Clustering-Based Extraction of Near Border Data Samples for Remote Sensing Image Classification

Bian, Xiaoyong; Zhang, Xiaolong; Yan, Luxin; Li, Bo

doi:10.1007/s12559-012-9147-2

Cited by 13 publications

(8 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Use AP algorithm for the initial clustering of initial data and initialize a cache area; (6). Calculate the density and eigenvector of each cluster built at initial clustering; (7). Reading each dimension attributes of new data x ¼ x 1 ; x 2 ; .…”

Section: Convergence and Computational Complexity Analysis Of Adstreamentioning

confidence: 99%

“…Traditional clustering algorithms have been unable to meet the clustering requirements of dynamic data streams [5,6]. On the one hand, in terms of fitting or predicting future data, we cannot use the learning machine which is trained by historical data to test future data directly like traditional learning problems, as the independent identical distribution hypothesis is not true; on the other hand, from the view of modeling, the probability of the sample set cannot be simply written as the product of each sample's probability, for lacking of independent and identical distribution [7]. In order to cluster the data from data streams, we need to modify and improve traditional theory or method, even propose new clustering algorithms.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An Adaptive Density Data Stream Clustering Algorithm

et al. 2015

View full text Add to dashboard Cite

Now we are in the age of big data. Huge amount of data and information are generated every time. Traditional data stream algorithms are suit for the data streams with low dimension and simple structure. However, with the development of information technology, the produced data streams are becoming more and more complicated. It is particularly important to study how to find new associations and patterns from complex data to achieve the cognition ability and judgment ability like human brain. Clustering data streams with mixed attributes of irregular distribution is a big challenge in data mining. To solve this problem, we present an adaptive density data stream clustering algorithm-ADStream. ADStream is based on the online-off-line clustering framework. It can automatically recognize the initial clusters by passing messages between data points. Then a novel time-decay density clustering strategy is designed to group and update the continuously arriving data streams. Comprehensive experimental results demonstrate that ADStream is adaptive to the evolving data streams and may generate highquality clusters with fast processing rate.

show abstract

Section: Convergence and Computational Complexity Analysis Of Adstreamentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

An Adaptive Density Data Stream Clustering Algorithm

et al. 2015

View full text Add to dashboard Cite

show abstract

“…With rapid advances in storage devices and mobile networks, large-scale multimedia data have become available to ordinary users. However, in practical applications, e.g., human action recognition [17] [18] [19], scene classification [20], text categorization [1], and video annotation and retrieval, the labeled samples always insufficient, though vast amounts of unlabeled samples are readily accessible and provide auxiliary information. Semi-supervised learning (SSL) aiming to exploit both labeled data and the structure imposed by the unlabeled data is designed to address such problem.…”

Section: Introductionmentioning

confidence: 99%

Ensemble p-Laplacian Regularization for Scene Image Recognition

Liu

Tao

et al. 2019

Cogn Comput

View full text Add to dashboard Cite

Recently, manifold regularized semi-supervised learning (MRSSL) received considerable attention because it successfully exploits the geometry of the intrinsic data probability distribution including both labeled and unlabeled samples to leverage the performance of a learning model. As a natural nonlinear generalization of graph Laplacian, -Laplacian has been proved having the rich theoretical foundations to better preserve the local structure. However, it is difficult to determine the fitting graph -Lapalcian i.e. the parameter which is a critical factor for the performance of graph -Laplacian.Therefore, we develop an ensemble -Laplacian regularization (EpLapR) to fully approximate the intrinsic manifold of the data distribution. EpLapR incorporates multiple graphs into a regularization term in order to sufficiently explore the complementation of graph -Laplacian. Specifically, we construct a fused graph by introducing an optimization approach to assign suitable weights on different -value graphs. And then, we conduct semi-supervised learning framework on the fused graph.Extensive experiments on UC-Merced data set demonstrate the effectiveness and efficiency of the proposed method.

show abstract

“…In past decades, many different techniques were presented in the extensive literature for the classification of HRS images, ranging from generative methods to discriminative classifiers based on trained models. For example, Polarization-based [2] and support vector machine (SVM) [3,4] methods have been used. The gray-level co-occurrence matrix (GLCM) is a popular statistical procedure for analyzing texture that is successful in the analysis of remote sensing data.…”

mentioning

confidence: 99%

“…The classifiers of K-nearest-neighbor (K-NN) [12], Gauss process maximum likelihood (GP-ML) [7], and SVM are employed for comparison. A one-against-all [4] classification scheme is adopted for the SVM classification, and a grid searching of fivefold cross validation is performed to obtain optimal regularization parameters C and Gauss kernel width σ. The configuration for MLGBP is the following: P, R parameters are tuned in the range P 8; 16; 32, R 1; 2; 3, respectively.…”

mentioning

confidence: 99%

Spatial–spectral method for classification of hyperspectral images

et al. 2013

Self Cite

View full text Add to dashboard Cite

Spatial-spectral approach with spatially adaptive classification of hyperspectral images is proposed. The rotation-invariant spatial texture information for each object is exploited and incorporated into the classifier by using the modified local Gabor binary pattern to distinguish different types of classes of interest. The proposed method can effectively suppress anisotropic texture in spatially separate classes as well as improve the discrimination among classes. Moreover, it becomes more robust with the within-class variation. Experimental results on the classification of three real hyperspectral remote sensing images demonstrate the effectiveness of the proposed approach.

show abstract

Clustering-Based Extraction of Near Border Data Samples for Remote Sensing Image Classification

Cited by 13 publications

References 18 publications

An Adaptive Density Data Stream Clustering Algorithm

An Adaptive Density Data Stream Clustering Algorithm

Ensemble p-Laplacian Regularization for Scene Image Recognition

Spatial–spectral method for classification of hyperspectral images

Contact Info

Product

Resources

About