Search

A novel and effective change detection method based on manifold regularized joint non-negative matrix factorization (MJNMF) framework is proposed in this paper, which detects the changes that occurred in multi-temporal remote sensing images. Most change detection methods, including dictionary learning, principal component analysis (PCA), etc., do not consider the non-negativity among image pixels. However, image itself is a non-negative signal, and the non-negative constraint has better interpretability in practical applications. Nonnegative Matrix Factorization, which incorporates the non-negativity constraint and thus learns object parts, obtains the parts-based representation as well as enhancing the interpretability of the issue correspondingly. In this paper, our proposed approach based on MJNMF framework aims to establish a pair of joint basis matrices by unchanged training samples from unchanged area. Then, unchanged pixels can be well reconstructed by the corresponding basis matrix, while changed pixels cannot be reconstructed from the basis matrix corresponding to the knowledge of unchanged samples, or a larger reconstruction error can be generated even if changed pixels are reconfigurable. In order to suppress similar information and highlight different information, the cross-reconstruction error is used to generate the difference image. Finally, the binary image is obtained by the robust fuzzy local information c-means (FLICM) clustering algorithm. In addition, inspired by manifold learning, we incorporate manifold regularization into the proposed method to keep the geometric structure of data and improve the accuracy of change detection. Experimental results obtained on simulated and real remote sensing images confirm the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Clustering of Remote Sensing (RS) images is one of the most accurate methods for land cover Change Detection (CD). Despite its popularity for unsupervised CD, the necessity to determine the number of clusters is the main shortcoming of Fuzzy C-Means(FCM) for CD. Clustering the absolute difference image to pre-defined two clusters (includes change versus no-change), which is called binary FCM, is a conventional strategy to mitigate the cluster number problem. This strategy suffers from low CD accuracy because of its inability to detect multiple changes in RS images. In this paper, a novel framework for unsupervised CD is proposed, which is called Multiple FCM (MFCM). In this method, the optimal cluster number of MFCM is determined by analyzing the cluster-validity indices when clustering the difference image. After clustering the difference image, binarizing of the clustered image produces the Binary Change Map (BCM). The performance of the proposed approach along with two state-of-the-art binary CD methods (BFCM and PCA-Kmeans) were evaluated on two different datasets with 5 and 2 real cluster numbers. The implementation results show that the proposed approach not only detects multiple land cover changes but also improves the binary change detection accuracy. Utilizing the proposed MFCM, the total errors of BCM reduced by 20% and 24% as compared to BFCM and PCA- Kmeans methods in the first dataset with multiple land cover changes. In the second dataset, which has only two real clusters no significant difference was observed between the results of the proposed method and reference approaches. [ABSTRACT FROM AUTHOR]

Land use / land cover (LULC) has changed dramatically in recent years, especially in areas that have experienced severe climate change and population growth. Evaluation of LULC changes is the most accurate method to understand past land uses and predict their future changes. Analysis of multi-temporal optic and radar remotely sensed images is an effective tool for monitoring LULC changes, and fusion of these datasets has provided significant developments in the detection and modeling of land surface changes due to their unique features individually. Time-series image (TSI) analysis provides a history of changes in an area and allows for predicting future changes. This study aims to develop a method for modeling LULC changes and predicting future trends. To this end, a spatiotemporal regression model is proposed based on TSI. In this method, land surface changes are first modeled by considering the spatiotemporal behavioral patterns of cover changes in a given period. Then, future land covers are predicted by a spatiotemporal regression. To evaluate the proposed method, LULC changes of the Maharloo Lake (Fars province, Iran) were modeled using fused optical and radar TSI (12 consecutive image sets) and then predicted for time t = 13. Compared to ground truth data, the different land covers including; water bodies, vegetation cover, bare land, and saline area were predicted by accuracies of 84.9, 98.4, 90.7, and 97.3%, respectively, indicating the remarkable efficiency of the model proposed for predicting LULC. [ABSTRACT FROM AUTHOR]

Detection of change through remote sensing (RS) is widely used in Earth observation and environment surveys, whereas the introduction of bibliometric methods to the development, application, and identification of trends in RS change detection (RSCD) remain limited. Based on the 5,012 published academic studies in the Web of Science Core Collection (WOSCC) database between 2000 and 2022, and CiteSpace software, the publications built the RSCD knowledge mapping about cooperation network, literature co-citation, keyword co-occurrence, and burst detection analyses. The result shown that: (1) There has been a significant increasing trend in RSCD-related literature over the last two decades. Among these, Remote Sensing and IEEE journals had the highest number of publications. (2) China, the United States, and Italy, are the top three in the number of publications, mainly by various universities and research institutes. Bruzzone Lorenzo, Gong Maoguo, Bovolo Francesca, and other authors have made important contributions. (3) Among highly cited literature, the "changed object" was the first focus of CD research, and 17 research clusters were identified, including semantic, terrain correction, land cover, synthetic aperture radar, and the unsupervised. (4) Main research topics included CD models, unsupervised CD algorithms, and land cover classification. Research hots included deep learning, misregistration, image segmentation, and Google Earth Engine. This study provided the multidimensional references for researchers, practitioners, and institutions in the current trends, topics, and hots of RSCD. [ABSTRACT FROM AUTHOR]