Synthetic aperture radar (SAR) and optical image fusion can leverage the strength of both data sources for improved ground object identification. However, existing fusion methods often encounter challenges such as spatial information loss and spectral distortion, and few studies address scenarios where fused images struggle to distinguish between distinct ground objects simultaneously, such as crops at different growth stages and low woodland. To address these challenges, we propose a hybrid fusion method for Sentinel-1/2 images and construct a temporal phenology feature for improved crop mapping. To efficiently preserve spatial and spectral information, our method combines principal component analysis (PCA) transform, and utilizes convolution sparse representation (CSR) and generative adversarial network (GAN) to fuse the high- and low-frequency coefficients of nonsubsampled shearlet transform (NSST), respectively. Furthermore, leveraging the annual normalized difference vegetation index (NDVI) curve to capture the crop growth cycle, we analyze ground object change patterns across a growth cycle and construct temporal phenology feature based on fused images to enhance crop classification accuracy. Results demonstrate that our method outperforms six classical methods and five state-of-the-art deep learning methods across both subjective and objective evaluation performance. When compared with independent SAR, optical, and fused images, the combination of fused images and temporal phenology feature improves the distinction between cropland and woodland and achieves the highest classification accuracy. We also expand the application scenarios to a total of twelve agricultural regions in China, Canada, the United States and France, yielding satisfactory results.

Human intelligence can retrieve any person according to both visual and language descriptions. However, the current computer vision community studies specific person re-identification (ReID) tasks in different scenarios separately, which limits the applications in the real world. This paper strives to resolve this problem by proposing a new instruct-ReID task that requires the model to retrieve images according to the given image or language instructions. Our instruct-ReID is a more general ReID setting, where existing 6 ReID tasks can be viewed as special cases by designing different instructions. We propose a large-scale OmniReID benchmark and an adaptive triplet loss as a baseline method to facilitate research in this new setting. Experimental results show that the proposed multi-purpose ReID model, trained on our OmniReID benchmark without finetuning, can improve +0.5%, +0.6%, +7.7% mAP on Market1501, MSMT17, CUHK03 for traditional ReID, +6.4%, +7.1%, +11.2% mAP on PRCC, VC-Clothes, LTCC for clothes-changing ReID, +11.7% mAP on COCAS+ real2 for clothes template based clothes-changing ReID when using only RGB images, +24.9% mAP on COCAS+ real2 for our newly defined language-instructed ReID, +4.3% on LLCM for visible-infrared ReID, +2.6% on CUHK-PEDES for text-to-image ReID. The datasets, the model, and code are available at https://github.com/hwz-zju/Instruct-ReID.

Accurately predicting large-scale surface displacements is a vital task in the prevention and control of geological hazards in mining areas. Small baseline subset interferometric synthetic aperture radar (SBAS-InSAR) faces challenges in capturing significant deformations, and implementing extensive global navigation satellite system (GNSS) monitoring can only obtain a single point of deformation. To combine the strengths of both techniques, we proposed a novel approach named the LL-SIG (Integrating single SBAS-InSAR pair, GNSS and Log-logistic model), which leverages the understanding of mining subsidence patterns and integrates single-orbit InSAR data with limited GNSS data to achieve overall and high-gradient three-dimensional (3-D) deformation monitoring and prediction in mining areas. Firstly, we employed empirical Bayesian kriging interpolation to integrate GNSS and SBAS-InSAR data at a consistent time interval. This allowed us to obtain an overall large-gradient deformation pattern of the mining area based on its subsidence characteristics. Subsequently, we combined the Log-logistic model with the SIP (single InSAR pair) model to calculate the 3-D deformation resulting from mining activities at any moment. The proposed method was tested in the Fenxi Ruitai mine. The root mean square error (RMSE) of the fitted deformation in the line-of-sight (LOS) direction was within 10% of the maximum deformation. Compared to the Logistic model and Gompertz model, the Log-logistic model show superior accuracy and efficiency in mine deformation prediction under the same testing conditions. The RMSE of deformation for this method was 3.20 cm, 5.85 cm, and 3.73 cm in the vertical, northward, and eastward directions, respectively.

Satisfactory extraction of buildings from aerial images has long been a challenging task. In the recently fully convolutional network (FCN)-based methods, the locality of the convolution operation is detrimental to handling global long-range dependencies for complex buildings, such as those shaded by trees, obscured by the shadow of high-rise ones, and blurred by the high-similarity pixels. Although graph neural networks (GNN) show clear-cut advantages in modeling semantic correlations between different segments or instances, existing FCN-GNN methods cannot directly perform graph reasoning for spatial features in an end-to-end framework due to two chief limitations: 1) it is costly to construct large fully-connected graphs; 2) there is not a standard pixel-based graph reasoning paradigm. We therefore developed a hybrid end-to-end FCN-like network to perform better building extraction on complex scenes, termed the Dual Spatial-Graph Refinement Network (DSRNet). Specifically, we first proposed a spatial-graph reasoning module (SR), which effectively constructs adjacent relations for dense pixels, to make fast graph reasoning for grid-based features possible. Considering SR as the basis, we further developed a dual spatial-graph refinement module (DSR), consisting of body and structure SRs (BSR and SSR), to make SR attentively perceive global buildings’ spatial-semantic relationships from two complementary perspectives. BSR was designed to enhance consistency within buildings and SSR models correlations of buildings’ structural information, to extract buildings with more coherent bodies and tight-fitting edge contours. Finally, we devised a Contour Alignment loss function (CA) to encourage the segmentation result to align correctly with the contours of the ground truths. DSRNet outperforms state-of-the-art FCN-based methods on Christchurch and Tokyo high-resolution building datasets and consistently shows improvements in building segmentation and contour extraction, especially in the case of complex scenes.

Cropland classification can be used to monitor cropland distribution and its change over time. In this letter, a new superpixel-based cropland classification method is proposed for synthetic aperture radar (SAR) imagery through the integration of statistical texture, polarization, and spatial information. First, the method combines random forest algorithm and superpixels, which are generated using simple linear iterative clustering algorithm with polarization features of Pauli decomposition and spatial information. Superpixel-based spatial context information is used to reduce the influence of coherent speckle and misclassification in cropland blocks. Second,$G^{0}$statistical texture feature is used to reduce the interference of background targets such as woodland in cropland classification. Comparison experiments of different methods using C-band airborne SAR (AIRSAR) polarimetric data acquired in early July show that the proposed method has better classification performance, with an overall accuracy of 88.62%. The classification accuracy of corn and soybean is above 95% and 91%, respectively. The$G^{0}$statistical texture feature is helpful to eliminate woodland that may cause crop misclassification using single-date SAR image.

The freeze/thaw (F/T) state of the soil is an essential indicator for permafrost monitoring. However, current soil F/T products with a coarse spatial resolution (>1 km) have limited their use on a fine scale. In this letter, a new approach integrating two microwave sensors [i.e., Sentinel-1 and advanced microwave scanning radiometer 2 (AMSR-2)] is developed to identify the soil F/T state at a spatial resolution of 10 m in the Qinghai-Tibet engineering corridor (QTEC). Using a linear regression model to integrate the coarse AMSR-2 data with the finer Moderate Resolution Imaging Spectroradiometer (MODIS) land surface temperature (LST), the frozen frequency product at a 1-km resolution can be obtained. Then, the high-spatial-resolution F/T map based on Sentinel-1 synthetic aperture radar (SAR) time-series images can be produced using the threshold extracted from each pixel of frozen-frequency products. We tested soil F/T results via both visual and quantitative evaluations. The overall accuracy of the 10-m soil F/T map achieves 84.63% and 77.09% for ascending and descending orbits based on four meteorological stations, respectively.

In order to detect power transmission tower efficiently in synthetic aperture radar (SAR) images, a new constant false alarm rate (CFAR) detection algorithm based on integrated superpixel window and adaptive statistical model has been proposed. Firstly, the SEEDS algorithm is used to extract superpixels from the SAR image, and the segmented objects are merged into CFAR detection windows. Subsequently, the power transmission towers are detected by the CFAR algorithm with the adaptive statistical model in each window. The experiment results using UAVSAR images prove that the proposed algorithm can achieve good detection performance and suppress false alarms at the same time.

Polarimetric SAR classification is an effective approach in image understanding. This paper proposes a novel semantic method for classification of Polarimetric SAR data. The method combines superpixels and semantic model to benefit from both the object-oriented classification and the high-level semantic information. Firstly, pixels was grouped into superpixels via Simple Linear Iterative Clustering (SLIC). Secondly, the feature vector was generated within the superpixels by considering both polarimetric information and textures. To incorporate semantic information, the feature vectors were further processed via probabilistic Latent Semantic Analysis (pLSA). Finally, Supporting Vector Machine (SVM) was utilized to obtain classification results. The results were evaluated with respect to the accuracy of classification and spatial preservation. The results of this work were analyzed by means of RADARSAT-2 data.

It is still a challenge to obtain the collapsed building distribution from post-disaster polarimetric synthetic aperture radar (SAR) data. This paper proposed a novel approach for extracting the spatial distribution of collapsed buildings using post-disaster RADARSAT-2 SAR data. In this method, non-building areas are removed by using eigen-values λ2+ λ3. Then, the modified Freeman decomposition which includes deorientation selectively and surface scattering characteristic parameter constraint is presented for building area. The contribution of the double-bounce component (PD/span) is used to extract the collapsed building spatial distribution. The method was tested on RADARSAT-2 fine-mode polarimetric SAR imagery from the Yushu earthquake which acquired on April 21, 2010. By comparison with other methods, the results confirm the validity of the proposed method.

To obtain accurate classification results of polarimetric SAR images in different heterogeneity areas, a novel unsupervised classification method is proposed, which combines an advanced distribution with spatial contextual information based on stochastic expectation maximization (SEM) algorithm. Specifically, the probabilities of class membership are calculated by L distribution, and a neighborhood function is defined to describe spatial contextual information. Then the probabilities of class membership are altered by the predefined neighborhood function via probabilistic label relaxation (PLR) technique. Moreover, RADARSAT-2 and EMISAR data are used to verify the effectiveness of the proposed method. The experiment results show it can accurately classify different heterogeneity areas and obtain more consistent results compared with other algorithms.

Entropy, alpha, and anisotropy (H/α̅/A) of Cloude decomposition are effective in polarimetric SAR image understanding and geophysical information inversion. As an incoherent target decomposition, the inner sample covariance matrix estimation severely affects the estimated parameters. The contradiction between details preservation and accurate parameters estimation is still a challenge task. In this article, we propose adaptive coherency matrix estimation based on local heterogeneity coefficients, and utilize it to parameters estimation of Cloude decomposition. The results were evaluated with respect to details preservation and the accuracy of parameters estimation. The results of this work were analyzed by means of AIRSAR data.

Polarimetric synthetic aperture radar (SAR) images usually contain a mixture of homogeneous and heterogeneous regions, which makes estimation of the coherency matrix a very challenging task. In this paper, we propose an adaptive coherency matrix estimation method that employs local heterogeneity coefficient and leverages the sample covariance matrix estimation to the homogeneous components and the fixed-point estimation to the heterogeneous components. Evaluations were conducted with synthetic polarimetric data and real-world SAR imagery, including UAVSAR, RADARSAT-2, and ESAR. Our experimental results demonstrated that the heterogeneity coefficient effectively characterizes the scattering property of ground objects, which enables adaptive estimation of the coherency matrix in high-resolution polarimetric SAR imagery. Our method was able to handle single- and multilook polarimetric SAR imagery gracefully. Compared with the sample covariance matrix estimator, the fixed-point estimator, and the Lee sigma filtering, our method achieved the best performance for retaining the spatial structure, suppressing speckles, and preserving polarimetric information of SAR imagery with different degrees of heterogeneity.

In this paper, we proposed an Integrated Four-component Model-based Decomposition (IFMD) based on multi-look covariance matrix integrating the selective de-orientate and the generalized volume scattering. Firstly, a generalized volume scattering model (GVSM) is adopted to substitute the volume scattering of Freeman Decomposition. Then, the Cross Polarization Correlation Coefficient is adopted as a criterion to judge whether the target is reflection symmetric and we only de-orientate those non-reflection symmetric targets. For targets still being non-reflection symmetric after the de-orientation, the helix scattering is exploit to compensate the non-reflection symmetric components. Finally, the power constrain procedures are added to eliminate the negative pixels completely. The effectiveness of the IFMD is demonstrated by the L band airborne E-SAR data.