A5006526093- Advanced Neural Network Applications
- Domain Adaptation and Few-Shot Learning
- Advanced Image and Video Retrieval Techniques
- Remote-Sensing Image Classification
- Advanced Measurement and Detection Methods
- Aerospace Engineering and Energy Systems
- Image Processing Techniques and Applications
- Infrared Target Detection Methodologies
- Anomaly Detection Techniques and Applications
- Radioactive element chemistry and processing
- Robotics and Sensor-Based Localization
- Gas Sensing Nanomaterials and Sensors
- Remote Sensing and Land Use
- Thallium and Germanium Studies
- Target Tracking and Data Fusion in Sensor Networks
- Data Stream Mining Techniques
- Autonomous Vehicle Technology and Safety
- Advanced Chemical Sensor Technologies
- Ultrasound in Clinical Applications
- Text and Document Classification Technologies
- Icing and De-icing Technologies
- High-pressure geophysics and materials
- COVID-19 diagnosis using AI
- Solar Radiation and Photovoltaics
- IoT and GPS-based Vehicle Safety Systems
Shanxi Normal University
2024
Kunming University of Science and Technology
2024
Xiamen University of Technology
2024
Sichuan University
2018-2024
China Institute of Water Resources and Hydropower Research
2024
University of Electronic Science and Technology of China
2020-2023
Beihang University
2014-2023
Shanghai Medical College of Fudan University
2023
Fudan University
2023
Jiangsu University
2022
Semantic segmentation for unmanned aerial vehicle (UAV) remote sensing images has become one of the research focuses in field at present, which could accurately analyze ground objects and their relationships. However, conventional semantic methods based on deep learning require large-scale models that are not suitable resource-constrained UAV tasks. Therefore, it is important to construct a light-weight method images. With this motivation, we propose neural network model with fewer...
Many practical applications of high-resolution remote sensing images (HRRSIs) are based on semantic segmentation. However, due to the complex ground object information contained in images, it is difficult make precise segmentation HRRSIs. In this letter, we proposed a hierarchical context aggregation network (HCANet) for The HCANet has an encoder-decoder structure which similar UNet. HCANet, designed two Compact Atrous Spatial Pyramid Pooling (CASPP and CASPP+) modules. CASPP modules replace...
Currently, the most advanced high-resolution remote sensing image (HRRSI) semantic labeling methods rely on deep neural networks. However, HRRSIs naturally have a serious class imbalance problem which is not yet well solved by current method. The cross-entropy (CE) loss often used to guide training of networks for HRRSIs, but it essentially dominated major classes in image, resulting poor predictions minority class. Based prediction results, Focal Loss (FL) effectively suppresses negative...
Introduction Lung image segmentation plays an important role in computer-aid pulmonary disease diagnosis and treatment. Methods This paper explores the lung CT method by generative adversarial networks. We employ a variety of networks used their capability translation to perform segmentation. The network is employed translate original into segmented image. Results networks-based tested on real data set. Experimental results show that proposed outperforms state-of-the-art method. Discussion...
Semantic segmentation of high-resolution remote sensing (HRRS) images becomes more and important at present. Popular approaches use deep learning to solve this task, which depends on a large amount labeled data powerful computing resources. When resources or the are insufficient, their performance will be severely degraded. To deal with problem, we proposed light-weight network attention modules for semantic HRRS images. The depth width designed, has small number parameters ensure efficiency...
Stratospheric airships are long-endurance aerostats and have broad applications. All of the energy required for their operation is obtained from solar radiation, which makes accurate calculation output array crucial to design flight planning airships. However, status each photovoltaic module in may differ due airship curvature, resulting mismatch losses lowered power, has not been widely studied. In this paper, an irradiation model a thermal established based on actual arrangement modules....
Online action detection plays a vital role in video understanding and can be widely used various analysis applications. This task aims to detect actions at the current moment within long untrimmed streams. However, accurately identifying action-background transitions that are ambiguous terms of time during challenging due similarity between background clips, adding difficulty finding suitable division them. To address this issue, we propose hard clip mining method based on deep metric...
A dynamically and thermodynamically stable Fe-rich compound, Fe<sub>2</sub>Mg, reveals that Mg may be a light element candidate in the earth's inner core.
Deep learning has achieved widespread success in medical image analysis, leading to an increasing demand for large-scale expert-annotated datasets. Yet, the high cost of annotating images severely hampers development deep this field. To reduce annotation costs, active aims select most informative samples and train high-performance models with as few labeled possible. In survey, we review core methods learning, including evaluation informativeness sampling strategy. For first time, provide a...
Abstract Objectives Conventional lung biopsy for critically ill patients who under invasive mechanical ventilation (IMV) is limited due to high risks of procedure-related complications. We developed a novel technique named bronchus-blocked ultrasound-guided percutaneous transthoracic needle (BUS-PTNB) mitigate the IMV, but its safety and efficacy have not been prospectively evaluated. Methods In this prospective, single-arm trial (Chictr.org, ChiCTR2100054047), invasively ventilated with...
Most of target motion models have uncertainties and strong nonlinear characteristics, in the uncertainty processing, effect Unscented Kalman Filter (UKF) is remarkable. In order to achieve accurate tracking, based on UKF algorithm, this paper uses ability dynamic non-linear approximation Elman neural network compensate filtering results UKF. addition, explains shortcomings Extended Filter(EKF) algorithm systems from mechanism. By tracking targets different situations, emulation indicate that...
Introduction Precise delineation of glioblastoma in multi-parameter magnetic resonance images is pivotal for neurosurgery and subsequent treatment monitoring. Transformer models have shown promise brain tumor segmentation, but their efficacy heavily depends on a substantial amount annotated data. To address the scarcity data improve model robustness, self-supervised learning methods using masked autoencoders been devised. Nevertheless, these not incorporated anatomical priors structures....
In order to automate the coil delivery for cold rolling process and reduce manual operation error, this paper presents a method locate track steel using monocular camera image processing, core circle of is used as tracking target. This proposed firstly determines initial target location information by RHT, then configure qualification conditions detection next time step based on parameters from last step. Canny edge detect contour candidates, limited domain method, separation quadrant are...
This paper introduces the principle and features of Beidou positioning system, proposes a new compensation method for dynamic performance temperature sensor in order to solve problems haze monitoring. The model established with this can greatly improve response speed sensor. experimental results show that prediction not only effectively sensor, but also has good robustness it stand strong anti-interference outside noise while monitoring is realized.
Shooting is the regular army military training course which often used as an artificial target, and it has disadvantages such poor safety low target efficiency. However, some existing automatic target-reading devices have inaccurate data, high price not easy to carry. Therefore, of great significance develop a fast accurate carry device for modernization. In this paper, research object chest round target. order complete image acquisition, processing, bullet hole identification ring value...