In this letter, a new superpixel-based constant-false-alarm-rate (CFAR) target detection algorithm for high-resolution synthetic aperture radar (SAR) images is proposed. The consists of three stages, i.e., segmentation, detection, and clustering. the segmentation stage, superpixel-generating utilized to segment SAR image. based on superpixels generated, clutter distribution parameters each pixel can be adaptively estimated, even in multitarget situations. Then, two-parameter CFAR test...

Thanks to the medium-to-high resolution and wide coverage imaging ability, satellite synthetic aperture radar (SAR) systems are momentously developed for intelligent maritime surveillance, especially ship classification in SAR images. Previous researchers mostly utilized geometric, radiometric, structural features combined with traditional machine-learning (ML) methods conduct high-resolution (HR) However, handcrafted showed weak representation medium-resolution (MR) images, normal ML less...

Target discrimination has been one of the hottest issues in interpretation synthetic aperture radar (SAR) images. However, presence speckle noise and absence robust features make SAR difficult to deal with. Recently, convolutional neural network obtained state-of-the-art results pattern recognition. In this letter, we propose a target framework that jointly uses intensity edge information This contains three parts, namely, feature extraction block, fusion final classification block....

To detect ships robustly and automatically in monitoring the marine areas, polarimetric synthetic aperture radar imagery is more important. In this letter, three superpixel-level dissimilarity measures are developed to enhance contrast between ship targets sea clutter, which then used construct an automatic detection algorithm. proposed method, multiscale superpixels first generated. Second, measurements a certain superpixel surrounding ones calculated. The transformed from level pixel...

Ship classification from synthetic aperture radar (SAR) images tends to be a hotspot in the remote sensing community. Currently, more efforts have been made single-polarization (single-pol) SAR ship with limited performance. This letter proposes explore dual-polarization (dual-pol) for better classification. To specific, novel group bilinear convolutional neural network (GBCNN) model is developed deeply extract discriminative second-order representations of targets pairwise VH and VV...

Transformer-based approaches have demonstrated remarkable performance in image processing tasks due to their ability model long-range dependencies. Current mainstream methods typically confine self-attention computation within windows reduce computational burden. However, this constraint may lead grid artifacts the reconstructed images insufficient cross-window information exchange, particularly super-resolution tasks. To address issue, we propose Multi-Scale Texture Complementation Block...

In this letter, the local scattering mechanism difference based on regression kernel (LSMDRK) is developed as a discriminative feature for ship detection. The LSMDRK measures dissimilarity of center pixel to its neighboring pixels. A detection scheme proposed LSMDRK. consists two stages. extraction stage, polarimetric target decomposition required improve ability descriptor. saliency strategy utilized construct map. Then, maximum employed. Finally, an adaptive threshold method designed...

Ship classification using synthetic aperture radar (SAR) images plays a core role in modern maritime surveillance. Traditional methods mainly applied the handcrafted features for ship representation SAR images, which can hardly deal with resolution-limited well. Recently, deep learning methodology opens new door effective and efficient classification. This paper is dedicated to make exploration on convolutional neural networks (CNNs) First, novel two stream CNN framework proposed...

As one of crucial remote sensing applications, ship classification using synthetic aperture radar (SAR) images has increasingly been studied in modern maritime surveillance. Nowadays, the prevailing paradigm for SAR targets is to utilize deep network models, which presents superior performance over traditional handcrafted feature driven methods. Of method densely connected convolutional neural networks (CNNs) among state-of-the-art. However, general CNNs cannot fully explore representations,...

With the development of deep learning, fine-grained image classification task has made remarkable achievements, but it largely depends on a large number annotated data samples. However, in practical applications, such as public safety, medicine, endangered species and other professional fields, samples with annotations are difficult to obtain. When insufficient, traditional learning training method is easy yield over fitting phenomenon, which greatly limits applicability algorithms. The...

Abstract In the field of single image super‐resolution, prevalent use convolutional neural networks (CNN) typically assumes a simplistic bicubic downsampling model for degradation. This assumption misaligns with complex degradation processes encountered in medical imaging, leading to performance gap when these algorithms are applied real scenarios. Addressing this critical discrepancy, our study introduces novel comparative learning framework meticulously designed nuanced characteristics...

With the rapid development of modern world, it is imperative to achieve effective and efficient monitoring for territories interest, especially broad ocean area. For surveillance ship targets at sea, a common powerful approach take advantage satellite synthetic aperture radar (SAR) systems. Currently, using SAR images classification challenging issue due complex sea situations imaging variances ships. Fortunately, emergence advanced sensors has shed much light on automatic target recognition...

At present, the image super-resolution method based on convolutional neural network has achieved a very high PSNR, but high-frequency information obtained by using mean square error as loss function is not sufficient, and when scale factor large, detail texture of restored blurred, it completely consistent with human visual perception. Therefore, this paper proposes an algorithm GAN. We modify residual block original SRGAN generator into three modules: Edge-Reconstruction network,...

In the field of hyperspectral anomaly detection, autoencoder (AE) have become a hot research topic due to their unsupervised characteristics and powerful feature extraction capability. However, autoencoders do not keep spatial structure information original data well during training process, is affected by anomalies, resulting in poor detection performance. To address these problems, method based on with superpixel manifold constraints proposed. Firstly, segmentation technique used obtain...