.87%（IKONOS），虚警率和漏检率最大降幅为30.22%和23.30%（高分二号）、17.08%和23.87%（IKONOS），能够有效提高变化检测精度。Abstract:Objectives A method of deep learning change detection with domain knowledge as an optimization strategy was proposed to improve the change detection precision of high-resolution remote sensing images.Methods The improved change vector analysis algorithm and grey-level co-occurrence matrix algorithm were used to obtain the spectral and texture changes of images, and reasonable thresholds were set to divide the changed samples from the unchanged samples based on the spectral and texture change intensity maps. The pattern shape index and spectral knowledge in domain knowledge were introduced as an optimization strategy to filter the changed samples for obtaining high-quality training samples. The deep belief network model was constructed and trained, and the results of deep learning change detection were optimized by the optimization strategy to reduce the influence of "salt and pepper noise" and false change zones on the detection accuracy.Results The Results of change detection experiments show that the accuracies of Gaofen-2 and IKONOS imageswere increased by 7.58% and 14.69% and the recall by 17.08% and 23.87%, respectively, while the false alarms and were decreased by 30.22% and 23.30% and the missing alarms by 17.08% and 23.87%, respectively.Conclusions Compared with the method before the optimization strategy was adopted, the proposed method in this paper can effectively improve the precision of change detection, and it provides a new way of using remote sensing images to improve the precision of deep learning change detection.
-
Keywords:
- remote sensing /
- change detection /
- deep belief networks /
- domain knowledge /
- Gaofen-2(GF-2)
-
-
![]()
图 4 执行优化策略前后变化样本标记图(实验1)
Figure 4. Change Sample Marker Image Before and After Executing the Optimization Strategy (Experiment 1)
![]()
图 7 优化前后变化样本标记图(实验2)
Figure 7. Change Sample Marker Images Before and After Executing the Optimization Strategy (Experiment 2)
表 1 训练样本优化前后精度分析表(实验1)
Table 1 Precision Analysis Before and After Optimizing Training Samples (Experiment 1)
数量/个 准确率/% 召回率/% 虚警率/% 漏检率/% 正样本 负样本 A B C A B C A B C A B C 5 000 5 000 88.70 93.28 93.77 65.30 76.30 80.10 60.42 42.42 39.76 34.70 23.70 19.90 10 000 10 000 88.63 94.09 94.59 66.59 73.75 78.68 60.45 37.37 34.30 33.41 26.25 21.32 30 000 30 000 88.26 92.45 93.07 66.60 78.95 83.68 61.46 46.57 43.70 33.40 21.05 16.32 50 000 50 000 87.72 91.74 92.80 68.21 79.16 83.45 62.62 49.44 44.96 31.79 20.84 16.55 80 000 80 000 86.90 94.42 94.48 69.14 73.14 75.21 64.37 35.09 34.15 30.86 26.86 24.79 
下载: 导出CSV
表 2 训练样本优化前后精度分析表(实验2)
Table 2 Precision Analysis Before and After Optimizing Training Samples (Experiment 2)
数量/个 准确率/% 召回率/% 虚警率/% 漏检率/% 正样本 负样本 A B C A B C A B C A B C 5 000 5 000 73.27 83.57 86.29 64.58 73.37 77.79 45.42 27.75 23.23 35.42 26.63 22.21 10 000 10 000 72.20 83.83 85.70 66.89 77.86 79.02 47.13 28.99 25.26 33.11 22.14 20.98 30 000 30 000 72.28 84.37 86.97 57.26 79.39 81.13 46.57 28.42 23.27 42.74 20.61 18.87 50 000 50 000 73.83 83.93 86.52 65.20 80.86 82.28 44.60 29.88 24.87 34.80 19.14 17.72 80 000 80 000 72.95 84.51 87.46 63.06 81.81 84.87 45.80 29.04 23.83 36.94 18.19 15.13
下载: 导出CSV
-
[1] 佟国峰, 李勇, 丁伟利, 等. 遥感影像变化检测算法综述[J]. 中国图象图形学报, 2015, 20(12): 1561-1571 doi: 10.11834/jig.20151201 Tong Guofeng, Li Yong, Ding Weili, et al. Review of Remote Sensing Image Change Detection[J]. Journal of Image and Graphics, 2015, 20(12): 1561-1571 doi: 10.11834/jig.20151201
[2] 张鑫龙, 陈秀万, 李飞, 等. 高分辨率遥感影像的深度学习变化检测方法[J]. 测绘学报, 2017, 46(8): 999-1008 https://www.cnki.com.cn/Article/CJFDTOTAL-CHXB201708010.htm Zhang Xinlong, Chen Xiuwan, Li Fei, et al. Change Detection Method for High Resolution Rmote Sensing Images Using Deep Learning[J]. Ata Geodaetica et Cartographica Sinica, 2017, 46 (8): 999-1008 https://www.cnki.com.cn/Article/CJFDTOTAL-CHXB201708010.htm
[3] 李胜. 联合领域知识和深度学习的城市地表覆盖变化检测方法[D]. 武汉: 武汉大学, 2018 Li Sheng. Detection Method of Urban Land Cover Change Based on Combining Domain Knowledge and Deep Learning[D]. Wuhan: Wuhan Universty, 2018
[4] Zhao Z Q, Guo L, Jia M, et al. The Generalized Gamma-DBN for High-Resolution SAR Image Clasification[J]. Remote Sensing, 2018, 10(6): 878 doi: 10.3390/rs10060878
[5] Traore B B, Kamsu-Foguem B, Tangara F. Deep Convolution Neural Network for Image Recognition [J]. Ecological Informatics, 2018, 48: 257-268 doi: 10.1016/j.ecoinf.2018.10.002
[6] Simonyan K, Zisserman A. Very Deep Convolutional Networks for Large-Scale Image Recognition[EB/ OL]. [2018-09-15]. https://arxiv.org/abs/1409.1556
[7] Hou B, Wang Y H, Liu Q J. Change Detection Based on Deep Features and Low Rank[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14 (12): 2418-2422 doi: 10.1109/LGRS.2017.2766840
[8] He K M, Zhang X Y, Ren S Q, et al. Deep Residual Learning for Image Recognition[C]//IEEE Coference on Computer Vision and Pattern Recogntion, Las Vegas, NV, USA, 2016
[9] 季顺平, 田思琦, 张驰. 利用全空洞卷积神经元网络进行城市土地覆盖分类与变化检测[J]. 武汉大学学报·信息科学版, 2020, 45(2): 233-241 doi: 10.13203/j.whugis20180481 Ji Shunping, Tian Siqi, Zhang Chi. Urban Land Cover Classification and Change Detection Using Fully Atrous Convolutional Neural Network[J]. Geomatics and Information Science of Wuhan Unversity, 2020, 45(2): 233-241 doi: 10.13203/j.whugis20180481
[10] Peng D F, Zhang Y J, Guan H Y. End-to-End Change Detection for High Resolution Satellite Images Using Improved UNet++[J]. Remote Sensing, 2019, 11(11), DOI: 10.3390/rs11111382
[11] Mou L C, Bruzzone L, Zhu X X. Learning SpectraSpatial-Temporal Features via a Recurrent Convoltional Neural Network for Change Detection in Mutispectral Imagery[J]. IEEE Transactions on Gescience and Remote Sensing, 2019, 57(2): 924-935 doi: 10.1109/TGRS.2018.2863224
[12] 眭海刚, 冯文卿, 李文卓, 等. 多时相遥感影像变化检测方法综述[J]. 武汉大学学报·信息科学版, 2018, 43(12): 1885-1898 doi: 10.13203/j.whugis20180251 Sui Haigang, Feng Wenqing, Li Wenzhuo, et al. Rview of Change Detection Methods for Multi -Tem - poral Remote Sensing Imagery[J]. Geomatics and Information Science of Wuhan University, 2018, 43 (12): 1885-1898 doi: 10.13203/j.whugis20180251
[13] 赵生银, 安如, 朱美如. 联合像元-深度-对象特征的遥感图像城市变化检测[J]. 测绘学报, 2019, 48(11): 1452-1463 https://www.cnki.com.cn/Article/CJFDTOTAL-CHXB201911012.htm Zhao Shengyin, An Ru, Zhu Meiru. Urban Change Detection by Aerial Remote Sensing Using Com - bining Features of Pixel-Depth-Object[J]. Acta Gedaetica et Cartographica Sinica, 2019, 48(11): 1452-1463 https://www.cnki.com.cn/Article/CJFDTOTAL-CHXB201911012.htm
[14] Huang X, Liu X B, Zhang L P. A Multichannel Gray Level Co-occurrence Matrix for Multi/Hypespectral Image Texture Representation[J]. Remote Sensing, 2014, 6(9), DOI: 10.3390/rs6098424
[15] Kupidura P. The Comparison of Different Methods of Texture Analysis for Their Efficacy for Land Use Classification in Satellite Imagery[J]. Remote Sensing, 2019, 11(10), DOI: 10.3390/rs11101233
[16] Park K, Kim D H. Accelerating Image Classifiction Using Feature Map Similarity in Convolutional Neural Networks[J]. Applied Sciences, 2018, 9 (1), DOI: 10.3390/app9010108
[17] Xiao P F, Zhang X L, Wang D G, et al. Change Detection of Built-Up Land: A Framework of Com - bining Pixel-Based Detection and Object-Based Recognition[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2016, 119: 402-414 doi: 10.1016/j.isprsjprs.2016.07.003
[18] Li J J, Xi B B, Li Y S, et al. Hyperspectral Classifcation Based on Texture Feature Enhancement and Deep Belief Networks[J]. Remote Sensing, 2018, 10(3), DOI: 10.3390/rs10030396
[19] Gao X G, Li F, Wan K F. Accelerated Learning for Restricted Boltzmann Machine with a Novel Mometum Algorithm[J]. Chinese Journal of Electronics, 2018, 27(3): 483-487 doi: 10.1049/cje.2018.03.013
-
期刊类型引用(4)
1. 朱杰，郑加柱，陈红华，杨静，胡平昌，陆敏燕. 结合POI数据的南京市商业中心识别与集聚特征研究. 现代测绘. 2022(06): 34-39 . 
百度学术
2. 金澄，安晓亚，陈占龙，马啸川. 矢量居民地多边形多级图划分聚类方法. 武汉大学学报(信息科学版). 2021(01): 19-29 . 百度学术
3. 张铭龙，何贞铭. 基于因子分析法的城市商业中心抽取研究. 地理空间信息. 2021(08): 58-60+64+5 . 百度学术
4. 李卫东，张铭龙，段金龙. 基于POI数据的南京市空间格局定量研究. 世界地理研究. 2020(02): 317-326 . 百度学术
其他类型引用(3)
计量
- 文章访问数: 703
- HTML全文浏览量: 172
- PDF下载量: 83
- 被引次数: 7
