A5025442255- Remote-Sensing Image Classification
- Remote Sensing and Land Use
- Advanced Image Fusion Techniques
- Advanced Image Processing Techniques
- Robotics and Sensor-Based Localization
- Face and Expression Recognition
- Advanced Vision and Imaging
- Advanced Chemical Sensor Technologies
- Image Processing Techniques and Applications
- Image and Signal Denoising Methods
- Control Systems in Engineering
- Remote Sensing and LiDAR Applications
- MRI in cancer diagnosis
- Sensorless Control of Electric Motors
- Optical measurement and interference techniques
- Image and Object Detection Techniques
- Iterative Learning Control Systems
- Target Tracking and Data Fusion in Sensor Networks
- 3D Shape Modeling and Analysis
- Advanced Neural Network Applications
- Geochemistry and Geologic Mapping
- Smart Agriculture and AI
- Infrared Target Detection Methodologies
- 3D Surveying and Cultural Heritage
- Lanthanide and Transition Metal Complexes
Dalian University of Technology
2024
Nanjing University of Science and Technology
2016-2021
Nanjing University of Information Science and Technology
2021
Beijing Institute of Technology
2020
Tianjin University of Technology
2020
Wuhan University of Technology
2014-2016
Jilin University
2015
The emergence of a convolutional neural network (CNN) has greatly promoted the development hyperspectral image (HSI) classification technology. However, acquisition HSI is difficult. lack training samples primary cause low performance. traditional CNN-based methods mainly use 2-D CNN for feature extraction, which makes interband correlations HSIs underutilized. 3-D extracts joint spectral-spatial information representation, but it depends on more complex model. Also, too deep or shallow...
Purpose To develop a reproducible and fast method to reconstruct MR fingerprinting arterial spin labeling (MRF‐ASL) perfusion maps using deep learning. Method A fully connected neural network, denoted as DeepMARS, was trained simulation data added Gaussian noise. Two MRF‐ASL models were used generate the data, specifically single‐compartment model with 4 unknowns parameters two‐compartment 7 unknown parameters. The DeepMARS evaluated from healthy subjects (N = 7) patients Moymoya disease 3)....
Ensembles of extreme learning machine (ELM) have been widely used for hyperspectral image classification. The previous studies shown that the goal ensemble is to train accurate but diverse component classifiers improve generalization performance. To approach this goal, article proposes a novel framework construct an ELM model. proposed relies on multiview-based random rotation pruning (MVRR-EP) and has several features. First, ensure subsets spectral bands can sufficiently learn target...
As the hyperspectral images (HSIs) usually have a low spatial resolution, HSI super-resolution has recently attracted more and attention to enhance resolution of HSIs. A common method is fuse low-resolution (LR) with multispectral image (MSI) whose higher than HSI. In this article, we proposed novel adaptive nonnegative sparse representation-based model an its corresponding MSI. First, basing linear spectral unmixing, structured representation estimates codes desired high-resolution from...
Convolutional neural networks (CNNs), a kind of feedforward network with deep structure, are one the representative methods in hyperspectral image (HSI) classification. However, redundant information and interclass interference common challenging problems HSI In addition, if spectral spatial is not properly extracted analyzed, it will affect classification performance to great extent. Aiming at these issues, this article proposes an method based on adaptive hash attention mechanism lower...
Graph-based semisupervised hyperspectral image (HSI) classification methods have obtained extensive attention. In graph-based methods, a graph is first constructed, and then the label propagation carried out on constructed to obtain labels for unknown samples. However, results of may be unreliable, especially in case very limited labeled To address above problem, we propose dual sparse representation collaborative (DSRG-CP) HSI classification. Specifically, DSRG-CP adopts (SR) construct...
The frontal‐parallel assumption is made by many matching algorithms, but this fails for slanted surfaces. This study proposes a algorithm intended to improve the results First, mathematical model constructed prove that surfaces in environment have corresponding disparity space image, and help find proper plane parameters of support windows, then improved cost aggregation post‐processing methods are proposed. tested using Middlebury Karlsruhe Institute Technology Toyota Technical at Chicago...
In crop disease image segmentation, traditional convolutional neural networks have the problem of low accuracy. For this reason, paper proposes a segmentation method that combines conditional random fields with segnet networks. First, input training set in data into network for training, and obtain initial result updated parameters; secondly, segmented by is sent to field form pixels classification vectors corresponding At same time, construct an energy function represent relative...
Recently, hyperspectral image (HSI) classification has become a popular research direction in remote sensing. The emergence of convolutional neural networks (CNNs) greatly promoted the development this field and demonstrated excellent performance. However, due to particularity HSIs, redundant information limited samples pose huge challenges for extracting strong discriminative features. In addition, addressing how fully mine internal correlation data or features based on existing model is...
In military training, the shooting target reporting system faces several challenges, including difficulty grasping surface, low real-time judgment of bullet hole ring value, and susceptibility to external environmental factors. To address these issues, this work proposes an approach that combines deep learning detection semantic segmentation algorithms with traditional YOLOv5 model. The model is enhanced into a multi-task architecture, featuring backbone network encoder two decoders, detect...
This paper deals with the gear shift control of 2-speed AMT driving system Electric Vehicle backlash.A motor torque strategy is proposed to eliminate impact backlash when adjust speed system.The confirmed through simulation tests on a complete power train model.The results show that it can reduce backlash.
Welcome to the WIT Press eLibrary - home of Transactions Wessex Institute collection, providing on-line access papers presented at Institute's prestigious international conferences and from its State-of-the-Art in Science & Engineering publications.
An automatic change detection method based on conditional random field (CRF) is presented for high resolution remote sensing images in this paper. Marginalized denoising autoencoder used to generate the difference image. The clustering results of Fuzzy C-means are applied initialize unary potentials CRF. A scaled squared Euclidean distance between neighboring pixels observed introduced define pairwise CRF, which avoid training parameters and help improve accuracy degree automation....
In order to solve the problems of insufficient SAR data and difficulty in labeling image recognition task, this paper proposes a method based on parallel vision. We present an improved deep convolutional generative adversarial network model that is used generate large number diverse virtual samples. By mixing real images images, we carry out computational experiments for recognition. A public dataset MSTAR verify effectiveness vision applied Combining learning can further promote...