计算机科学 ›› 2023, Vol. 50 ›› Issue (6A): 220500040-5.doi: 10.11896/jsjkx.220500040
李滔, 王海瑞
LI Tao, WANG Hairui
摘要: 针对利用传统语义分割网络进行城市建设用地变化检测过程中出现的欠分割或者过分割、边缘分割粗糙等问题,文中提出了一种基于孪生注意力网络的高分辨率遥感影像变化检测方法。该方法在编码部分使用孪生神经网络进行特征采集,以保留更多的不同时相影像特征;深层编码阶段引入空洞卷积特征金字塔实现多尺度特征的提取与融合,增大网络感受野;解码部分使用注意力机制CBAM突出有用特征以增强有用信息,提高边缘分割精度;最后在娄底市土地利用变化数据集上进行实验。实验结果表明,该方法在娄底市土地利用变化检测数据集上的准确率达到92.56%,精确率达到89.15%,召回率达到85.61%,IoU达到77.53%,MIoU达到83.76%,F1分数达到87.34%,Kappa系数达到31.42%,性能指标优于FCN网络、U-Net网络、CBAM U-Net网络。实验结果表明,该方法可以有效解决变化检测结果欠分割或者过分割、边缘分割粗糙的问题。
中图分类号:
[1]ZHU X X,NING X G,WANG H,et al.Land use classification for optimization segmentation based on high-precision land cover data[J].Science of Surveying and Mapping,2021,46(6):140-149. [2]SONG K Q,JIE J.AGCDetNet:an attention-guided network for building change detection in high-resolution remote sensing images[J].IEEE Journal of Selected Topics in Applied Earth Observation and Remote Sensing,2021,14:4816-4831. [3]GAO C,FENG D J,HU J L,et al.Collapse and landslide extraction from remote sensing image based on improved feature pyramid network[J].Science of Surveying and Mapping,2021,46(11):32-38,46. [4]HU B.Land dynamic change monitoring based on remote sen-sing image processing[J].Beijing Surveying and Mapping,2021,35(1):70-73. [5]CHEN F,ZHANG J,ZENG B.Habitation change detectionbased on image algebra and effectiveness evaluation using ZY-3 satellite image[J].Bulletin of Surveying and Mapping,2015(5):38-41. [6]ZHU E Z,SONG W D,DAI J G.Road extraction of high-resolution remote sensing image based on improved SVM[J].Science of Surveying and Mapping,2016,41(12):224-228. [7]WANG H,DAN X F,LI Z Y,et al.Extraction of winter leisure fields in Jingmen city based on random forest[J].Science of Surveying and Mapping,2020,45(5):101-105,108. [8]LONG J,SHELHAMER E,DARRELL T.Fully convolutionnetworks for semantic segmentation[C]//2015 IEEE Confe-rence on Computer.2015. [9]RONNEBERGER O,FISCHER P,BROX T.U-net:convolution networks for biomedical image segmentation[M]//Lecture Notes in Computer Science.Cham:Springer International Publishing.2015:234-241. [10]RONNEBERGER O,FISCHER P,BROX T.U-net:convolution networks for biomedical image segmentation[M]//Lecture Notes in Computer Science.Cham:Springer International Publishing.2015:234-241. [11]CHEN L C,PAPANDREOU G,KOKKINOS I,et al.Semantic image segmentation with deep convolutional nets and fully connected CRFs[J].Computer Science,2014(4):357-361. [12]HE K,GKIOXARI G,DOLLAR P,et al.Mask R-CNN[J].IEEE Transaction on Pattern Analysis and Machine Intelligence,2020,42(2):386-397. [13]JI S P,WEI S Q.Building extraction via convolutional neural networks from an open remote sensing building dataset[J].Acta Geodaetica et Cartographica Sinica,2019,48(4):448-459. [14]HAN X,HAN L,LI L Z,et al.High-resolution remote sensingimage building change detection based on deep learning[J/OL].Laser and Optoelectronics Progress:1-14.[2022-04-20].http://kns.cnki.net/kcms/detail/31.1690.TN.20210823.1125.002.html. [15]WANG M C,ZHU C Y,CHEN X Y,et al.Building change detection high resolution remote sensing images based on FPN Res-Unet[J].Journal of Jilin University(Earth Science Edition),2021,51(1):296-306. [16]XIANG Y,ZHAO Y D,DONG J H.Remote sensing image mi-ning area change detection based on improvedUNet siamese network[J].Journal of China Coal Society,2019,44(12):3773-3780. [17]JI S P,TIAN S Q,ZHANG C.Urban land cover classification and change detection using fullyatrous convolutional neural network[J].Geomatics and Information Science of Wuhan University,2020,45(2):233-241. [18]WOO S,PARK J,LEE J Y,et al.CBAM:convolutional block attention module[C]//European Conference on Computer Vision.2018:3-19. |
|