Recently a novel active contour model referred to as fuzzy-based active contour model embedded with edge detector shows its good performance in the segmentation of images, however, it cannot be applied to Synthetic Aperture Radar (SAR) images directly due to speckle noise. In this model, the edge detector is not suited to detect boundary and the used intensity average information cannot distinguish different regions in SAR images. To this end, this paper proposes a modified fuzzy active contour model that incorporates edge and texture information for SAR image segmentation. First, number of false alarm defined in the improved line segment detector is modified as edge detection operator. Second, texture information obtained based on the textural image attained by Gabor filter is used instead of intensity average information. Third, edge and texture information obtained by the abovementioned operators is embedded in the energy functional related to the fuzzy active contour model and the segmentation is then achieved by minimizing this functional using a numerical iteration process. In the experiment, the segmentation results qualitative and quantitative validate the effectiveness of the proposed model.

Semantic change detection (SCD) based on remote sensing images can provide an effective solution for large-scale land use monitoring. The existing change detection (CD) networks face limitations due to their limited receptive fields, which make it difficult to provide a comprehensive and consistent response to changes in similar large objects. Moreover, these limitations often cause the network to ignore weak changes localized in the images. To address these limitations, we propose change-guided similarity pyramid network (CG-SPNet), a decoupled multitask architecture that integrates three key components, including the multiscale positional similarity pyramid (MSSP), the symmetry-enhanced fusion CD unit (SFCD), and the change-guided feature interaction module (CGM). MSSP module captures positional correlation of semantic features at multiple scales by introducing an attention pyramid during pooling downsampling. SFCD utilizes a convolutional attention mechanism to enhance local features and highlight weak changes in dual images, and the symmetric fusion (SF) approach reduces the network’s sensitivity to timing. CGM is used to address the challenge of effective feature interaction, which combines cosine similarity loss to leverages cross-attention to compute the positional relevance of change features to embed a priori information for semantic features. Through rigorous experiments and analyses, we have successfully validated the essential components of CG-SPNet. It achieves the state-of-the-art performance on the SECOND dataset as well as our self-built CZWZ dataset.

The completeness of road extraction is very important for road application. However, existing deep learning (DP) methods of extraction often generate fragmented results. The prime reason is that DP-based road extraction methods use square kernel convolution, which is challenging to learn long range contextual relationships of roads. The road often produce fractures in the local interference area. Besides, the quality of extraction results will be subjected to the resolution of remote sensing (RS) image. Generally, an algorithm will produce worse fragmentation when the used data differs from the resolution of the training set. To address these issues, we propose a novel road extraction framework for RS images, named the Multi-Scale and Multi-Direction Feature Fusion Network (MSMDFF-Net). This framework comprises three main components: the Multi-Directional Feature Fusion (MDFF) Initial Block, the Multi-Scale Residual (MSR) encoder, and the Multi-Directional Combined Fusion (MDCF) decoder. Firstly, according to the road’s morphological characteristics, we develop a strip convolution module with a direction parameter (SCM-D). Then, to make the extracted result more complete, four SCM-D with different directions are used to MDFF-Initial Block and MDCF-decoder. Finally, we incorporate an additional branch into the ResNet encoding module to build MSR-encoder for improving the generalization of the model on different resolution RS image. Extensive experiments on three popular datasets with different resolution (Massachusetts, DeepGlobe, and SpaceNet datasets) show that the proposed MSMDFF-Net achieves new state-of-the-art results. The code will be available at https://github.com/wycloveinfall/MSMDFF-NET.

Correction of polarimetric synthetic aperture radar (PoSAR) data obtained from steep terrains is critical in SAR applications. Polarization rotation angle (POA) decides the data correction accuracy while the accurate estimation of it is challenging over forested regions. Aiming at this problem, this paper proposes a SAR data correction method for the region of mountainous forests based on a six-component scattering decomposition technique. First, POA is roughly estimated using orthogonal circular polarization method. Second, the six-component scattering decomposition technique is carried out to obtain helix scattering, oriented dipole scattering, and compound dipole scattering. Third, POA is precisely estimated according to the values of obtained scattering powers and co-herency matrix elements. Finally, SAR data is corrected using the estimated POA. The experiment carried out on a fully Pol-SAR image obtained from the region of mountainous forest verifies the effectiveness of the proposed method.

In this paper, the decomposition and reconstruction principle - two scale model (DRP-TSM) is firstly proposed to analyze microwave scattering of burned ground surface. According to the different states of ground surface before and after forest fires, we use the microwave scattering model to study the calculation of electromagnetic process of burned ground surface. Meanwhile, multi-band scattering coefficients of different polarization are measured to exploiting the temporal and spatial responses of forest fires effect on ground surface. Simultaneously, the small perturbation model is constructed in the DRP-TSM, which described as the combination of large-scale roughness and small-scale roughness for the accurate simulation of burned ground surface characteristics. Finally, the field experiments based on ground-based radar scatterometer (GBRS) are carried out and the observation is used for validation.

The performance of most supervised classification methods for synthetic aperture radar (SAR) images is largely tired to the number of samples, while labeled samples are usually very difficult and costly to obtain in the remotely sensing field. Semi-supervised methods, which are achieved by using pseudo samples, have thus been proposed to deal with this problem. Most of these methods, however, cannot be directly applied to fully polarization SAR (PolSAR) images due to the complexity of PolSAR data and it is also inefficient. To this end, this paper proposes a fast algorithm for the generation of pseudo samples of fully PolSAR data used in the classification, which is developed based on the complex Wishart distribution and chaotic maps. First, a weighting parameter that estimates the importance degree of labeled samples is defined in terms of the complex Wishart distribution. Second, chaotic maps are introduced to randomly and non-repeatedly select proper samples from the labeled sample set. Third, based on the selected samples with the weighting parameter, pseudo samples are generated. Finally, combined with appropriate classifiers, classification is attained. The experiment carried out on a fully PolSAR image verifies the effectiveness of the proposed algorithm.

Radiometric correction is a necessary pre-processing step in synthetic aperture radar (SAR) applications. In this step, geometric distorted areas caused by side-looking imaging system of SAR are the main reason for incorrect SAR data. This case is worse for rugged terrain since three different distorted phenomena including shortening, overlay, and shadowing can be observed. Generally, these three phenomena are roughly identified based on echo time of signal. However, many applications show that this method is not accurate that may affect subsequent data processing. To this end, this paper proposes a fast identification algorithm for the abovementioned three geometric distorted areas, which is attained based on the geometry model by using digital elevation model (DEM) data. In the proposed method, shadowing area is firstly identified by the defined look-up table in terms of geometry. Active overlay areas are then identified based on local incident angles. Passive overlay areas are finally identified using the traversing method. The experiment shows the efficiency and the effectiveness of the proposed method.

Slope stability models developed based on the physical mechanism of landslides show the effectiveness in landslide probability assessment, while they have rarely been applied in the field of radar remote sensing. Inspired by the related work, this article proposes a new quantitative method for rainfall-induced landslide probability assessment based on safety factors (SFs) using soil moisture estimation from synthetic aperture radar (SAR) images. In order to combine slope stability models with SAR measurement, first, soil moisture that plays a vital role in slope stability models is estimated by SAR techniques from vegetated slope terrain. In this article, we propose a new SAR data processing model for potential landslide areas and a modified physical-based scattering model for short vegetation. The estimated results are qualitatively verified by the tropical rainfall measuring mission (TRMM) instrument and are quantitatively verified by the field investigation. Second, we study the water table level that plays another vital role in slope stability models and cannot be retrieved from SAR data. The analysis indicates that it can be treated as a constant in the case of unsaturated soil moisture. Combining with other geotechnical parameters that do not change with external circumstances, we simplify the slope stability model, of which effectiveness is tested by the visual interpretation. Finally, the SF maps are obtained by the simplified slope stability model using soil moisture estimated from the corresponding SAR images. The field investigation shows that all the observed landslides are located in the unstable areas, indirectly verifying the proposed method.

In the previous work, the presented multi-texture-based segmentation model based on level sets for synthetic aperture radar (SAR) images generally can attain good results. However, the weighting parameters in this model have non-ignorable influences on segmentation performance, while such parameters usually are determined by the past experiences, which decreases the flexibility of this method. To address this problem, based on this model, we propose a SAR image segmentation method with the optimal level sets using chaotic whale optimization algorithm (CWOA). First, various image segmentation results are attained by the multi-texture-based model with several random sets of weighting parameters, samples are then automatically generated by comparing these segmentation results. Second, search agents (humpback whales) are defined with respect to the weighting parameters that need to be optimized, and the fitness function (prey) is associated with the samples. The optimal level sets are then established by integrating the multi-texture-based model and CWOA. Finally, the experimental result achieved from a SAR image shows the effectiveness of the proposed method.

Dual-polarization synthetic aperture radar (SAR) data acquired over steep terrain topography is commonly calibrated using the correction methods for single-polarization intensity SAR data. Such kinds of correction methods are generally attained by the illuminated area normalization based on terrain geometry while the case of polarization rotation induced by steep terrain is ignored. To address this issue, this paper proposes a correction method for dual-polarization SAR data over steep terrain, in which both effects of geometry and polarization rotation are taken into account. In this method, correction is achieved based on the original definition of backscattering coefficient by using only intensity SAR values obtained from two channels (co- and cross-polarization) without involving any other amplitude or phase information. The experiments on a set of dual-polarization data correction show that the corrected results generally agree with the facts of relation among SAR data distortion, terrain geometry, and polarization rotation with respect to azimuth and range slope angles of radar, which verifies the effectiveness of the proposed correction method.

This article proposes an unsupervised multiregion segmentation method for fully polarimetric synthetic aperture radar (polSAR) images based on the improved fuzzy active contour model. Different from most of the active contour models that are based on the utilization of only statistical information, the proposed method makes better use of information from polarimetric data. In addition to the statistical information, an edge detector modified from the ratio of exponentially weighted averages (ROEWA) operator, a sliding window algorithm for the total received power, and a ratio operator with respect to scattering mechanisms are integrated to the proposed active contour model. We then present a layer-based fuzzy active contour framework to solve our model. The general fuzzy active contour framework is computationally much more efficient compared with the level set-based framework; however, it cannot be applied to the multiregion segmentation of SAR images due to its low robustness to strong noise. The proposed approach includes the advantages of the general fuzzy active contour framework and has good robustness. Using two fully polSAR images demonstrates that the proposed method can achieve higher efficiency and a better segmentation performance in comparison with the commonly used active contour methods.

The objective of this article is to investigate a multiphase segmentation framework for synthetic aperture radar (SAR) images, which is proposed based on the idea of the fuzzy region competition-based method. The fuzzy region competition-based framework is highly efficient and can attain good segmentation performances for conventional images. The framework is achieved based on its convexity, which not only ensures the existence of a globally optimized solution but also enables the convex optimization theory-based solving algorithms that are feasible. However, the constraint conditions of the framework that guarantee this convexity probably cannot be satisfied in the segmentation of images corrupted with strong noise. Therefore, applying this method to an SAR image probably produces an unsatisfactory segmentation result. To address this problem, we propose an improved fuzzy region competition-based framework in terms of the hierarchical strategy, such that the framework is always convex during the iterative calculation. The proposed framework inherits the advantages of the fuzzy region competition-based method, as well as that it is able to be applied to the segmentation of images with strong noise. Several experiments are then carried out to test and verify the performance and the robustness of the proposed framework. It demonstrates that the proposed segmentation framework can be applied to various types of SAR images and achieves satisfactory segmentation results.

One of the greatest challenges in automated driving is the ability to acquire, access and query the data pertaining to high resolution 3D objects from multiple heterogeneous sources. Specifically, the information extraction needs to be done by fusing data from both sensors and databases, and with real-time constraints. Existing structures and algorithmic approaches designed for regular maps - or even regular features in High Definition maps - are not optimal to handle the various challenges. In this paper, we review the importance and roles of high resolution 3D objects in High Definition maps being used in autonomous driving applications and summarize the characteristics of 3D objects compared to other regular map features. We also describe an end-to-end pipeline of a system targeting such problems and emphasize the challenges and feasible solutions to each part of the pipeline. Last but not least, we define the quantified evaluation metrics for each task and introduce the dataset that we built for this objective.

Vegetation monitoring is important in earth science. It plays an essential role in biomass estimation, soil moisture retrieval, irrigation planning, and it also can help us better understand climate change, among others. Sentinel-1 data acquired since 2014 with a particularly short revisit period is well suited to monitor vegetation growth. In this paper, we analyze the temporal behavior of backscattering coefficients as well as the cross-pol ratio (VH/VV) of an agricultural area and a forest obtained from Sentinel-1A data. Normalized Difference Vegetation Index (NDVI) data derived from Landsat 8 data is used to roughly differentiate vegetated and non-vegetated areas. This study verifies the feasibility of Sentinel-1 data to vegetation monitoring, and gives the physical interpretation of the observed experimental results, which provides a useful information for the next phase of radar data applications, such as image classification.

More and more malicious apps and mobile rootkits are found to perform sensitive operations on behalf of legitimate users without their awareness. Malware does so by either forging user inputs or tricking users into making unintended requests to online service providers. Such malware is hard to detect and generates large revenues for cybercriminals, which is often used for committing ad/click frauds, faking reviews/ratings, promoting people or business on social networks, etc.

Synthetic Aperture Radar (SAR) image segmentation is a difficult problem due to the presence of strong multiplicative noise. To attain multi-region segmentation for SAR images, this paper presents a parametric segmentation method based on the multi-texture model with level sets. Segmentation is achieved by solving level set functions obtained from minimizing the proposed energy functional. To fully utilize image information, edge feature and region information are both included in the energy functional. For the need of level set evolution, the Ratio of Exponentially Weighted Averages (ROEWA) operator is modified to obtain edge feature. Region information is obtained by the Improved Edgeworth Series Expansion (IESE), which can adaptively model a SAR image distribution with respect to various kinds of regions. The performance of the proposed method is verified by three high resolution SAR images. The experimental results demonstrate that SAR images can be segmented into multiple regions accurately without any speckle pre-processing steps by the proposed method.

This paper presents an unsupervised segmentation method based on the variational model for fully polarimetric Synthetic Aperture Radar (PolSAR) images. Considering that fully PolSAR images contain much more information than optical or single-channel SAR images, we used the characteristics vector of PolSAR images instead of statistical parametric models in the variational model. To fully utilize characteristics information, we propose a ratio operator with respect to scattering mechanisms and a sliding window algorithm for total received power. Combining these two operators with statistical information, the variational model with respect to an energy functional is defined and the segmentation is then achieved by solving such functional in terms of fuzzy membership functions and dual projection method. The experimental results indicate that the proposed method can attain a better segmentation compared with the classical cluster algorithm based on the complex Wishart distribution and the variation model only using statistical information.

A classifier based on the Least Square Support Vector Machine (LS-SVM) with Fruit Fly Optimization Algorithm (FOA) for polarimetirc Synthetic Aperture Radar (SAR) image classification is proposed in this paper. This method uses pixel-based information and region-based information as the features of land cover. The former one comes from the integration of multiple polarimetric parameters obtained by various polarimetric decomposition techniques, and the latter one is derived from the Grey Level Co-occurrence Matrix (GLCM). Kernel Principal Component Analysis (KPCA) is afterwards used to reduce the dimensionality of the multi-feature data. Additionally, this method uses LS-SVM as the classifier. Due to the fact that the classification performance is dependent on the input parameters of LS-SVM, FOA is adopted to obtain the optimized input parameters. Finally, compared with the method without using FOA and the supervised Wishart method, the classification performance of a fully polarimetric SAR image is much better by using the proposed method.

This paper presents a landslide prediction method based on soil moisture estimation obtained by fully polarimetric Synthetic Aperture Radar (SAR) data and surface topography acquired by Shuttle Radar Topographic Mission (SRTM). In order to solve the problem of geometric distortion caused by topography, the study area is classified as measurable and non-measureable areas in terms of the terrain slope with respect to the SAR flight path. The polarimetric backscattering coefficients are corrected through a polarization transformation that depends on the direction of the unit normal for each image pixel. Areas of tall vegetation are excluded. A radiative transfer model for short vegetation is used for soil moisture estimation assuming the surface roughness is a fixed parameter. A soil stability model with soil moisture and slope as a parameter is employed to predict landslide. Finally, the model is applied to the Radarsat-2 images acquired from Maoxian, China, and the extracted soil moisture data is compared with Tropical Rainfall Measuring Mission (TRMM) data for validation. The soil stability model is then used for determination of areas for high possibility of landslide.

Persistent scatterers synthetic aperture radar interferometry (PSI) is a powerful remote sensing technique to detect the subsidence and landslides with an accuracy of millimeters. Distributed scatterers (DS) can be extracted to increase the measurement points with preserved Persistent scatterers (PSs), especially in non-urban areas covered with vegetation. In the experiment, we selected a set of SAR images of Radarsat-2 satellite, covered Mao country area, to detect the subsidence during the six months. An improved method has been presented to optimize measurement points for applying the InSAR technique to monitor the deformation of mountain areas through processing small scales full-polarization SAR Images.

Flood is one of serious natural disasters in the word. Synthetic aperture radar (SAR) has become a popular tool to detect the flood disaster for its distinct benefits such as retrieval of surface information, penetrability, and availability in all weathers. This paper aims to analyze microwave scattering characteristics of soil from low water content to semi-flooded status based on ground-scatterometer radar measurement. The research can demonstrate the scattering characteristic of soil at different status and be helpful to retrieve and monitor flood areas in the disaster. A regressive model combined with the radar data and the submerged proportion of soil has been presented.