Due to the broad usage and widespread popularity of drones, demand for a more accurate object detection algorithm images captured by drone platforms has become increasingly urgent. This article addresses this issue first analyzing unique characteristics datasets related drones. We then select widely used YOLOv7 as foundation conduct comprehensive analysis its limitations, proposing targeted solution. In order enhance network’s ability extract features from small objects, we introduce...

This paper presents a multi-Helmholtz unit series–parallel acoustic metamaterial with segmented noise control effect designed by taking advantage of the adjustable frequency band an metamaterial, aiming to change main spectrum characteristics under different driving conditions military equipment such as tanks. Based on transfer matrix method, theoretical model that can predict hybrid structure multiple Helmholtz resonator (HR) units is established, and its feasibility verified through finite...

In this paper, an object detection and recognition method based on improved YOLOv5 is proposed for application unmanned aerial vehicle (UAV) images. Firstly, we the traditional Gabor function to obtain convolutional kernels with better edge enhancement properties. We used eight enhance edges from directions, enhanced image has obvious features, thus providing best area subsequent deep feature extraction work. Secondly, added a coordinate attention (CA) mechanism backbone of YOLOv5. The...

A frequency domain imaging algorithm, featured as joint two-dimensional (2D) time and resampling, used for high-resolution high-squint large-scene (HHL) spaceborne sliding spotlight synthetic aperture radar (SAR) processing is proposed in this paper. Due to the nonlinear beam rotation during HHL data acquisition, Doppler centroid varies nonlinearly with azimuth traditional sub-aperture approaches two step approach fail remove inertial aliasing of SAR data. In addition, curved orbit effect...

Purpose Highlights in academic papers serve as condensed summaries of the author’s key work, allowing readers to quickly grasp paper’s focus. However, many journals do not currently offer highlights for their articles. To address this gap, some scholars have explored using supervised learning methods extract from papers. A significant challenge approach is need substantial amounts training data. Design/methodology/approach This study examines effectiveness prompt-based generating highlights....

Abstract To address the issues of weak robustness and low accuracy traditional SLAM data processing algorithms in texture environments such as light contrast, this paper first studies improves feature extraction method, optimizing AGAST-based algorithm to adaptively adjust threshold according gradient size different features. Meanwhile, a fusion-based incremental loop closure detection method is proposed, which integrates similarity scores multi-dimensional features based on Borda counting...

In military operations, the accurate detection of enemy camouflaging targets is great significance to improve our combat capability. Most existing camouflage target methods are aimed at a single kind natural image, and their ability detect image in complex environment limited. order solve this problem, paper proposes two-branch multi-source remote sensing method. By combining fine-grained features infrared visible images, method fuses information Experiments show that accuracy, recall rate...

The performance of HTS transformers is integral to predictability outputs in transformer-rectifier flux pumps. To date, little attention has been given compact using stacked-tape cables. Current sharing behaviour with cable secondary windings studied here experimentally. Secondaries 1-4 tapes are tested a 500:1 transformer. Two modes coupling identified. Firstly, primary inductive determines current during changes current. Secondly, shared between loops when held constant. This caused by...

High-temperature superconducting (HTS) magnets offer a novel pathway to magnetically confined fusion energy generation. Reactor-scale devices require tens of kA current generate large toroidal fields for magnetic confinement. Operating HTS at kA-current levels is difficult due unavoidable joint resistances that active driving current. The footprint and demand solid-state power supply multi-kA leads present significant constraints on the size, cost efficiency tokamak. An alternative approach...

The technology for object detection in remote sensing images finds extensive applications production and people’s lives, improving the accuracy of image is a pressing need. With that goal, this paper proposes range improvements, rooted widely used YOLOv7 algorithm, after analyzing requirements difficulties images. Specifically, we strategically remove some standard convolution pooling modules from bottom network, adopting stride-free to minimize loss information small objects transmission....

More and more useful drug controlled release profiles have been reported in literature for potential applications of "efficacious, safe convenient" delivery [1][2][3][4][5][6][7][8][9].These include immediate release, pulsatile delayed sustained zero-order kinetic dual-stage multiple-phase terms the rate initiating time [10][11][12][13][14][15][16][17].Among these profiles, is composed a first rapid quickly eliminating uncomfortable symptoms (such as ache, fever depression) later long period...

The objects in UAV aerial images have multiple scales, dense distribution, and occlusion, posing considerable challenges for object detection. In order to address this problem, paper proposes a real-time multi-scale detection method based on an improved YOLOv7 model (ATS-YOLOv7) images. First, introduces feature pyramid network, AF-FPN, which is composed of adaptive attention module (AAM) enhancement (FEM). AF-FPN reduces the loss deep information due reduction channels convolution process...

When working with objects on a smaller scale, higher detection accuracy and faster speed are desirable features. Researchers aim to endow drones these attributes in order improve performance when patrolling controlled areas for object detection. In this paper, we propose an improved YOLOv7 model. By incorporating the variability attention module into backbone network of original model, association between distant pixels is increased, resulting more effective feature extraction and, thus,...

UAV remote sensing (RS) image object detection is a very valuable and challenging technology. This article discusses the importance of key features proposes an network (URSNet) based on bidirectional multi-span feature pyramid capture mechanism. Firstly, (BMSFPN) constructed. In process sampling, bicubic interpolation cross layer fusion are used to filter out noise enhance details features. Secondly, designed polarization module (FPM) uses internal attention mechanism build powerful...

To overcome the limitations of two-dimensional simultaneous localization and mapping (2D-SLAM) technology, which struggles with closed-loop detection in large spatial environments unstructured poor lighting conditions, leading to accumulation positioning errors, information loss, low map construction accuracy, solutions need be found. We focused on multimodal fusion algorithm a depth camera LiDAR based particle filter at front end SLAM. The precision complex is maintained through utilization...

Abstract MRI (magnetic resonance imaging) images can effectively show the placental haemorrhage area. In view of special properties and real‐time detection requirements images, this paper has systematically improved single‐shot multi‐box detector (SSD) target algorithm (M‐SSD). First, taking advantage particularity image, maximum stable extremum region (MSER) was used as anchor proposal network which integrated information into feature layer SSD to avoid hungry traversal original algorithm....

Abstract In response to challenges faced by traditional detection methods such as image blurring, scarcity of insulator defect datasets, and insulators small targets, we propose an improved Deformable DETR network based on the algorithm. This model accurately identifies position information segments string. To address classification problem after detection, introduce a fused classifier with self-attention mechanism built behind model. The detected are classified into three categories. Due...

In recent years, with the rapid developments in science and technology which have taken place, wireless signals evolved from a simple communication medium to something can be used for environmentally aware tools. Many researchers applied research fields of human perception behavior recognition. At present, vision-based sensor-based action recognition represent two mainstream methods. However, former is sensitive levels light available environment, latter requires users wear (or deploy)...