Inspired by vision foundation models, a semantic-driven unsupervised change detection method is proposed to deal with the low accuracy of unsupervised change detection in remote sensing images. Firstly, the segment anything model (SAM) is used to perform unsupervised segmentation of remote sensing images to obtain different object regions. Secondly, the large language and vision assistant (LLaVA) is used to generate corresponding text descriptions for the object regions in two temporal remote sensing images. Thirdly, the word embedding model is further utilized to convert text descriptions into feature vectors, and the semantic similarity of the same object area in two temporal remote sensing images is calculated. According to the semantic differences of the same segmentation object area in different temporal remote sensing images, the threshold segmentation method is used to identify changing regions. Finally, the change results obtained by using remote sensing images from different temporal images as reference images are integrated to further improve the accuracy of change detection. The effectiveness of the proposed method is validated through extensive change detection experiments on two benchmark datasets, LEVIR-CD and WHU-CD.
KHELIFIL, MIGNOTTEM. Deep learning for change detection in remote sensing images: comprehensive review and meta-analysis[J]. IEEE Access, 2020,8:126385-126400.
[3]
BAIT, WANGL, YIND M, et al. Deep learning for change detection in remote sensing: a review[J]. Geo-Spatial Information Science, 2022,26(3):262-288.
[4]
蒋汾龙.基于差异表示学习的遥感影像变化检测[D].西安电子科技大学,2022:1-132.
[5]
ZHUX X, TUIAD, MOUL C, et al. Deep learning in remote sensing: a comprehensive review and list of resources[J]. IEEE Geoscience and Remote Sensing Magazine, 2018,5(4):8-36.
[6]
PARELIUSE J. A review of deep-learning methods for change detection in multispectral remote sensing images[J]. Remote Sensing, 2023,15(8):No.2092.
[7]
SHELHAMERE, LONGJ, DARELLT, et al. Fully convolutional networks for semantic segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016,39(4):640-651.
[8]
HUANGJ, SHENQ, WANGM, et al. Multiple attention siamese network for high-resolution image change detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022,60:No.5406216
[9]
KIRILLOVA, MINTUNE, RAVIN, et al. Segment anything[DB/OL]. (2023-04-05)[2024-10-15].
[10]
ZHANGY H, JIANGH, MIUURAY, et al. Contrastive learning of medical visual representations from paired images and text[DB/OL]. (2022-09-19)[2024-10-15].
[11]
CELIKT. Unsupervised change detection in satellite images using principal component analysis and k-means clustering[J]. IEEE Geoscience and Remote Sensing Letters, 2009,6(4):772-776.
[12]
AMINA M E, LIUQ J, WANGY H. Convolutional neural network features based change detection in satellite images[C]∥Proceedings of the First International Workshop on Pattern Recognition. Washington, USA: SPIE, 2016:181-186.
[13]
DUB, RUL X, WUC, et al. Unsupervised deepslow feature analysis for change detection in multitemporal remote sensing images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019,57(12):9976-9992.
[14]
TANGX, ZHANGH Y, MOUL C, et al. An unsupervised remote sensing change detection method based on multiscale graph convolutional network and metric learning[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022,60:No.5609715.