Search

Earth is constantly changing due to some natural events and human activities that threaten our environment. Thus, accurate and timely monitoring of these changes is of great importance for properly coping with their consequences. In this regard, this research presented a new framework for hyperspectral change detection (HCD) based on dynamic time warping (DTW) and multi-level spectral unmixing. The proposed method included two parts. The first part provided the binary change map based on Otsu and DTW algorithms. The DTW algorithm plays the role of a robust predictor for HCD purposes and the Otsu algorithm selects the threshold for detecting changeand no-changeareas. The second part presented a multiple change map based on the local spectral unmixing procedure and the output of the image differencing (ID) algorithm. The second part, at the first step, uses the ID to predict changeand no-changeareas and then employs the binary change map for mask no-changepixels. The endmember estimation and extraction was applied to changepixels, and the correlation coefficient among the bands was calculated simultaneously. Next, changepixels were divided into many parts based on the correlation among the bands. In addition, the abundance map was estimated, and then the labeling process was applied for each part. Finally, the multiple change map was obtained by the fusion of the labels of all parts. The result of HCD was compared to those of other robust HCD methods by two real bi-temporal hyperspectral datasets. Based on the result of HCD in binary and multiple change maps, the proposed method had high performance compared to other methods and its overall accuracy and kappa coefficient were more than 90% and 0.77, respectively.

AbstractThe estimation of water depth in coastal areas and shallow waters is crucial for marine management and monitoring. However, direct measurements using fieldwork methods can be costly and time-consuming. Therefore, remote sensing imagery is a promising source of geospatial information for coastal planning and development. To this end, this study investigates advanced machine learning (ML) methods and redesigned morphological profiles for water depth estimation using high-resolution Sentinel-2 satellite imagery. The proposed framework involves three main steps: (1) morphological feature generation, (2) model training using several ML methods (Decision Tree, Random Forest, eXtreme Gradient BOOSTing, Light Gradient Boosting Machine, Deep Neural Network, and CatBoost), and (3) model interpretation using eXplainable Artificial Intelligence (XAI). The performance of the proposed method was evaluated in two different coastal areas (port and jetty) with reference data from accurate hydrographic data (Echo-sounder and differential global positioning systems). The statistical analysis revealed that the proposed method had high efficiency for depth estimation of the coastal area, achieving a best R2value of 0.96 and Root Mean Square Error (RMSE) of 0.27 m in water depth estimation in the shallow water of Chabahar Bay in the Oman Sea. Additionally, the higher impact and interaction of the morphological features were verified using XAI for water depth mapping.

AbstractOne of the most important applications of remote sensing is change detection (CD). The accurate detection of changes is of great significance for the optimal management of available resources. This article presents an unsupervised ‘multiple-change detection’ method using multi-temporal hyperspectral imaging based on the integration of an unmixing technique, multi-resolution segmentation, similarity measure methods, and the Otsu algorithm. The proposed method is presented in the context of two main scenarios: the first scenario is hyperspectral change detection (HSCD) at the sub-pixel level with no ancillary data; and the second is the HSCD at the sub-pixel level based on ancillary data (high-resolution panchromatic (PAN) data). The main advantages of the proposed method are that it is unsupervised, easy to use, providing ‘multiple-change’ maps at the sub-pixel level with high accuracy. To evaluate the performance of the proposed method, real bi-temporal hyperspectral Hyperion images and high spatial resolution of PAN data belonging to the Advanced Land Imager (ALI) sensor with a variety of land cover classes were used. The results show the overall accuracy improved by >5%, and kappa coefficient improved by >0.13 with respect to the results obtained from their original resolution.