A5101599198- Remote-Sensing Image Classification
- Advanced Image Fusion Techniques
- Remote Sensing and Land Use
- Geochemistry and Geologic Mapping
- Spectroscopy and Chemometric Analyses
- Advanced Image and Video Retrieval Techniques
- Indoor and Outdoor Localization Technologies
- Anomaly Detection Techniques and Applications
- Image Retrieval and Classification Techniques
- Engine and Fuel Emissions
- Time Series Analysis and Forecasting
- Underwater Vehicles and Communication Systems
- 3D IC and TSV technologies
- Advanced Image Processing Techniques
- Machine Fault Diagnosis Techniques
- Electric Vehicles and Infrastructure
- Advanced Data Storage Technologies
- Solar Radiation and Photovoltaics
- Building Energy and Comfort Optimization
- Speech and Audio Processing
- Graph Theory and Algorithms
- Energy Efficient Wireless Sensor Networks
- Semiconductor materials and devices
- Advanced Adaptive Filtering Techniques
- Blind Source Separation Techniques
Nanjing University of Posts and Telecommunications
2019-2023
Beijing Normal University
2017-2021
Anhui University
2021
Harbin Institute of Technology
2013-2019
North China Electric Power University
2016
Tsinghua University
2015
Fusarium head blight, caused by a fungus, can cause quality deterioration and severe yield loss in wheat. It produces highly toxic deoxynivalenol, which is harmful to human animal health. In order quickly accurately detect the severity of fusarium method detecting disease using continuous wavelet analysis particle swarm optimization support vector machines (PSO-SVM) proposed this paper. First, seven features for blight detection were extracted based on hyperspectral reflectance wheat ears....
How to classify hyperspectral images using few training samples is an important and challenging problem because the collection of difficult expensive. Because semi-supervised approaches can utilize information contained in unlabeled labeled samples, it a suitable choice. A novel spectral-spatial classification method for data based on generative adversarial network (GAN) proposed this paper. First, we use custom one-dimensional GAN train obtain spectral features. After new small...
Classification of stellar spectra from voluminous is a very important and challenging task. In order to better classify spectra, inspired by the principle deep convolutional neural network (CNN), we propose supervised algorithm for classification based on 1D (1D SSCNN). SSCNN, modify traditional 2D into adapt spectral classification. On basis using convolution algorithm, features are extracted used We first use data train SSCNN obtain well-trained model, then apply model unknown spectra. To...
With the development of information technology, human activity recognition and localization has received much attentions, since they can be utilized in many fields. In this paper, a new algorithm using channel state (CSI) measurement is proposed. The problem are formulated as machine learning which solve with support vector (SVM) approach. off-line phase, after data normalization principal component analysis (PCA) preprocessing CSI measurements, (CSI measurement, label activity) training set...
Since indoor location information can be utilized in many aspects, localization using existing wireless network has received much attentions. In this paper, a novel algorithm signal strength indicator (RSSI) measurements and statistical feature of the channel state (CSI) is proposed. off-line phase, with obtained CSI measurement, principal component analysis (PCA) pre-processing for dimension reduction at first. Then, features are extracted. Next, RSSI measurements, extracted reference...
Band selection is a very important hyperspectral image preprocessing before using data. A novel bands method for data based on convolutional neural network (CNN) proposed in this paper. In way, we use custom one-dimensional CNN to train the obtain well-trained model. After testing band combinations, model test precision of different and finally combination with highest as selected bands. This measure new criterion selection. first application selection, our can select better combinations...
Classifying Hyperspectral images with few training samples is a challenging problem. The generative adversarial networks (GAN) are promising techniques to address the problems. GAN constructs an game between discriminator and generator. generator generates that not distinguishable by discriminator, determines whether or sample composed of real data. In this paper, introducing multilayer features fusion in dynamic neighborhood voting mechanism, novel algorithm for HSIs classification based on...
Low-light image enhancement is a crucial area of research in computer vision, aimed at recovering normally exposed images from low-light to facilitate high-level vision tasks such as target detection, tracking, and recognition. However, the convolutional neural networks commonly used this field have bias towards extracting low-frequency local structural features spatial domain, resulting unclear texture details enhanced images. To address limitation, paper proposes novel frequency-domain...
Abstract Targeting the problem of high real-time requirements in astronomical data processing, this paper proposes a early warning model for light curves based on Gated Recurrent Unit (GRU) network. Using memory function GRU network, prediction curve is established, and trained using collected data, so that can predict star magnitude value next moment historical data. In paper,we calculate difference between actual observation set threshold. If exceeds threshold, at considered to be an...
Hyperspectral remote sensing data contains near continuous spectral information of the object, which is very suitable for mineral classification and geologic body mapping. However, collecting a lot labeled hyperspectral expensive, time-consuming labor-intensive. We choose semi-supervised method to classify based on generative adversarial nertwork (GAN), just use small amount data, named HSGAN. The GAN made up generator discriminator, generates similar real so that discriminator cannot tell...
This 72 Mb synchronous-link DRAM (SLDRAM) is a proof-of-concept vehicle for next-generation memory. SLDRAM packet-protocol-based memory that employs source-synchronous busses with push-pull I/O signaling integrity. devices are calibrated on power-up by the controller so individual do not have to meet tight timing specifications and compensate interconnect loading variations.
Detecting supernova remnant (SNR) candidates in the interstellar medium is a challenging task because SNRs have weak radio signals and irregular shapes. The use of convolutional neural network deep learning method that can help us extract various features from images. To astronomical images estimate positions SNR candidates, we design SNR-Net model composed training component detection component. In addition, transfer used to initialize parameters, which improves speed accuracy training. We...
Hyperspectral data contains abundant information in spectral domain, which is very useful for mineral classification and geological body mapping. But, due to the lack of labeled data, it difficult get an acceptable result by just using small number data. We adopt a semi-supervised method called CNN, can effectively extract inner features hyperspectral image classify However, with constraint size receptive field, hardly higher level features. propose dilated CNN At same kernels, has bigger...
In the process of aeroengine anomaly detection, there is always an unbalance distribution among samples gas path performance parameters, that is, number normal much larger than abnormal samples. addition, this imbalance will worsen with time, which leads to classifier paying too attention in model training. Thus, recognition rate reduce significantly. To solve above problems, adaptive decision threshold support vector machine (ADT-SVM) proposed and applied detection aeroengine. Firstly,...