Automotive synthetic aperture radar (SAR) systems are rapidly emerging as a candidate technological solution to enable a high-resolution environment awareness for autonomous driving. This paper deals with the SAR imaging of a low-speed target using a Multiple-Input Multiple-Output (MIMO) Frequency Modulated Continuous Wave (FMCW) radar platform, which is quite common in the general vehicle automatic driving systems. The low-resolution SAR image is constructed with time domain back projection. And the final SAR image is obtained by coherently summing all the complex-valued low-resolution MIMO images along the synthetic aperture. We analyze the influence of target slow motion on SAR image quality with simulated data. And field experiments have shown that the proposed process can effectively realize the SAR perception of low speed targets.

Noise reduction is a critical research area in current remote sensing image processing. Existing denoising techniques for remote sensing images often encounter challenges such as blurred edges and excessive smoothing. To overcome these limitations, we propose a novel hierarchical denoising model based on an Autoencoder. Our model effectively addresses the issue of distinguishing low-rank residuals and preserving essential details in remote sensing images, while also extracting edge features from the residuals with high efficiency. To validate the effectiveness of our approach, we conduct comprehensive experimental tests on a representative remote sensing dataset. The results demonstrate that our method successfully preserves edge details while achieving superior denoising performance compared to state-of-the-art techniques.

Traditional low-rank methods overlook residuals as corruptions, but we discovered that low-rank residuals actually keep image edges together with corrupt components. Therefore, filtering out such structural information could hamper the discriminative details in images, especially in heavy corruptions. In order to address this limitation, this paper proposes a novel method named ESL-LRR, which preserves image edges by finding image projections from low-rank residuals. Specifically, our approach is built in a manifold learning framework where residuals are regarded as another view of image data. Edge preserved image projections are then pursued using a dynamic affinity graph regularization to capture the more accurate similarity between residuals while suppressing the influence of corrupt ones. With this adaptive approach, the proposed method can also find image intrinsic low-rank representation, and much discriminative edge preserved projections. As a result, a new classification strategy is introduced, aligning both modalities to enhance accuracy. Experiments are conducted on several benchmark image datasets, including MNIST, LFW, and COIL100. The results show that the proposed method has clear advantages over compared state-of-the-art (SOTA) methods, such as Low-Rank Embedding (LRE), Low-Rank Preserving Projection via Graph Regularized Reconstruction (LRPP_GRR), and Feature Selective Projection (FSP) with more than 2% improvement, particularly in corrupted cases.

This paper considers designing an optimal policy for deadline-constrained access in cognitive radio networks, where a secondary user needs to complete a packet transmission over the vacant spectrum within a delivery deadline. To minimize the total access cost, it is desirable to design an optimal opportunistic access policy by utilizing channel dynamics and sensing outcomes. We take non-negligible switching overheads, a state-dependent overtime penalty, and practical switching operations into consideration in the Markov decision process formulation of such an access problem under wide-band sensing. Moreover, we establish the existence of monotone optimal decision rules to reduce the complexity of computing an optimal policy. Simulation results verify our theoretical studies and the cost advantage over other policies.

The changes in urban built-up areas can reflect the process of urban expansion and have significant implications for evaluating sustainable urban development. The use of remote sensing to detect building areas is an important technology for land monitoring and urban management. Compared with optical images, SAR is less affected by interference and is an important data source for extracting building information in cloudy and rainy areas. Currently, high-precision automated extraction of built-up areas from SAR images remains a challenge. To address this issue, this paper proposes a semantic segmentation framework for building area extraction based on HRNet. Using multiple temporal SAR images acquired from Hainan Province as an example, the paper analyzes changes in the built-up areas of the province by combining the results of built-up area extraction over several years.

In inverse synthetic aperture radar (ISAR) imaging of maneuvering targets, traditional high-order phase compensation methods are usually limited by the heavy computational burden of parameter estimation, making it hard to apply in real-time cases. In this letter, an efficient ISAR motion compensation method is proposed based on fast parameter estimation. A novel analytic expression of the image entropy is derived with the help of a compensation matrix based on sinc function interpolation, which can eliminate the high-order phase term. Hence, the parameter estimation is transformed to an optimization problem where the gradient descent algorithm can be adopted to accelerate the computation speed. Compared with other recently proposed methods, our method is superior in high robustness and low computing cost. Experiments with simulated and real data have verified that without affecting the image quality, the computing time of our method can be reduced to about 1/4 of the traditional search algorithm.

Synthetic aperture radar (SAR) image despeckling is a challenging task as speckle noise is spatially correlated and signal-dependent, and appears as a grainy texture superimposed on images. Although traditional low-rank SAR image despeckling methods have shown promising performance, they have the problem of producing over-smoothed images with blurred edges due to their low-rank characteristics. In this paper, we propose a novel edge preserved SAR despeckling method named EP-LRSID, which can keep rich edge details while reducing speckle noise. Specifically, EP-LRSID takes a fresh look at the low-rank model, i.e., we can obtain structural edge information from residuals which is viewed as noise and simply disregarded by the traditional low-rank methods. To obtain discriminative edge information from residuals, the edge subspace is obtained in a manifold framework by using a dynamic affinity graph regularization. Moreover, a new hierarchical prior knowledge regulation is designed to make different kinds of pixels processed hierarchically, especially the strong scattering points in SAR images. By introducing this prior knowledge, our low-rank model can obtain more confidential low-rank parts and edge parts, thus structural information including edges can be better preserved in this way. Extensive experiments on several real and synthetic datasets demonstrate that EP-LRSID can achieve the highest despeckling performance with edge preservation than other state-of-the-art despeckling algorithms.

Manifold learning is a widely used technique for dimensionality reduction as it can reveal the intrinsic geometric structure of data. However, its performance decreases drastically when data samples are contaminated by heavy noise or occlusions, which leads to unsatisfying data processing performance. We propose a novel robust dimensionality reduction method via low-rank Laplacian graph learning for classification and clustering tasks to solve the above problem. First, we construct a low-rank Laplacian graph by combining manifold learning and subspace learning. This graph can capture both global and local structural information of the data. And we introduce rank constraints for the Laplacian graph to make it more discriminative. Second, we put the learning of projection matrix and sample affinity graph into a unified framework. The projection matrix is embedded into a robust low-rank Laplacian graph so that the low-dimensional mapping of data can maintain the structural information in the graph well. Finally, we add a regularization term to the projection matrix to make it have the ability of both feature extraction and feature selection. Therefore, the proposed model can resist the interference of noise or data damage to learn the optimal projection to achieve better performance in dimensionality reduction through such a data dimensionality reduction joint framework. Comprehensive experiments on various benchmark datasets with varying degrees of occlusions or corruptions are carried out to evaluate the performance of the proposed method. Compared with the state-of-the-art dimensionality reduction methods in the literature, the experimental results are inspiring, showing our method’s effectiveness and robustness in classification and clustering, especially in object recognition scenarios with noise or occlusions.

Federated learning (FL) has become a promising machine learning (ML) paradigm for training machine learning models over distributed datasets, owing to its low communication costs and privacy preserving property. To date, the most commonly adopted model fusion mechanism in FL is average aggregation. However, it has been shown that this average aggregation mechanism performs poorly in heterogeneous systems, especially for non-independent and identically distributed (NonIID) data. In order to address this challenge, we propose a weighted FL model aggregation strategy for each client based on clustering, termed CluFL. Specifically, CluFL first measures the similarities among uploaded models from clients through their parameters using a spectral clustering algorithm. Then, CluFL assigns aggregation weights according to the similarity of the intra-cluster global model for each cluster and the average model across the clusters. Further, we derive a convergence bound on the CluFL algorithm considering a practical nonconvex setting of neural network training. This bound reveals that the proposed CluFL algorithm can achieve a convergence speed in the order of O(1/T). Extensive experiments have been conducted on both FashionMNIST and CIFAR-10 datasets and show that CluFL outperforms the state-of-the-art FL algorithms in terms of accuracy and communication efficiency.

Automatic target recognition is a crucial task for SAR remote sensing. Unlike other methods, the unsupervised representation learning based on deep architecture can obtain robust high-level features directly from raw data. A drawback of most unsupervised representation learning methods in SAR ATR is that they only deal with amplitude images. In addition, many methods utlize a single layer architecture to extract pixel-level/mid-level features which are probably sensitive to condition variation. In this paper, a feature extraction method based on modified stacked convolutional denoising auto-encoder (MSCDAE) for complex SAR images is proposed, where convolutional kernels of MSCDAE are learned by 1-D modified denoising auto-encoders. By stacking the convolutional layers and pooling layers, high-level representation of objects are learned. The features are subsequently sent to a trained SVM for object classification. Experimental results demonstrate that the proposed method can provide a significant improvement in the ATR performance.

Ship classification with spaceborne high resolution synthetic aperture radar (SAR) has wide applications in maritime traffic monitoring, fishing law-enforcement operation, marine security, etc. Deep learning, which has the ability of learning features itself, is successfully used in computer vision and artificial intelligence, and introduced into remote sensing field in recent years. In this study, the Italian COSMO-SkyMed SAR images acquired on Jul. 12-15, 2010 were used for ship classification with convolution neural networks in the Google's TensorFlow environment. The results show that cargo ships could be discriminated from non-cargo ships. Due to variations of radar illumination directions and ship poses, more data are necessary for sub-category classification.

Ship detection using synthetic aperture radar (SAR) is an important application in maritime transportation monitoring. However, azimuth ambiguities are inevitably caused by the undersampling of the echo signals, which can be easily mistaken as ship targets in the detection result. To solve the problem mentioned above, a fast object-oriented method for azimuth ambiguities removal based on quantitative analysis is proposed. The main idea is to manage to find the ambiguities' accurate locations in the image based on the imaging parameters, so that these “ghosts” can be removed easily. It is shown that the displacement distances between the ship and its corresponding azimuth ambiguities can be calculated approximately. In case of removal of some small ships with weak backscattering power, the energy decay of the ship is also taken into account. After the ambiguities' location were fixed, a searching procedure was executed to remove them. The experimental results show the proposed method is able to remove azimuth ambiguities efficiently.

This letter depicts a ship detection scheme for synthetic aperture radar images, utilizing a segmentation based global iterative censoring algorithm. In the proposed scheme, the fuzzy local information c-means clustering (RFLICM) algorithm is adopted to partition the inhomogeneous SAR image into numerous homogeneous sub-regions, thereby eliminating the performance degradation caused by SAR image inhomogeneity. Subsequently, successively applying the GIC algorithm base on a parametric clutter model database to the sub-regions, the optimal clutter models and the initial outlier map of the sub-regions are generated. A sliding window CFAR detector based on the selected clutter models and the initial outlier map is utilized to detect ships in the SAR image. In our experiment, we tested the proposed method on spaceborne SAR data, and its effectiveness was successfully demonstrated.

In this letter, an improved kernel density estimation (KDE) constant false alarm rate (CFAR) method is proposed for ship detection in single polarization synthetic aperture radar (SAR) images. The proposed method consists of a target enhancement filter, an adaptive KDE bandwidth estimation method and an improved KDE-CFAR. The gravity-based target enhancement filter is utilized to remove the inhomogeneity in SAR images, and thereby meet the requirement of the KDE bandwidth estimation method. The proposed method provides an automatic training sample selection scheme, avoiding the manual intervention in conventional method. In addition, the KDE-CFAR is improved, employing the exponential function as the kernel since it provides an analytical solution for the CFAR criterion, which is unavailable for the Gaussian kernel. Experimental results with six spaceborne SAR images demonstrated that the proposed method is effective and efficient for ship detection application.

A wavelet based automatic targets detection method for high resolution airborne SAR data is described in this article to receive faster and more accuracy detection. This method is based on the assumption that man-made objects are easily detectable at low resolution because their scattering is more persistent than that of natural objects. The algorithm involves an improved wavelet soft threshold filter (IWSTF) and a wavelet based RCCFAR detector. In order to retain the target feature, the wavelet soft threshold filter is improved by the strategy used in the enhanced Lee filter. Instead of using a global threshold, we adopted an adaptive threshold calculated according to the detail coefficients in each scale. To accelerate the RCCFAR detector, two RCCFAR detectors are used. One is first applied to the approximate coefficients to make a coarse detection. The other one is applied to the filtered images in those regions which are regarded as candidate targets. Performance of the algorithm is assessed by some high resolution airborne SAR image and it shows that the algorithm can effectively reduce false alarms caused by speckles.