Introduction
My research interests include machine learning, remote sensing, image processing and multi-objective optimization.
Research Experience
January 2019 - present
Nankai University

    School of Statistics and Data Science Tianjin, China

Position

    Professor (Associate)

Education
September 2013 - January 2019
Beihang University (BUAA)
Beihang University (BUAA)
Field of study

    Remote sensing, image processing, machine learning

September 2009 - June 2013
Beihang University (BUAA)
Beihang University (BUAA)
Field of study

    Remote sensing

Publications
Publications (33)
Simultaneously Multiobjective Sparse Unmixing and Library Pruning for Hyperspectral Imagery
Article

    Aug 2020

    Xia Xu
    Xia Xu
    Bin Pan
    Bin Pan
    Zongqing Chen
    Zongqing Chen[...]
    Tao Li
    Tao Li

Sparse hyperspectral unmixing has attracted increasing investigations during the past decade. Recent research has indicated that library pruning algorithms can significantly improve the unmixing accuracies by reducing the mutual coherence of the spectral library. Inspired by the good performance of library pruning, in this article we propose a new...
View
Fig. 1. Overview of our framework. Parameters on black arrows are...
Fig. 3. Illustration of the gate functions generated by different data...
Fig. 4. Illustration of the SGE module enhancement process. The...
Fig. 5. Illustration of our gated bidirectional fusion structure with SGE.
+1 Fig. 8. Activation response during detection. The objects in the red...
A Contextual Bidirectional Enhancement Method for Remote Sensing Image Object Detection
Article
Full-text available

    Aug 2020

    Jun Zhang
    Jun Zhang
    Changming Xie
    Changming Xie
    Xia Xu
    Xia Xu[...]
    Bin Pan
    Bin Pan

In remote sensing images, the backgrounds of objects include crucial contextual information that may contribute to distinguishing objects. However, there are at least two issues that should be addressed: not all the backgrounds are beneficial, and object information may be suppressed by backgrounds. To address these problems, in this paper, we prop...
View
An End-to-End Network for Remote Sensing Imagery Semantic Segmentation via Joint Pixel- and Representation-Level Domain Adaptation
Article

    Aug 2020

    Lukui Shi
    Lukui Shi
    Ziyuan Wang
    Ziyuan Wang
    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi

It requires pixel-by-pixel annotations to obtain sufficient training data in supervised remote sensing image segmentation, which is a quite time-consuming process. In recent years, a series of domain-adaptation methods was developed for image semantic segmentation. In general, these methods are trained on the source domain and then validated on the...
View
SRDA-Net: Super-Resolution Domain Adaptation Networks for Semantic Segmentation
Preprint

    May 2020

    Enhai Liu
    Enhai Liu
    Zhenjie Tang
    Zhenjie Tang
    Bin Pan
    Bin Pan[...]
    Zhenwei Shi
    Zhenwei Shi

Recently, Unsupervised Domain Adaptation (UDA) was proposed to address the domain shift problem in semantic segmentation task, but it may perform poor when source and target domains belong to different resolutions. In this work, we design a novel end-to-end semantic segmentation network, Super- Resolution Domain Adaptation Network (SRDA-Net), which...
View
Domain Adaptation Based on Correlation Subspace Dynamic Distribution Alignment for Remote Sensing Image Scene Classification
Article

    Apr 2020

    Jun Zhang
    Jun Zhang
    Jiao Liu
    Jiao Liu
    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi

Remote sensing image scene classification refers to assigning semantic labels according to the content of the remote sensing scenes. Most machine learning-based scene classification methods assume that training and testing data share the same distributions. However, in real application scenarios, this assumption is difficult to guarantee. Domain ad...
View
Fig. 1: Overall architecture of the proposed framework with...
Fig. 2: Two major characteristics of high-resolution remote sensing...
Fig. 3: Schematic diagram of forward and back propagation for SSCL...
Fig. 7: Confusion matrices of the proposed SSCL algorithm. (a) UCM data...
+1 Fig. 8: Overall classification accuracies of different scene...
Semi-Supervised Center Loss for Remote Sensing Image Scene Classification
Article
Full-text available

    Mar 2020

    Jun Zhang
    Jun Zhang
    Min Zhang
    Min Zhang
    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi

High-resolution remote sensing image scene classification is a scene-level classification task. Driven by a wide range of applications, accurate scene annotation has become a hot and challenging research topic. In recent years, convolutional neural networks have achieved promising performance among a variety of supervised classification methods. Ho...
View
Figure 2. Multiscale features to spatial features. The red word...
Figure 3. Indian pines dataset, (a) shows the image of first three...
Figure 4. Pavia University dataset, (a) shows the first three...
Fig. 11. Accuracy vary with feature dimension, the spatial feature...
+9 Figure 13. Accuracy with different numbers of training samples on the...
Multiscale Deep Spatial Feature Extraction Using Virtual RGB Image for Hyperspectral Imagery Classification
Article
Full-text available

    Jan 2020

    Liqin Liu
    Liqin Liu
    Zhenwei Shi
    Zhenwei Shi
    Bin Pan
    Bin Pan[...]
    Xianchao Lan
    Xianchao Lan

In recent years, deep learning technology has been widely used in the field of hyperspectral image classification and achieved good performance. However, deep learning networks need a large amount of training samples, which conflicts with the limited labeled samples of hyperspectral images. Traditional deep networks usually construct each pixel as...
View
DSSNet: A Simple Dilated Semantic Segmentation Network for Hyperspectral Imagery Classification
Article

    Jan 2020

    Bin Pan
    Bin Pan
    Xia Xu
    Xia Xu
    Zhenwei Shi
    Zhenwei Shi[...]
    Xianchao Lan
    Xianchao Lan

View
Figure 1. Overall architecture of the proposed method. Firstly, we...
Figure 2. R c and fusion process of different feature points.
Figure 3. Sample images of the UCM dataset: (1) Agricultural, (2)...
Figure 4. Sample images of the AID dataset: (1) Airport, (2) Bare land,...
+8 Figure 5. Sample images of the NWPU dataset: (1) Airplane, (2) Bridge,...
A Multi-Scale Approach for Remote Sensing Scene Classification Based on Feature Maps Selection and Region Representation
Article
Full-text available

    Oct 2019

    Jun Zhang
    Jun Zhang
    Min Zhang
    Min Zhang
    Lukui Shi
    Lukui Shi[...]
    Bin Pan
    Bin Pan

Scene classification is one of the bases for automatic remote sensing image interpretation. Recently, deep convolutional neural networks have presented promising performance in high-resolution remote sensing scene classification research. In general, most researchers directly use raw deep features extracted from the convolutional networks to classi...
View
Fig. 1. Flowchart of CM-MoSU.
A Classification-Based Model for Multi-Objective Hyperspectral Sparse Unmixing
Article
Full-text available

    Aug 2019

    Xia Xu
    Xia Xu
    Zhenwei Shi
    Zhenwei Shi
    Bin Pan
    Bin Pan
    Xuelong Li
    Xuelong Li

Sparse unmixing has become a popular tool for hyperspectral imagery interpretation. It refers to finding the optimal subset of a spectral library to reconstruct the image data and further estimate the proportions of different materials. Recently, multi-objective based sparse unmixing methods have presented promising performance because of their adv...
View
Simultaneous Super-Resolution and Segmentation for Remote Sensing Images
Conference Paper

    Jul 2019

    Sen Lei
    Sen Lei
    Zhenwei Shi
    Zhenwei Shi
    Xi Wu
    Xi Wu[...]
    Hongxun Hao
    Hongxun Hao

View
Tropical Cyclone Intensity Prediction Based on Recurrent Neural Networks
Article

    Jan 2019

    Bin Pan
    Bin Pan
    Xia Xu
    Xia Xu
    Zhenwei Shi
    Zhenwei Shi

The accurate prediction for the tropical cyclone (TC) intensity is a recognised challenge. Researchers usually develop dynamical models to address this task. However, since the TC intensity is highly influenced by various factors such as ocean and atmosphere conditions, it is difficult to build the very model which can explicitly describe the mecha...
View
Analysis for the Weakly Pareto Optimum in Multiobjective-Based Hyperspectral Band Selection
Article

    Jan 2019

    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi
    Xia Xu
    Xia Xu

View
CoinNet: Copy Initialization Network for Multispectral Imagery Semantic Segmentation
Article
Full-text available

    Dec 2018

    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi
    Xia Xu
    Xia Xu[...]
    Xinzhong Zhu
    Xinzhong Zhu

Remote sensing imagery semantic segmentation refers to assigning a label to every pixel. Recently, deep convolutional neural networks (CNNs)-based methods have presented an impressive performance in this task. Due to the lack of sufficient labeled remote sensing images, researchers usually utilized transfer learning (TL) strategies to fine tune net...
View
Multiobjective-Based Sparse Representation Classifier for Hyperspectral Imagery Using Limited Samples
Article

    Aug 2018

    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi
    Xia Xu
    Xia Xu

View
Fig. 1. Binary coding of spectral library. Here we simply suppose the...
Fig. 2. An illustration about the difference of ideal points in the...
Fig. 3. An illustration about the difference between SMoSU and the...
Fig. 4. Endmember missing and redundancy of SMoSU under different k and...
+4 Fig. 5. The concentration process of MOEA/D-SU and SMoSU when the...
ℓ 0 -based sparse hyperspectral unmixing using spectral information and a multi-objectives formulation
Article
Full-text available

    Jul 2018

    Xia Xu
    Xia Xu
    Zhenwei Shi
    Zhenwei Shi
    Bin Pan
    Bin Pan

Sparse unmixing aims at recovering pure materials from hyperpspectral images and estimating their abundance fractions. Sparse unmixing is actually ℓ0 problem which is NP-h ard, and a relaxation is often used. In this paper, we attempt to deal with ℓ0 problem directly via a multi-objective based method, which is a non-convex manner. The characterist...
View
Towards Weakly Pareto Optimal: An Improved Multi-Objective Based Band Selection Method for Hyperspectral Imagery
Conference Paper

    Jul 2018

    Bin Pan
    Bin Pan
    Liming Wang
    Liming Wang
    Xia Xu
    Xia Xu
    Zhenwei Shi
    Zhenwei Shi

View
Robust Sparse Hyperspectral Unmixing Based on Multi-Objective Optimization
Conference Paper

    Jul 2018

    Xia Xu
    Xia Xu
    Liming Wang
    Liming Wang
    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi

View
Single-Sample Aeroplane Detection in High-Resolution Optimal Remote Sensing Imagery
Conference Paper

    Jul 2018

    Bin Pan
    Bin Pan
    Liming Wang
    Liming Wang
    Xinran Yu
    Xinran Yu
    Zhenwei Shi
    Zhenwei Shi

View
Fig. 1. Unfolding an RNN by the BPTT.
Prediction errors (km) of different methods
Results of the experiment with various number of samples in the...
A nowcasting model for the prediction of typhoon tracks based on a long short term memory neural network
Article
Full-text available

    May 2018

    Song Gao
    Song Gao
    Peng Zhao
    Peng Zhao
    Bin Pan
    Bin Pan[...]
    Zhenwei Shi
    Zhenwei Shi

It is of vital importance to reduce injuries and economic losses by accurate forecasts of typhoon tracks. A huge amount of typhoon observations have been accumulated by the meteorological department, however, they are yet to be adequately utilized. It is an effective method to employ machine learning to perform forecasts. A long short term memory (...
View
A supervised abundance estimation method for hyperspectral unmixing
Article

    Apr 2018

    Xia Xu
    Xia Xu
    Zhenwei Shi
    Zhenwei Shi
    Bin Pan
    Bin Pan

Abundance estimation is one of the key steps in hyperspectral unmixing. Usually, abundance estimation is based on linear mixing. However, in real hyperspectral image, this assumption is not physically rigorous enough, because nonlinear mixture may be observed. Nonlinear models present an improvement by considering the microscopic interactions. Howe...
View
Hyperspectral Image Classification Based on Deep Forest and Spectral-Spatial Cooperative Feature
Chapter

    Dec 2017

    Mingyang Li
    Mingyang Li
    Ning Zhang
    Ning Zhang
    Bin Pan
    Bin Pan[...]
    Zhenwei Shi
    Zhenwei Shi

Recently, deep-learning-based methods have displayed promising performance for hyperspectral image (HSI) classification. However, these methods usually require a large number of training samples, and the complex structure and time-consuming problem have restricted their applications. Deep forest, a decision tree ensemble approach with performance h...
View
Longwave infrared hyperspectral image classification via an ensemble method
Article

    Nov 2017

    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi
    Xia Xu
    Xia Xu

Longwave infrared hyperspectral images (LWIR-HSIs) classification is challenging, due to the poor imaging quality and low signal-to-noise ratio. A popular viewpoint is that abundant spatial contextual information can significantly improve the classification accuracies. However, it is quite difficult to determine what degree of spatial information i...
View
Fig. 1. The flowchart of MugNet. We take Indian Pines data set for...
Fig. 2. An illustration for spectral MugNet.
Fig. 6. Grss_dfc_2014 data set. (a) A false composite image with R-G-B...
Fig. 7. Classification accuracies by spectral MugNet on (a) Indian...
+4 Fig. 8. Classification accuracies by spatial MugNet on (a) Indian...
MugNet: Deep learning for hyperspectral image classification using limited samples
Article
Full-text available

    Nov 2017

    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi
    Xia Xu
    Xia Xu

In recent years, deep learning based methods have attracted broad attention in the field of hyperspectral image classification. However, due to the massive parameters and the complex network structure, deep learning methods may not perform well when only few training samples are available. In this paper, we propose a small-scale data based method,...
View
Figure 1. The flowchart of the H 2 F based method.
Table 1 . Classification accuracies of different methods on Indian...
Figure 2. An illustration for the hashing based hierarchical feature...
Figure 3. Cont.
+3 Figure 4. Classification maps by compared methods for Indian Pines data...
Hashing Based Hierarchical Feature Representation for Hyperspectral Imagery Classification
Article
Full-text available

    Oct 2017

    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi
    Xia Xu
    Xia Xu
    Yi Yang
    Yi Yang

Integrating spectral and spatial information is proved effective in improving the accuracy of hyperspectral imagery classification. In recent studies, two kinds of approaches are widely investigated: (1) developing a multiple feature fusion (MFF) strategy; and (2) designing a powerful spectral-spatial feature extraction (FE) algorithm. In this pape...
View
A New Unsupervised Hyperspectral Band Selection Method Based on Multiobjective Optimization
Article
Full-text available

    Oct 2017

    Xia Xu
    Xia Xu
    Zhenwei Shi
    Zhenwei Shi
    Bin Pan
    Bin Pan

Unsupervised band selection methods usually assume specific optimization objectives, which may include band or spatial relationship. However, since one objective could only represent parts of hyperspectral characteristics, it is difficult to determine which objective is the most appropriate. In this letter, we propose a new multiobjective optimizat...
View
Fig. 1. Flowchart of HiFi-We.
Hierarchical Guidance Filtering-Based Ensemble Classification for Hyperspectral Images
Article
Full-text available

    Apr 2017

    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi
    Xia Xu
    Xia Xu

Joint spectral and spatial information should be fully exploited in order to achieve accurate classification results for hyperspectral images. In this paper, we propose an ensemble framework, which combines spectral and spatial information in different scales. The motivation of the proposed method derives from the basic idea: by integrating many in...
View
R-VCANet: A New Deep-Learning-Based Hyperspectral Image Classification Method
Article
Full-text available

    Feb 2017

    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi
    Xia Xu
    Xia Xu

Deep-learning-based methods have displayed promising performance for hyperspectral image (HSI) classification, due to their capacity of extracting deep features from HSI. However, these methods usually require a large number of training samples. It is quite difficult for deep-learning model to provide representative feature expression for HSI data...
View
Hyperspectral Image Classification Based on Nonlinear Spectral–Spatial Network
Article

    Sep 2016

    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi
    Ning Zhang
    Ning Zhang
    Shaobiao Xie
    Shaobiao Xie

View
Fig. 1. Flowchart of the proposed method.
Fig. 2. Candidates selection for (a) data without whitening and (b)...
Fig. 3. Illustration for distance-based vertices extraction. 
Fig. 4. GOCI and HJ-1B data used in this paper. (a) False composite...
+3 Fig. 5. Evaluation for green algae blooms distribution. (a) NDVI on...
A Novel Spectral-Unmixing-Based Green Algae Area Estimation Method for GOCI Data
Article
Full-text available

    Jul 2016

    Bin Pan
    Bin Pan
    Zhenwei Shi
    Zhenwei Shi
    Zhenyu An
    Zhenyu An[...]
    Yi Ma
    Yi Ma

Geostationary Ocean Color Imager (GOCI) data have been widely used in the detection and area estimation of green algae blooms. However, due to the low spatial resolution of GOCI data, pixels in an image are usually “mixed,” which means that the region a pixel covers may include many different materials. Traditional index-based methods can detect wh...
View
Table 1 . Recognition accuracy for different ships (%)
Ship Recognition Based on Active Learning and Composite Kernel SVM
Conference Paper
Full-text available

    Jun 2015

    Bin Pan
    Bin Pan
    Zhiguo Jiang
    Zhiguo Jiang
    Junfeng Wu
    Junfeng Wu[...]
    Penghao Luo
    Penghao Luo

Aiming at recognizing ship target efficiently and accurately, a novel method based on active learning and the Composite Kernel Support Vector Machines (CK-SVM) is proposed. First, we build a ship recognition dataset which contains the major warship models and massive civil ships. Second, to reduce the cost of manual labeling, active learning algori...
View
Improved grey world color correction method based on weighted gain coefficients
Article

    Oct 2014

    Bin Pan
    Bin Pan
    Zhiguo Jiang
    Zhiguo Jiang
    Haopeng Zhang
    Haopeng Zhang[...]
    Junfeng Wu
    Junfeng Wu

Grey world algorithm is a simple but widely used global white balance method for color cast images. However, this algorithm only assumes that the mean values of the R, G, and B components tend to be equal, which may lead to false alarms in some normal images with large areas of single color background, for example, images in ocean background. Anoth...
View
Shadow detection in remote sensing images based on weighted edge gradient ratio
Conference Paper

    Jul 2014

    Bin Pan
    Bin Pan
    Junfeng Wu
    Junfeng Wu
    Zhiguo Jiang
    Zhiguo Jiang
    Xiaoyan Luo
    Xiaoyan Luo

This paper presents a novel shadow detection method in remote sensing images based on edge feature description of candidate regions. Edge gradient ratio is defined and used to represent the inherent properties of shadow regions. To improve the detection result, weighted edge gradient ratio (WEGR) is addressed, where the weight of a region is determ...