Ship detection aims to automatically identify whether there are ships in the images, precisely classifies and localizes them. Regardless of utilizing early manually designed methods or deep learning technology, ship is dedicated exploring inherent characteristics enhance recall. Nowadays, high-precision plays a crucial role civilian military applications. In order provide comprehensive review optical remote-sensing images (SDORSIs), this paper summarizes challenges as guide. These include...

Deep learning has made significant achievements in remote sensing object detection tasks. However, small weak objects located complex scenes are still not effectively addressed. The lack of feature information and negligible contributions during the optimization stage main reasons. To solve indicated issues, a novel remote-sensing method is proposed this letter. Firstly, context integration module (CFIM) designed to extract implicit clues co-occurring with compensate for features objects....

Recently, remote sensing image object detection based on convolutional neural networks (CNNs) has made significant advancements. However, small objects remains a major challenge in this field. Because the size of makes it difficult to extract their features and these are further weakened after downsampling network. In order improve accuracy images, letter provides feature super-resolution fusion framework cross-scale distillation. Specifically, we design sub-pixel pyramid network (SSRFPN)...

The complementary characteristics of SAR and optical images are beneficial in improving the accuracy land cover classification. Deep learning-based models have achieved some notable results. However, how to effectively extract fuse unique features multi-modal for pixel-level classification remains challenging. In this article, a two-branch supervised semantic segmentation framework without any pretrained backbone is proposed. Specifically, novel symmetric attention module designed with...

Graph convolutional networks (GCNs) can extract features of samples in non-Euclidean space, which be used for hyperspectral image (HSI) classification collaboration with neural (CNNs). The GCNs and CNNs are incompatible to a certain extent, traditional graph convolution methods use single channel two-dimensional matrix features. As result, it is difficult explore the relationships fully flexibly between samples. To further exploit potential these two collaborative extraction HSI features, we...

Object detection plays a crucial role in remote sensing due to the urgent demands of various applications, such as urban planning and environmental monitoring. Despite notable progress, current methods still struggle with detecting challenging small objects. At object level, limited pixel representation, blurred details, background interference objects impose greater on feature extractors. network resource bias fails provide adequate learning signals for these In this paper, we propose...

In deep learning-based hyperspectral remote sensing image classification tasks, random sampling strategies are typically used to train model parameters for testing and evaluation. However, this approach leads strong spatial autocorrelation between the training set samples surrounding test samples, some unlabeled data directly participate in of network. This leaked information makes overly optimistic. Models trained under these conditions tend overfit a single dataset, which limits range...

In the field of remote sensing image (RSI) object detection, oriented box annotation can accurately locate objects with arbitrary orientation and obtain information. However, detection based on bounding (OBB) is still a challenging task, effect needs to be improved. RSI, distribution extremely uneven, situation aggregation easy occur. Some researchers believe that characteristics dense are one reasons for difficulty detection. impact performance has not been studied in depth. To address this...

Face recognition is constructed based on facial feature extraction and classification, the taken in face region according to different parts of characteristic, it prone confuse issue, this paper proposes a algorithm neural network. In order reduce interference background noise. Reduce post point location complexity binarization image denoising method for denoising, noise reduction output extraction, extract value peak valley two-dimensional features, can get edge regions around border, BP...

Hyperspectral image super-resolution technology has made remarkable progress due to the development of deep learning. However, technique still faces two challenges, i.e., imbalance between spectral and spatial information extraction, parameter deviation high computational effort associated with 3D convolution. In this article, we propose a method for remote sensing hyperspectral images based on multi-domain multi-scale fusion (MSSR). Specifically, inspired by degree self-similarity images,...

Chemical vapor deposition (CVD) is the most promising method for preparation of high-quality and large-area graphene films, especially epitaxial growth on single-crystal Cu foils. While foils are normally achieved by thermally annealing commercial polycrystalline foils, their size therefore films grown them limited to reaction chamber. We report a simple feasible prepare with decimeter grains rolled-up where layers separated thin porous carbon fiber cloths. The cloths prevent from sticking...

On-line inspection of PCBs requires acquisition and processing gigabytes image data in a matter few seconds, especially when multi-layer very high-resolution boards are used. To meet the demands for speed accuracy, our system uses run-length encoding (RLE) storage operations an scheme which exploits availability artwork comparison purposes. The is suitable parallel consists four parts: (1) segmentation feature extraction, (2) acquisition, (3) blank areas, (4) trace areas. First, available as...

Image difference operation is commonly used in on-line automated printed circuit board (PCB) inspection systems as well many other image processing applications. In this paper, we describe a new systolic algorithm and its system architecture which computes differences run-length encoded (RLE) format. The efficiency of greatly affects the overall performance system. It shown that, for images with high similarity measure, time complexity small constant. A formal proof correctness also given paper.

This paper presents a fast printed circuit board (PCB) inspection methodology based on the segmentation of PCB image into basic patterns and context sensitive filtering difference for functional defect detection. The system consists three parts: (1) artwork test images using CAD data artwork, (2) comparison windows with corresponding windows, (3) context-sensitive obtained in step determination defects their locations. For representation, runlength encoding (RLE) is used processing done...

The construction of intrusion risk assessment model intelligent power network communication system is studied, designed for detection improving the security in network, and ability connectivity fault system. A method systems proposed based on homomorphic routing modulation adaptive empirical mode decomposition. multi-path restructuring signal, signal using fuzzy beam about spectrum decomposition method, choose a wavelet function to realize constraint control homomorphism beamforming node,...

Low-light enhancement is a challenging task.With the image brightness increasing, noises are amplified, and with contrast detail false information generated.In order to solve this problem, paper proposes novel end-to-end attention-guided method (A-MBLLEN) based on multi-branch convolutional neural network.The proposed network composed module (EM) Convolutional Block Attention Module (CBAM).The attention can make CNN structure gradually focus weak light area in image, fully highlight feature...

In order to solve the problems of long filter time, low efficiency and utilisation rate filtered information in traditional data filtering methods, a method falsified smart grid wireless communication based on SVM is proposed. initial stage population search, chaos model introduced increase diversity individuals, adaptive factors are added into updating mechanism global search capability, feature fitness function adjust classification accuracy number features by using penalty factors. At...

Combining organic and inorganic fertilizers is critical for increasing yield, reducing greenhouse gas (GHG) emissions, improving soil fertility. However, the effect of combined on GHG emissions in hilly apple orchards not clear. Furthermore, studies slope agriculture mostly ignore runoff. Hence, a two-year field experiment was conducted orchard north Shaanxi to explore effects management practices surface water temperature, runoff, fruit quality. Three were implemented: (1) chemical...