A5101642220- Advanced Image Fusion Techniques
- Remote-Sensing Image Classification
- Advanced Image and Video Retrieval Techniques
- Quantum Information and Cryptography
- Image and Signal Denoising Methods
- Satellite Image Processing and Photogrammetry
- Adversarial Robustness in Machine Learning
- Neutrino Physics Research
- Advanced Image Processing Techniques
- Quantum Mechanics and Applications
- Web Data Mining and Analysis
- Remote Sensing and Land Use
- Anomaly Detection Techniques and Applications
- Astrophysics and Cosmic Phenomena
- Remote Sensing and LiDAR Applications
- Advanced Malware Detection Techniques
- Atmospheric aerosols and clouds
- Text and Document Classification Technologies
- Advanced Neural Network Applications
- Quantum Computing Algorithms and Architecture
- Video Analysis and Summarization
- Domain Adaptation and Few-Shot Learning
- Topic Modeling
- Air Quality Monitoring and Forecasting
- Robotics and Sensor-Based Localization
Jiangsu University of Science and Technology
2022-2025
University of Chicago
2023
Inner Mongolia University of Technology
2022-2023
Dalian University of Technology
2023
University of Science and Technology of China
2010-2022
Microbiology Institute of Shaanxi
2022
Institute of Microbiology
2022
Chinese Academy of Sciences
2020-2022
Istituto Superiore per la Protezione e la Ricerca Ambientale
2022
Aerospace Information Research Institute
2020-2022
Measurement-device-independent quantum key distribution (MDIQKD) protocol is immune to all attacks on detection and guarantees the information-theoretical security even with imperfect single photon detectors. Recently, several proof-of-principle demonstrations of MDIQKD have been achieved. Those experiments, although novel, are implemented through limited distance a rate less than 0.1 bps. Here, by developing 75 MHz clock fully-automatic highly-stable system, superconducting nanowire...
Anthropogenic waste deposition in aquatic environments precipitates a decline water quality, engendering pollution that adversely impacts human health, ecological integrity, and economic endeavors. The evolution of underwater robotic technologies heralds new era the timely identification extraction submerged litter, offering proactive measure against scourge pollution. This study introduces refined YOLOv8-based algorithm tailored for enhanced detection small-scale debris, aiming to mitigate...
Zero-shot remote sensing scene classification aims to solve the problem on unseen categories and has attracted numerous research attention in field. Existing methods mostly use shallow networks for visual semantic feature learning, encoder are usually fixed during zero-shot learning process, thus failing capture powerful representations classification. In this work, we introduced a vision-language model based contrastive supervision. Our method is capable of semantic-aware using loss...
The main type of obstacles practical applications quantum key distribution (QKD) network are various attacks on detection. Measurement-device-independent QKD (MDIQKD) protocol is immune to all these attacks, and thus, a strong candidate for security. Recently, several proof-of-principle demonstrations MDIQKD have been performed. Although novel, those experiments implemented in the laboratory with secure rates less than 0.1 b/s. Besides, they need manual calibration frequently maintain system...
This letter addresses the issue of accurate object detection in large-area remote sensing images. Although many convolutional neural network (CNN)-based models can achieve high accuracy small image patches, perform poorly images due to large quantity false and missing detections that arise from complex backgrounds diverse groundcover types. To address this challenge, proposes a sample update-based CNN (SUCNN) framework for The proposed contains two stages. In first stage, base...
Accurate predictions of remaining useful life (RUL) important components play a crucial role in system reliability, which is the basis prognostics and health management (PHM). This paper proposed an integrated deep learning approach for RUL prediction turbofan engine by integrating autoencoder (AE) with convolutional generative adversarial network (DCGAN). In pretraining stage, reconstructed data AE not only participate its error reconstruction but also take part DCGAN parameter training as...
This study on soil salinity inversion in coastal tidal flats based Sentinel-2 remote sensing imagery is significant for improving saline–alkali soils and advancing flat agriculture. proposes an improved approach using a new enhanced chaotic mapping adaptive whale optimization neural network (CIWOABP) algorithm. Novel spectral indices were developed to enhance correlations with salinity, significantly outperforming traditional indexes. The CIWOABP model achieved superior validation accuracy...
The actual driving condition and fuel consumption rate gaps between lab real-world are becoming larger. In this paper, we demonstrate an approach to determine the most important factors that may influence prediction of light-duty vehicles. A multilayer perceptron (MLP) method is developed for since it provides accurate classification results despite complicated properties different types inputs. model considers parameters external environmental factors, manipulation vehicle companies,...
Urban functional zones are important space carriers for urban economic and social function. The accurate rapid identification of is great significance to planning resource allocation. However, the factors considered in existing zone methods not comprehensive enough, recognition stops at their categories. This paper proposes a framework that combines multisource heterogeneous data identify categories draw portraits zones. comprehensively describes features from four aspects: building-level...
Current convolutional neural network (CNN)-based methods for remote sensing image segmentation require a large number of densely annotated images model training and have limited generalization abilities unseen object categories. In this letter, we propose novel few-shot learning-based method the semantic images. Our can perform labeling categories with only few samples. More specifically, our starts by using deep CNN to extract high-level features. The prototype representation each class is...
Abstract The adjacency effect can blur and reduce the contrast of satellite images. For low visibility, has a crucial impact on submeter-scale spatial resolution optical (SM) Therefore, SM images are preprocessed with an atmospheric correction before performing visual interpretation quantitative research. This includes (expressed by average background reflectance, ABR) intrinsic reflectance correction. ABR is related to distance from central pixel its surrounding pixels (SDCS) affected...
Deep neural networks (DNNs) are used in various domains, such as image classification, natural language processing and face recognition, etc. However, the presence of malicious examples, generated by specific methods, could result DNNs misclassification. Such maliciously modified examples called adversarial examples. So far, most work about mainly focuses on multi-class classification tasks, only a little has been done field multi-label classification.In this study, we have proposed novel...
Studies have shown that deep neural networks (DNNs) are susceptible to adversarial attacks, which can cause misclassification. The attack problem be regarded as an optimization problem, then the genetic algorithm (GA) is problem-independent naturally designed solve generate effective examples. Considering dimensionality curse in image processing field, traditional algorithms high-dimensional problems often fall into local optima. Therefore, we propose a GA with multiple fitness functions...
In this article, we deal with the problem of change detection in cloudy and rainy areas using multisource remote sensing images. While previous methods mostly focus on pixel or super-pixel levels, introduce concept geo-parcel use it as basic processing unit for our method. Concretely, first extract from an optical high spatial resolution image. Then, divide each into fine-grained segments refined boundaries image segmentation methods. These are used units After that, unsupervised...
We present an innovative approach to mitigating brightness variations in the unmanned aerial vehicle (UAV)-based 3D reconstruction of tidal flat environments, emphasizing industrial applications. Our work focuses on enhancing accuracy and efficiency neural radiance fields (NeRF) for scene synthesis. introduce a novel luminance correction technique address challenging illumination conditions, employing convolutional network (CNN) image enhancement cases overexposure underexposure....
Recent progress has shown great potential of visual prompt tuning (VPT) when adapting pre-trained vision transformers to various downstream tasks. However, most existing solutions independently optimize prompts at each layer, thereby neglecting the usage task-relevant information encoded in tokens across layers. Additionally, structures are prone interference from task-irrelevant noise input images, which can do harm sharing information. In this paper, we propose a novel VPT approach,...
Short text analysis is a challenging task as far the sparsity and limitation of semantics. The semantic extension approach learns meaning short by introducing external knowledge. However, for randomness descriptions in microblogs, traditional methods cannot accurately mine semantics suitable microblog theme. Therefore, we use prominent refined hashtag information microblogs well complex social relationships to provide implicit guidance text. Specifically, design deep hash model based on...