With the rapid development of deep learning technology, many synthetic aperture radar (SAR) target recognition algorithms based on convolutional neural networks have achieved exceptional performance various datasets. However, conventional are repeatedly iterated a fixed dataset until convergence, and once they learn new tasks, large amount previously learned knowledge is forgotten, leading to significant decline in old tasks. This article presents an incremental method strong separability...

Synthetic aperture radar automatic target recognition (SAR-ATR) has made great progress in recent years. Most of the established methods are supervised, which have strong dependence on image labels. However, obtaining labels images is expensive and time-consuming. In this paper, we present a semi-supervised learning method that based standard deep convolutional generative adversarial networks (DCGANs). We double discriminator used DCGANs utilize two discriminators for joint training....

The deep convolution neural network (CNN), which has prominent advantages in feature learning, can learn and extract features from data automatically. Existing polarimetric synthetic aperture radar (PolSAR) image classification methods based on the CNN only consider polarization information of image, instead incorporating image’s spatial information. In this paper, a novel method dual-branch (Dual-CNN) is proposed to realize PolSAR images. built two CNNs: one used 6-channel real matrix (6Ch)...

Speckle noise is an inherent but annoying property in the synthetic aperture radar (SAR) imaging. In this paper we investigate influence of speckle on classical convolutional neural network (CNN) for SAR target classification. Then a dual stage coupled CNN architecture, named despeckling and classification CNNs (DCC-CNNs), proposed to distinguish multiple categories ground targets images with strong varying speckle. It first applies sub-network reduction. After that, residual features as...

Accurate range model with acceleration, the coupling phase terms, and spatial-variant (SV) Doppler parameters are main issues to be solved in high-speed-high-squint SAR (HSHS-SAR) a curved trajectory. For these issues, an extended Omega-K (EOK) algorithm is developed this paper. The proposed EOK mainly includes following four aspects. Firstly, modified (AMRM) considering three-dimension acceleration for trajectory established. Then, between azimuth direction removed by Stolt mapping (MSM)....

We explore a cutting-edge concept known as <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">C</i> lass Incremental Learning in xmlns:xlink="http://www.w3.org/1999/xlink">N</i> ovel Category Discovery for Synthetic Aperture Radar xmlns:xlink="http://www.w3.org/1999/xlink">T</i> argets (CNT). This innovative task involves the challenge of identifying categories within unlabeled datasets by utilizing provided labeled dataset reference. In...

In recent years, with the improvement of synthetic aperture radar (SAR) imaging resolution, it is urgent to develop methods higher accuracy and faster speed for ship detection in high-resolution SAR images. Among all kinds methods, deep-learning-based algorithms bring promising performance due end-to-end automated feature extraction. However, several challenges still exist: (1) standard deep learning detectors based on anchors have certain unsolved problems, such as tuning anchor-related...

In this paper, a 3D inverse synthetic aperture radar (ISAR) imaging method based on an antenna array configuration is proposed. The performance of conventional interferometric ISAR system using three antennas poor, as the positions scatterers, which have same range-Doppler value and projected onto plane synthesis scatterer, cannot be correctly estimated. However, by two arrays perpendicular to each other, system's ability separate these scatterers can improved. criterion for selection range...

Accurate detection of rivers plays a significant role in water conservancy construction and ecological protection, where airborne synthetic aperture radar (SAR) data have already become one the main sources. However, extracting river information from efficiently accurately still remains an open problem. The existing methods for detecting are typically based on rivers' edges, which easily mixed with those artificial buildings or farmland. In addition, pixel-based image processing approaches...

Common horizontal bounding box-based methods are not capable of accurately locating slender ship targets with arbitrary orientations in synthetic aperture radar (SAR) images. Therefore, recent years, based on oriented box (OBB) have gradually received attention from researchers. However, most the recently proposed deep learning-based for OBB detection encounter boundary discontinuity problem angle or key point regression. In order to alleviate this problem, researchers propose introduce some...

In this paper, an estimation method of micromotion parameters for free rigid targets using micro-Doppler (mD) features is investigated. These include spin rate, precession nutation angle, and inertia ratio. They represent the microdynamic characteristics intrinsic properties targets. The time variation mD frequency found complicated yet valuable to estimate parameters. From viewpoint spectra mixed time-frequency (TF) data sequences, theoretical analysis mathematical derivation are conducted...

Convolutional neural network (CNN) can be applied in synthetic aperture radar (SAR) object recognition for achieving good performance. However, it requires a large number of the labelled samples its training phase, and therefore performance could decrease dramatically when are insufficient. To solve this problem, paper, we present novel active semisupervised CNN algorithm. First, learning is used to query most informative reliable unlabelled extend initial dataset. Next, method developed by...

Deep learning techniques have attracted much attention in the radar automatic target recognition. In this paper, we investigate an acceleration method of convolutional neural network (CNN) on field-programmable gate array (FPGA) for embedded application millimeter-wave (mmW) radar-based human activity classification. Considering micro-Doppler effect caused by a person's body movements, spectrogram mmW echoes is adopted as CNN input. After that, according to architecture and properties FPGA...

As an important step of synthetic aperture radar image interpretation, segmentation aims at segmenting into different regions in terms homogeneity. Because the deficiency labeled samples and existence speckling noise, is a challenging task. We present new method for this article. Due to large size original image, we first divide input small slices. Then slices are attention-based fully convolutional network obtaining results. Finally, connected conditional random field adopted improving...

Convolutional Neural Network (CNN) has been widely applied in the field of synthetic aperture radar (SAR) image recognition. Nevertheless, CNN-based recognition methods usually encounter problem poor feature representation ability due to insufficient labeled SAR images. In addition, large inner-class variety and high cross-class similarity images pose a challenge for classification. To alleviate problems mentioned above, we propose novel few-shot learning (FSL) method recognition, which is...

There are several unresolved issues in the field of ship instance segmentation synthetic aperture radar (SAR) images. Firstly, inshore dense area, problems missed detections and mask overlap frequently occur. Secondly, scenes, false alarms occur due to strong clutter interference. In order address these issues, we propose a novel network based on dynamic key points information enhancement. detection branch network, module (DKPM) is designed incorporate target's geometric into parameters head...

Independent of daylight and weather conditions, synthetic aperture radar (SAR) images have been widely used for ship monitoring. The traditional methods SAR detection are highly dependent on the statistical models sea clutter or some predefined thresholds, generally require a multi-step operation, which results in time-consuming less robust detection. Recently, deep learning algorithms found wide applications from images. However, due to multi-resolution imaging mode complex background, it...