A5100750908- Remote-Sensing Image Classification
- Advanced Image and Video Retrieval Techniques
- Remote Sensing and LiDAR Applications
- Geochemistry and Geologic Mapping
- Advanced Neural Network Applications
- Remote Sensing in Agriculture
- Advanced Image Fusion Techniques
- Remote Sensing and Land Use
- Topic Modeling
- Land Use and Ecosystem Services
- Data Management and Algorithms
- Robotics and Sensor-Based Localization
- Image Retrieval and Classification Techniques
- Language, Metaphor, and Cognition
- Infrared Target Detection Methodologies
- Multimodal Machine Learning Applications
- Advanced Image Processing Techniques
- Autonomous Vehicle Technology and Safety
- Video Surveillance and Tracking Methods
- Sentiment Analysis and Opinion Mining
- Advanced Database Systems and Queries
- Automated Road and Building Extraction
- Infrastructure Maintenance and Monitoring
- Photoacoustic and Ultrasonic Imaging
- Human Pose and Action Recognition
China University of Geosciences
2016-2025
Ministry of Education of the People's Republic of China
2025
Shandong Sport University
2024
Hebei Medical University
2024
Second Hospital of Hebei Medical University
2024
Hebei General Hospital
2024
Baotou Teachers College
2024
China Meteorological Administration
2024
China Electronics Technology Group Corporation
2013-2024
Research Institute of Highway
2022-2024
The research community has increasing interest in autonomous driving research, despite the resource intensity of obtaining representative real world data. Existing self-driving datasets are limited scale and variation environments they capture, even though generalization within between operating regions is crucial to over-all viability technology. In an effort help align community's contributions with real-world problems, we introduce a new large scale, high quality, diverse dataset. Our...
In multimodal sentiment analysis (MSA), the performance of a model highly depends on quality synthesized embeddings. These embeddings are generated from upstream process called fusion, which aims to extract and combine input unimodal raw data produce richer representation. Previous work either back-propagates task loss or manipulates geometric property feature spaces favorable fusion results, neglects preservation critical task-related information that flows results. this work, we propose...
Remote sensing image scene classification has been widely applied and attracted increasing attention. Recently, convolutional neural networks (CNNs) have achieved remarkable results in classification. However, images complex semantic relationships between multiscale ground objects, the traditional stacked network structure lacks ability to effectively extract key features, resulting limited feature representation capabilities. By simulating way that humans understand perceive images,...
Abstract. Early-season crop identification is of great importance for monitoring growth and predicting yield decision makers private sectors. As one the largest producers winter wheat worldwide, China outputs more than 18 % global production wheat. However, there are no distribution maps over a large spatial extent with high resolution. In this study, we applied phenology-based approach to distinguish from other crops by comparing similarity seasonal changes satellite-based vegetation index...
Multimodal sentiment analysis aims to extract and integrate semantic information collected from multiple modalities recognize the expressed emotions in multimodal data. This research area's major concern lies developing an extraordinary fusion scheme that can key various modalities. However, previous work is restricted by lack of leveraging dynamics independence correlation between reach top performance. To mitigate this, we propose Bi-Bimodal Fusion Network (BBFN), a novel end-to-end...
The research community has increasing interest in autonomous driving research, despite the resource intensity of obtaining representative real world data. Existing self-driving datasets are limited scale and variation environments they capture, even though generalization within between operating regions is crucial to overall viability technology. In an effort help align community's contributions with real-world problems, we introduce a new large scale, high quality, diverse dataset. Our...
Detecting objects from LiDAR point clouds is an important component of self-driving car technology as provides high resolution spatial information. Previous work on point-cloud 3D object detection has re-purposed convolutional approaches traditional camera imagery. In this work, we present system called StarNet designed specifically to take advantage the sparse and nature cloud data. entirely point-based, uses no global information, data dependent anchors, sampling instead learned region...
High-resolution remote sensing (HRRS) image scene classification has attracted an enormous amount of attention due to its wide application in a range tasks. Due the rapid development deep learning (DL), models based on convolutional neural network (CNN) have made competitive achievements HRRS because excellent representation capacity DL. The labels images extremely depend combination global information and from key regions or locations. However, most existing CNN tend only represent features...
Object detection that focuses on locating objects of interest and categorizing them has long played a critical role in the development remote sensing imagery. Following significant improvements Earth observation technologies, high-resolution (HRRS) images show additional detailed information more complex patterns. Some applications, such as urban monitoring, military reconnaissance, national security, have urgent needs terms identifying small-scale (small) weak-feature-response (weak)...
Geological remote sensing interpretation can extract elements of interest from multiple types images, which is vital in geological survey and mapping, especially inaccessible regions. However, due to numerous classes, high interclass similarities, complex distributions, sample imbalances elements, the results machine-learning (ML)-based methods are understandably worse than manual visual interpretation. Additionally, scholars have mainly carried out their works interpret a single element...
Accurate bathymetric maps are essential to understand marine and coastal ecosystems. With the development of satellite sensor technology, satellite-derived bathymetry (SDB) has been widely used measure depth nearshore waters. Employment physics-based methods requires a series optical parameters water column seafloor, which limits application these shallow-water bathymetry. Due convenience, low costs, high efficiency, empirical based on <i>in situ</i> measurements imagery increasingly for...
Urban informal settlements (UIS) are high-density population areas with low urban infrastructure standards. UIS classification, which automates identifying UIS, is of great significance for various computing tasks. Fast and accurate extraction has the following difficulties. First, from a high-resolution perspective, buildings in settlement low-floor dense, complex spatial relationships. Second, settlements' remote sensing observation characteristics highly inconspicuous, caused by shooting...
A yolk-like nanocapsule with responsiveness to tumor microenvironment and NIR photons was invented by integrating a tumor-responsive photothermal agent on Mn-doped UCNPs@mSiO<sub>2</sub> nanospheres for multiple imaging guided thermo-chemotherapy.
Due to the trade-off of temporal resolution and spatial resolution, spatiotemporal image-fusion uses existing high-spatial-low-temporal (HSLT) high-temporal-low-spatial (HTLS) images as prior knowledge reconstruct high-temporal-high-spatial (HTHS) images. However, some algorithms ignore issue that information HTLS is insufficient support acquisition information, which leads unsatisfactory accuracy fusion result. To introduce more algorithm in this article Cycle-generative adversarial...
As the second largest producer of maize, China contributes 23% global maize production and plays an important role in guaranteeing markets stability. In spite its importance, there is no 30 m spatial resolution distribution map for all China. This study used a time-weighted dynamic time warping method to identify planting areas by comparing similarity series satellite-based vegetation index at each pixel with standard derived from known fields mapped 2016 2020 over 22 provinces accounting...
High-resolution (HR) remote sensing imagery plays a critical role in image interpretation, and single super-resolution (SISR) reconstruction technology is becoming increasingly valuable significant. The state-of-the-art deep-learning-based SISR methods have demonstrated remarkable advantages, while reconstructing complex texture details still remains big challenge. Besides, as typical ill-posed inverse problem, how to determine the optimal solution another important topic. To address these...
Dense time-series remote sensing images have transformed the traditional bitemporal land-cover change detection to continuous monitoring. Previous work mostly employs linear fitting, prediction, or decomposition methods, and accuracy is not high. The latest progress of deep learning (DL) shows its advantages in However, DL models are computationally expensive require lots labeled samples, resulting often employed prediction-threshold-based unsupervised method. determination a reasonable...