融合多尺度特征注意力的遥感影像变化检测方法

doi:10.11947/j.AGCS.2022.20200540

摘要/Abstract

摘要： 深度学习技术已经成为遥感影像变化检测研究的主流方法，现有的基于深度学习的变化检测方法主要是获取单一尺度的变化特征，而在现实场景中，变化区域的尺度具有多样性。为此，本文提出了融合多尺度特征注意力的遥感影像变化检测方法，通过关注多尺度融合策略来解决变化检测存在的多尺度问题。首先，利用特征金字塔网络自身的多尺度特性，使网络学习到不同尺度的变化特征，为了提升网络感受野和利用全局特征信息，在特征提取网络末端引入扩张卷积空间金字塔模块；然后，在不同变化特征融合时，使用变化特征融合模块来控制信息传播以减少特征融合时的差异性；最后，使用门控机制，将不同尺度预测的变化特征图进行加权求和，最终产生具有高精度的变化特征图。本文方法不仅能获取多尺度变化特征，还能利用全局信息和精确的空间细节来提升预测特征图的空间精度。对比试验表明，本文方法在变化检测基准数据集CDD和LEVIR-CD上取得了较好的结果，召回率分别提高了6.58%和5.26%。

关键词: 变化检测, 特征金字塔网络, 多尺度特征, 注意力融合, 门控机制

Abstract: Deep learning technology has become the mainstream method of remote sensing image change detection research. Existing change detection methods based on deep learning mainly obtain the change characteristics of a single scale. In the real scene, the scale of the change area is diverse. Therefore, we propose a change detection method of multi-scale feature attention fusion, which solves the multi-scale problem of change detection by focusing on multi-scale fusion strategy. We take advantage of the multi-scale characteristics of the feature pyramid network, the purpose is to enable the network to learn change features in different scales; meanwhile, in order to improve receptive field of network and exploit global information, atrous convolutional spatial pyramid module is introduced at the end of the feature extraction network; In addition, when different change features are fused, the change feature fusion module is used to control information flow to reduce the difference in feature fusion; Finally, the gating mechanism is utilized to perform weighted summation of the change feature maps predicted by different scales, and a high precision change feature map is generated. The proposed method can not only obtain multi-scale change features, but also use global information and precise spatial details to improve the spatial accuracy of the predicted feature maps. Experimental results show that our method has achieved competitive results on the change detection benchmark datasets CDD and LEVIR-CD, and the recall rate has increased by 6.58% and 5.26%, respectively.

Key words: change detection, feature pyramid network, multi-scale feature, attention fusion, gating mechanism

中图分类号:

P237

梁哲恒, 黎宵, 邓鹏, 盛森, 姜福泉. 融合多尺度特征注意力的遥感影像变化检测方法[J]. 测绘学报, 2022, 51(5): 668-676.

LIANG Zheheng, LI Xiao, DENG Peng, SHENG Sen, JIANG Fuquan. Remote sensing image change detection fusion method integrating multi-scale feature attention[J]. Acta Geodaetica et Cartographica Sinica, 2022, 51(5): 668-676.

参考文献 25

[1]	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. SUI Haigang, FENG Wenqing, LI Wenzhuo, et al. Review of change detection methods for multi-temporal remote sensing imagery[J]. Geomatics and Information Science of Wuhan University, 2018, 43(12):1885-1898.
[2]	TEWKESBURY A P, COMBER A J, TATE N J, et al. A critical synthesis of remotely sensed optical image change detection techniques[J]. Remote Sensing of Environment, 2015, 160:1-14.
[3]	钟家强,王润生.基于自适应参数估计的多时相遥感图像变化检测[J]. 测绘学报, 2005, 34(4):331-336. ZHONG Jiaqiang, WANG Runsheng. Multitemporal remote sensing image change detection based on adaptive parameter estimation[J]. Acta Geodaetica et Cartographica Sinica, 2005, 34(4):331-336.
[4]	黄世奇,刘代志,胡明星,等. 基于小波变换的多时相SAR图像变化检测技术[J]. 测绘学报, 2010, 39(2):180-186. HUANG Shiqi, LIU Daizhi, HU Mingxing, et al. Multi-temporal SAR image change detection technique based on wavelet transform[J]. Acta Geodaetica et Cartographica Sinica, 2010, 39(2):180-186.
[5]	BENEDEK C, SZIRANYI T. Change detection in optical aerial images by a multilayer conditional mixed Markov model[J]. IEEE Transactions on Geoscience and Remote Sensing, 2009, 47(10):3416-3430.
[6]	ZHAO Jiaojiao, GONG Maoguo, LIU Jia, et al. Deep learning to classify difference image for image change detection[C]//Proceedings of 2014 International Joint Conference on Neural Networks (IJCNN). Beijing, China:IEEE, 2014:411-417.
[7]	SHELHAMER E, LONG J, DARRELL T. Fully convolutional networks for semantic segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(4):640-651.
[8]	ZHAN Yang, FU Kun, YAN Menglong, et al. Change detection based on deep Siamese convolutional network for optical aerial images[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14(10):1845-1849.
[9]	ALCANTARILLA P F, STENT S, ROS G, et al. Street-view change detection with deconvolutional networks[J]. Autonomous Robots, 2018, 42(7):1301-1322.
[10]	PAPADOMANOLAKI M, VERMA S, VAKALOPOULOU M, et al. Detecting urban changes with recurrent neural networks from multitemporal sentinel-2 data[C]//Proceedings of 2019 International Geoscience and Remote Sensing Symposium. Yokohama, Japan:IEEE, 2019:214-217.
[11]	RONNEBERGER O, FISCHER P, BROX T. U-net:Convolutional networks for biomedical image segmentation[C]//Proceedings of 2015 International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham, Germany:Springer, 2015:234-241.
[12]	PENG Daifeng, ZHANG Yongjun, GUAN Haiyan. End-to-end change detection for high resolution satellite images using improved UNet++[J]. Remote Sensing, 2019, 11(11):1382.
[13]	ZHOU Z, SIDDIQUEE M M R, TAJBAKHSH N, et al. UNet++:a nested U-net architecture for medical image segmentation[J]. Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support, 2018, 11045:3-11.
[14]	MOU Lichao, BRUZZONE L, ZHU Xiao xiang. Learning spectral-spatial-temporal features via a recurrent convolutional neural network for change detection in multispectral imagery[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(2):924-935.
[15]	CAYE DAUDT R, LE SAUX B, BOULCH A. Fully convolutional Siamese networks for change detection[C]//Proceedings of the 25th IEEE International Conference on Image Processing. Athens, Greece:IEEE, 2018:4063-4067.
[16]	CHEN Jie, YUAN Ziyang, PENG Jian, et al. DASNet:dual attentive fully convolutional Siamese networks for change detection in high-resolution satellite images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, 14:1194-1206.
[17]	ZHANG Mengya, XU Guangluan, CHEN Keming, et al. Triplet-based semantic relation learning for aerial remote sensing image change detection[J]. IEEE Geoscience and Remote Sensing Letters, 2019, 16(2):266-270.
[18]	ZHANG Chenxiao, YUE Peng, TAPETE D, et al. A deeply supervised image fusion network for change detection in high resolution bi-temporal remote sensing images[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2020, 166:183-200.
[19]	CHEN Hao, SHI Zhenwei. A spatial-temporal attention-based method and a new dataset for remote sensing image change detection[J]. Remote Sensing, 2020, 12(10):1662.
[20]	LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, HI, USA:IEEE, 2017:936-944.
[21]	HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NV, USA:IEEE, 2016:770-778.
[22]	CHEN L C, PAPANDREOU G, SCHROFF F, et al. Rethinking atrous convolution for semantic image segmentation[EB/OL].[2021-01-22] https://arxiv.org/abs/1706.05587.
[23]	YU F, KOLTUN V. Multi-scale context aggregation by dilated convolutions[C]//Proceedings of 2016 International Conference on Learning Representations.[S.l.]:ICLR,2016.
[24]	KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60(6):84-90.
[25]	LEBEDEV M A, VIZILTER Y V, VYGOLOV O V, et al. Change detection in remote sensing images using conditional adversarial networks[J]. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2018, 42(2):1428-1441.