Junshi Xia

Multiple Classifier System for Remote Sensing Image Classification: A Review

Over the last two decades, multiple classifier system (MCS) or classifier ensemble has shown great potential to improve the accuracy and reliability of remote sensing image classification. Although there are lots of literatures covering the MCS approaches, there is a lack of a comprehensive literature review which presents an overall architecture of the basic principles and trends behind the design of remote sensing classifier ensemble. Therefore, in order to give a reference point for MCS approaches, this paper attempts to explicitly review the remote sensing implementations of MCS and proposes some modified approaches. The effectiveness of existing and improved algorithms are analyzed and evaluated by multi-source remotely sensed images, including high spatial resolution image (QuickBird), hyperspectral image (OMISII) and multi-spectral image (Landsat ETM+). Experimental results demonstrate that MCS can effectively improve the accuracy and stability of remote sensing image classification, and diversity measures play an active role for the combination of multiple classifiers. Furthermore, this survey provides a roadmap to guide future research, algorithm enhancement and facilitate knowledge accumulation of MCS in remote sensing community.

Hyperspectral Remote Sensing Image Classification Based on Rotation Forest

In this letter, an ensemble learning approach, Rotation Forest, has been applied to hyperspectral remote sensing image classification for the first time. The framework of Rotation Forest is to project the original data into a new feature space using transformation methods for each base classifier (decision tree), then the base classifier can train in different new spaces for the purpose of encouraging both individual accuracy and diversity within the ensemble simultaneously. Principal component analysis (PCA), maximum noise fraction, independent component analysis, and local Fisher discriminant analysis are introduced as feature transformation algorithms in the original Rotation Forest. The performance of Rotation Forest was evaluated based on several criteria: different data sets, sensitivity to the number of training samples, ensemble size and the number of features in a subset. Experimental results revealed that Rotation Forest, especially with PCA transformation, could produce more accurate results than bagging, AdaBoost, and Random Forest. They indicate that Rotation Forests are promising approaches for generating classifier ensemble of hyperspectral remote sensing.

Information fusion techniques for change detection from multi-temporal remote sensing images

In order to investigate the impacts of different information fusion techniques on change detection, a sequential fusion strategy combining pan-sharpening with decision level fusion is introduced into change detection from multi-temporal remotely sensed images. Generally, change map from multi-temporal remote sensing images using any single method or single kind of data source may contain a number of omission/commission errors, degrading the detection accuracy to a great extent. To take advantage of the merits of multi-resolution image and multiple information fusion schemes, the proposed procedure consists of two steps: (1) change detection from pan-sharpened images, and (2) final change detection map generation by decision level fusion. Impacts of different fusion techniques on change detection results are evaluated by unsupervised similarity metric and supervised accuracy indices. Multi-temporal QuickBird and ALOS images are used for experiments. The experimental results demonstrate the positive impacts of different fusion strategies on change detection. Especially, pan-sharpening techniques improve spatial resolution and image quality, which effectively reduces the omission errors in change detection; and decision level fusion integrates the change maps from spatially enhanced fusion datasets and can well reduce the commission errors. Therefore, the overall accuracy of change detection can be increased step by step by the proposed sequential fusion framework.

Fusion of Difference Images for Change Detection Over Urban Areas

As a result of urbanization, land use/land cover classes in urban areas are changing rapidly, and this trend increased in the recent years. Change information detected from multi-temporal remote sensing images can thus help to understand urban development and to effectively support urban planning. Differences in reflectance spectra, easily obtained by multi-temporal remote sensing images, are important indicators to characterize these changes. Although many algorithms were proposed to generate difference images, the results are usually greatly inconsistent. In order to integrate the merits of different algorithms to recognize spectral changes, fusion techniques merging multiple difference images are proposed and implemented in this paper. Feature and decision level fusion are used to combine simple change detectors, and to build an automatic change detection procedure. The proposed approach is tested with multi-temporal CBERS and HJ-1 images, and experimental results demonstrate its effectiveness and reliability. By integrating different change information, the appropriate fusion method can be selected according to the specific application in order to minimize the omission or the commission errors.

Spectral–Spatial Classification for Hyperspectral Data Using Rotation Forests With Local Feature Extraction and Markov Random Fields

In this paper, we propose a new spectral-spatial classification strategy to enhance the classification performances obtained on hyperspectral images by integrating rotation forests and Markov random fields (MRFs). First, rotation forests are performed to obtain the class probabilities based on spectral information. Rotation forests create diverse base learners using feature extraction and subset features. The feature set is randomly divided into several disjoint subsets; then, feature extraction is performed separately on each subset, and a new set of linear extracted features is obtained. The base learner is trained with this set. An ensemble of classifiers is constructed by repeating these steps several times. The weak classifier of hyperspectral data, classification and regression tree (CART), is selected as the base classifier because it is unstable, fast, and sensitive to rotations of the axes. In this case, small changes in the training data of CART lead to a large change in the results, generating high diversity within the ensemble. Four feature extraction methods, including principal component analysis (PCA), neighborhood preserving embedding (NPE), linear local tangent space alignment (LLTSA), and linearity preserving projection (LPP), are used in rotation forests. Second, spatial contextual information, which is modeled by MRF prior, is used to refine the classification results obtained from the rotation forests by solving a maximum a posteriori problem using the α-expansion graph cuts optimization method. Experimental results, conducted on three hyperspectral data with different resolutions and different contexts, reveal that rotation forest ensembles are competitive with other strong supervised classification methods, such as support vector machines. Rotation forests with local feature extraction methods, including NPE, LLTSA, and LPP, can lead to higher classification accuracies than that achieved by PCA. With the help of MRF, the proposed algorithms can improve the classification accuracies significantly, confirming the importance of spatial contextual information in hyperspectral spectral-spatial classification.

Random Subspace Ensembles for Hyperspectral Image Classification With Extended Morphological Attribute Profiles

Classification is one of the most important techniques to the analysis of hyperspectral remote sensing images. Nonetheless, there are many challenging problems arising in this task. Two common issues are the curse of dimensionality and the spatial information modeling. In this paper, we present a new general framework to train series of effective classifiers with spatial information for classifying hyperspectral data. The proposed framework is based on the two key observations: 1) the curse of dimensionality and the high feature-to-instance ratio can be alleviated by using random subspace (RS) ensembles; and 2) the spatial-contextual information is modeled by the extended multiattribute profiles (EMAPs). Two fast learning algorithms, i.e., decision tree (DT) and extreme learning machine (ELM), are selected as the base classifiers. Six RS ensemble methods, namely, RS with DT, random forest (RF), rotation forest, rotation RF (RoRF), RS with ELM (RSELM), and rotation subspace with ELM (RoELM), are constructed by the multiple base learners. Experimental results on both simulated and real hyperspectral data verify the effectiveness of the RS ensemble methods for the classification of both spectral and spatial information (EMAPs). On the University of Pavia Reflective Optics Spectrographic Imaging System image, our proposed approaches, i.e., both RSELM and RoELM with EMAPs, achieve the state-of-the-art performances, which demonstrates the advantage of the proposed methods. The key parameters in RS ensembles and the computational complexity are also investigated in this paper.

A New Pansharpening Method Based on Spatial and Spectral Sparsity Priors

The development of multisensor systems in recent years has led to great increase in the amount of available remote sensing data. Image fusion techniques aim at inferring high quality images of a given area from degraded versions of the same area obtained by multiple sensors. This paper focuses on pansharpening, which is the inference of a high spatial resolution multispectral image from two degraded versions with complementary spectral and spatial resolution characteristics: 1) a low spatial resolution multispectral image and 2) a high spatial resolution panchromatic image. We introduce a new variational model based on spatial and spectral sparsity priors for the fusion. In the spectral domain, we encourage low-rank structure, whereas in the spatial domain, we promote sparsity on the local differences. Given the fact that both panchromatic and multispectral images are integrations of the underlying continuous spectra using different channel responses, we propose to exploit appropriate regularizations based on both spatial and spectral links between panchromatic and fused multispectral images. A weighted version of the vector total variation norm of the data matrix is employed to align the spatial information of the fused image with that of the panchromatic image. With regard to spectral information, two different types of regularization are proposed to promote a soft constraint on the linear dependence between the panchromatic and fused multispectral images. The first one estimates directly the linear coefficients from the observed panchromatic and low-resolution multispectral images by linear regression while the second one employs the principal component pursuit to obtain a robust recovery of the underlying low-rank structure. We also show that the two regularizers are strongly related. The basic idea of both regularizers is that the fused image should have low-rank and preserve edge locations. We use a variation of the recently proposed split augmented Lagrangian shrinkage algorithm to effectively solve the proposed variational formulations. Experimental results on simulated and real remote sensing images show the effectiveness of the proposed pansharpening method compared with the state-of-the-art.

Rotation-Based Support Vector Machine Ensemble in Classification of Hyperspectral Data With Limited Training Samples

With different principles, support vector machines (SVMs) and multiple classifier systems (MCSs) have shown excellent performances for classifying hyperspectral remote sensing images. In order to further improve the performance, we propose a novel ensemble approach, namely, rotation-based SVM (RoSVM), which combines SVMs and MCSs together. The basic idea of RoSVM is to generate diverse SVM classification results using random feature selection and data transformation, which can enhance both individual accuracy and diversity within the ensemble simultaneously. Two simple data transformation methods, i.e., principal component analysis and random projection, are introduced into RoSVM. An empirical study on three hyperspectral data sets demonstrates that the proposed RoSVM ensemble method outperforms the single SVM and random subspace SVM. The impacts of the parameters on the overall accuracy of RoSVM (different training sets, ensemble sizes, and numbers of features in the subset) are also investigated in this paper.

(Semi-) Supervised Probabilistic Principal Component Analysis for Hyperspectral Remote Sensing Image Classification

In this paper, we have applied supervised probabilistic principal component analysis (SPPCA) and semi-supervised probabilistic principal component analysis (S2PPCA) for feature extraction in hyperspectral remote sensing imagery. The two models are all based on probabilistic principal component analysis (PPCA) using EM learning algorithm. SPPCA only relies on the labeled samples into the projection phase, while S2PPCA is able to incorporate both the labeled and unlabeled information. Experimental results on three real hyperspectral images demonstrate the SPPCA and S2PPCA outperform some conventional feature extraction methods for classifying hyperspectral remote sensing image with low computational complexity.

Spatiotemporal Pattern of PM 2.5 Concentrations in Mainland China and Analysis of Its Influencing Factors using Geographically Weighted Regression

Based on annual average PM2.5 gridded dataset, this study first analyzed the spatiotemporal pattern of PM2.5 across Mainland China during 1998–2012. Then facilitated with meteorological site data, land cover data, population and Gross Domestic Product (GDP) data, etc., the contributions of latent geographic factors, including socioeconomic factors (e.g., road, agriculture, population, industry) and natural geographical factors (e.g., topography, climate, vegetation) to PM2.5 were explored through Geographically Weighted Regression (GWR) model. The results revealed that PM2.5 concentrations increased while the spatial pattern remained stable, and the proportion of areas with PM2.5 concentrations greater than 35 μg/m3 significantly increased from 23.08% to 29.89%. Moreover, road, agriculture, population and vegetation showed the most significant impacts on PM2.5. Additionally, the Moran’s I for the residuals of GWR was 0.025 (not significant at a 0.01 level), indicating that the GWR model was properly specified. The local coefficient estimates of GDP in some cities were negative, suggesting the existence of the inverted-U shaped Environmental Kuznets Curve (EKC) for PM2.5 in Mainland China. The effects of each latent factor on PM2.5 in various regions were different. Therefore, regional measures and strategies for controlling PM2.5 should be formulated in terms of the local impacts of specific factors.