Synthetic aperture radar (SAR) ship detection is an important part of marine monitoring. With the development in computer vision, deep learning has been used for SAR images such as faster region-based convolutional neural network (R-CNN), single-shot multibox detector, and densely connected network. In field, much better performance than traditional methods on nearshore areas. This because need sea-land segmentation before detection, inaccurate mask decreases its performance. Though current...

<para xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> An adaptive and fast constant false alarm rate (CFAR) algorithm based on automatic censoring (AC) is proposed for target detection in high-resolution synthetic aperture radar (SAR) images. First, an global threshold selected to obtain index matrix which labels whether each pixel of the image a potential or not. Second, by using matrix, clutter environment can be determined adaptively prescreen...

With the development of deep learning (DL) and synthetic aperture radar (SAR) imaging techniques, SAR automatic target recognition has come to a breakthrough. Numerous algorithms have been proposed competitive results achieved in detecting different targets. However, due influence various sizes complex background ships, multiscale ships images is still challenging. To solve problems, novel network, called attention receptive pyramid network (ARPN), this article. ARPN two-stage detector...

Due to its great application value in the military and civilian fields, ship detection synthetic aperture radar (SAR) images has always attracted much attention. However, targets High-Resolution (HR) SAR show significant characteristics of multi-scale, arbitrary directions dense arrangement, posing enormous challenges detect ships quickly accurately. To address these issues above, a novel YOLO-based arbitrary-oriented detector using bi-directional feature fusion angular classification...

With the rapid development of earth observation technology, high-resolution synthetic aperture radar (HR SAR) imaging satellites could provide more observational information for maritime surveillance. However, there are still some problems to detect ship targets in HR SAR images due complex surroundings, defocusing, and diversity scales. In this article, an anchor-free method is proposed target detection images. First, fully convolutional one-stage object (FCOS) as base network applied...

Ship recognition in synthetic aperture radar (SAR) is an essential challenge SAR image interpretation. The measured ship targets often contain complex background such as port facilities and neighboring ships, which are easy to interfere with the model affect performance. To address this issue, a method based on dual-branch transformer fusion network proposed paper. First of all, feature extraction architecture designed paper, including significant (SFE), global (GFE), (D-BFF). Specifically,...

In this paper, the classification via sprepresentation and multitask learning is presented for target recognition in SAR image. To capture characteristics of image, a multidimensional generalization analytic signal, namely monogenic employed. The original signal can be then orthogonally decomposed into three components: 1) local amplitude; 2) phase; 3) orientation. Since components represent different kinds information, it beneficial by jointly considering them unifying framework. However,...

This paper studies the hyperspectral image (HSI) denoising problem under assumption that signal is low in rank. In this paper, a mixture of Gaussian noise and sparse considered. The includes stripes, impulse noise, dead pixels. task formulated as low-rank tensor recovery (LRTR) from noise. Traditional decomposition methods are generally NP-hard to compute. Besides, these based sensitive contrast, proposed LRTR method can preserve global structure HSIs simultaneously remove noise.The on new...

Several bandwise total variation (TV) regularized low-rank (LR)-based models have been proposed to remove mixed noise in hyperspectral images (HSIs). These methods convert high-dimensional HSI data into 2-D based on LR matrix factorization. This strategy introduces the loss of useful multiway structure information. Moreover, these TV-based exploit spatial information a separate manner. To cope with problems, we propose spatial–spectral TV tensor factorization (SSTV-LRTF) method HSIs. From...

This paper introduces a novel classification strategy based on the monogenic scale space for target recognition in Synthetic Aperture Radar (SAR) image. The proposed method exploits signal theory, multidimensional generalization of analytic signal, to capture characteristics SAR image, e.g., broad spectral information and simultaneous spatial localization. components derived from at different scales are then applied into recently developed framework, sparse representation-based (SRC)....

In this letter, the classification via sparse representation of monogenic signal is presented for target recognition in SAR images. To characterize images, which have broad spectral information yet spatial localization, performed. Then an augmented feature vector generated uniform down-sampling, normalization and concatenation components. The resulting fed into a recently developed framework, i.e., based (SRC). Specifically, vectors training samples are utilized as basis to code test sample...

Traditional two-dimensional (2D) Otsu method supposes that the sum of probabilities diagonal quadrants in 2D histogram is approximately one. This studies experiments and theory prove off-diagonal not always very small this could be neglected. Therefore assumption mentioned above inadequately reasonable. In study, an improved segmentation recursive algorithm are proposed. By calculating separately, modified acquired. Experimental results show proposed can obtain better performance than...

This work presents a robust graph mapping approach for the unsupervised heterogeneous change detection problem in remote sensing imagery. To address challenge that images cannot be directly compared due to different imaging mechanisms, we take advantage of fact share same structure information ground object, which is modality-invariant. The proposed method first constructs K-nearest neighbor represent each image, and then compares graphs within image domain by means calculate forward...

Change detection (CD) of remote sensing (RS) images is one the important problems in earth observation, which has been extensively studied recent years. However, with development RS technology, specific characteristics remotely sensed images, including sensor characteristics, resolutions, noises, and distortions imagery, make CD more complex. In this article, we propose a structure consistency-based method for CD, detects changes by comparing structures two rather than pixel values images....

Synthetic aperture radar (SAR) target recognition faces the challenge that there are very little labeled data. Although few-shot learning methods developed to extract more information from a small amount of data avoid overfitting problems, recent or limited-data SAR algorithms overlook unique imaging mechanism. Domain knowledge-powered two-stream deep network (DKTS-N) is proposed in this study, which incorporates domain knowledge related azimuth angle, amplitude, and phase vehicles, making...

Most deep-learning based target detection methods have high computational complexity and memory consumption, they are difficult to be deployed on edge devices with limited computing resources memory. To tackle this problem, paper proposes learn a lightweight detector named Light-YOLOv4, it is obtained from YOLOv4 through model compression. end, firstly, we perform sparsity training by applying L1 regularization the channel scaling factors, so less important channels layers can found....

Change detection of heterogeneous multitemporal satellite images is an important and challenging topic in remote sensing. Since the imaging mechanisms sensors are different, it not possible to directly compare detect changes as homogeneous images. To address this challenge, we propose unsupervised image regression-based change method based on structure consistency. The proposed first adaptively constructs a similarity graph represent pre-event image, then uses translate domain post-event...

Deep learning has been widely used in automatic target recognition (ATR) for synthetic aperture radar (SAR) recently. However, most of the studies are based on network structure optical images and lack full consideration inherent characteristics SAR targets, which limits improvement accuracy makes poor generalization ability. In addition, due to black-box characteristics, it is difficult effectively interpret ATR results. To conquer these problems, we propose an electromagnetic scattering...

In the context of ship monitoring in ocean, targets are usually sparsely distributed. Thus, synthetic aperture radar (SAR) imaging whole scene is quite redundant and costly. However, raw SAR echo data were considered to be useless before focusing. Few studies have attempted detect ships from data. It seems an impossible task since resolution too low. This article proposes a detection method for view nonimaging target sensing paradigm. The core idea that we can sense existence without...