• 대한공간정보학회지
• /
• 제14권4호통권38호
• /
• pp.71-77
• /
• 2006

통계학습이론에 기반하고 있는 Support Vector Machine(SVM)은 구조적 위험 최소화원리를 바탕으로 하는 학습 알고리즘이다. 일반적으로SVM은 비선형 경계를 결정하고 자료를 분류하기 위해서 커널(kernel)을 사용한다. 그러나 기존의 커널들은 두 벡터간의 내적이나 거리차를 이용하여 유사도를 측정하기 때문에 하이퍼스펙트럴 영상분류에 효과적으로 적용될 수 없다. 본 논문에서는 이를 해결하기 위해서 분광유사도커널(Spectral similarity kernel)을 제안한다. 분광유사도 커널은 두 벡터의 거리차와 각 차이를 모두 계산하는 지역적 커널로 하이퍼스펙트럴 영상의 분광특성을 효과적으로 고려할 수 있다. 이를 검증하기 위해서 Hyperion 영상에 polynomial kernel, RBF kernel을 사용한 SVM 분류기와 분광유사도 커널을 사용한 SVM 분류기를 적용하여 토지피복분류를 시행하였다. 분류결과를 통해서 분광유사도 커널을 사용한 SVM 분류기가 정량적, 공간적으로 가장 우수한 결과를 보임을 확인하였다.

• 대한원격탐사학회지
• /
• 제26권6호
• /
• pp.693-703
• /
• 2010

High spectral resolution of hyperspectral data enables analysis of complex natural phenomena that is reflected on the data nonlinearly. Although many manifold learning methods have been developed for such problems, most methods do not consider the spatial correlation between samples that is inherent and useful in remote sensing data. We propose a manifold learning method which directly combines the spatial proximity and the spectral similarity through kernel PCA framework. A gain factor caused by spatial proximity is first modelled with a heat kernel, and is added to the original similarity computed from the spectral values of a pair of samples. Parameters are tuned with intelligent grid search (IGS) method for the derived manifold coordinates to achieve optimal classification accuracies. Of particular interest is its performance with small training size, because labelled samples are usually scarce due to its high acquisition cost. The proposed spatial kernel PCA (KPCA) is compared with PCA in terms of classification accuracy with the nearest-neighbourhood classification method.

• ETRI Journal
• /
• 제38권3호
• /
• pp.540-550
• /
• 2016

Spectral clustering is a powerful tool for exploratory data analysis. Many existing spectral clustering algorithms typically measure the similarity by using a Gaussian kernel function or an undirected k-nearest neighbor (kNN) graph, which cannot reveal the real clusters when the data are not well separated. In this paper, to improve the spectral clustering, we consider a robust similarity measure based on the shared nearest neighbors in a directed kNN graph. We propose two novel algorithms for spectral clustering: one based on the number of shared nearest neighbors, and one based on their closeness. The proposed algorithms are able to explore the underlying similarity relationships between data points, and are robust to datasets that are not well separated. Moreover, the proposed algorithms have only one parameter, k. We evaluated the proposed algorithms using synthetic and real-world datasets. The experimental results demonstrate that the proposed algorithms not only achieve a good level of performance, they also outperform the traditional spectral clustering algorithms.

• ETRI Journal
• /
• 제44권5호
• /
• pp.769-779
• /
• 2022

Spectral clustering has become a typical and efficient clustering method used in a variety of applications. The critical step of spectral clustering is the similarity measurement, which largely determines the performance of the spectral clustering method. In this paper, we propose a novel spectral clustering algorithm based on the local similarity measure of shared neighbors. This similarity measurement exploits the local density information between data points based on the weight of the shared neighbors in a directed k-nearest neighbor graph with only one parameter k, that is, the number of nearest neighbors. Numerical experiments on synthetic and real-world datasets demonstrate that our proposed algorithm outperforms other existing spectral clustering algorithms in terms of the clustering performance measured via the normalized mutual information, clustering accuracy, and F-measure. As an example, the proposed method can provide an improvement of 15.82% in the clustering performance for the Soybean dataset.

• ETRI Journal
• /
• 제35권2호
• /
• pp.311-320
• /
• 2013

In most spectral clustering approaches, the Gaussian kernel-based similarity measure is used to construct the affinity matrix. However, such a similarity measure does not work well on a dataset with a nonlinear and elongated structure. In this paper, we present a new similarity measure to deal with the nonlinearity issue. The maximum flow between data points is computed as the new similarity, which can satisfy the requirement for similarity in the clustering method. Additionally, the new similarity carries the global and local relations between data. We apply it to spectral clustering and compare the proposed similarity measure with other state-of-the-art methods on both synthetic and real-world data. The experiment results show the superiority of the new similarity: 1) The max-flow-based similarity measure can significantly improve the performance of spectral clustering; 2) It is robust and not sensitive to the parameters.

• 한국멀티미디어학회논문지
• /
• 제11권9호
• /
• pp.1302-1309
• /
• 2008

K-means나 퍼지 군집화와 같은 전통적인 군집화 기법들이 원형(prototype)을 기반으로 하고 볼록한 형태의 집단들에 적합한 반면, 스펙트럼 군집화(spectral clustering)는 국부적인 유사성을 기반으로 전역적인 집단을 찾아내는 기법으로 오목한 형태의 집단들에도 적용할 수 있어 커널을 기반으로 하는 SVM과 더불어 각광을 받고 있다. 하지만 SVM이 그러하듯이 스펙트럼 군집화에서도 커널의 폭은 성능에 지대한 영향을 끼치는 요인으로, 이를 결정하기 위한 다양한 방법이 시도되었지만 여전히 휴리스틱에 의존하는 실정이다. 이 논문에서는 유사도 행렬이 보다 명백한 블록 대각 형태를 가지도록 하기 위해 국부적인 커널의 폭을 거리 히스토그램을 바탕으로 적응적으로 결정하는 방법을 제시한다. 제안한 방법은 스펙트럼 군집화에 사용되는 유사도 행렬(affinity matrix)이 블록 형태의 대각 행렬을 이룰 때 이상적인 결과를 낸다는 사실에 기반하고 있으며, 이를 위해서 전통적인 유클리디안 거리와 무작위 행보 거리(random walk distance)를 함께 사용한다. 제안한 방법은 기존의 방법들에서 사용하는 유사도 행렬에 비해 명확한 블록 대각 행렬을 나타내고 있음을 실험 결과를 통해 확인할 수 있다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제6권3호
• /
• pp.941-953
• /
• 2012

A novel and universal method of video image text detection is proposed. A coarse-to-fine text detection method is implemented. Firstly, the spectral clustering (SC) method is adopted to coarsely detect text regions based on the stationary wavelet transform (SWT). In order to make full use of the information, multi-parameters kernel function which combining the features similarity information and spatial adjacency information is employed in the SC method. Secondly, 28 dimension classifying features are proposed and support vector machine (SVM) is implemented to classify text regions with non-text regions. Experimental results on video images show the encouraging performance of the proposed algorithm and classifying features.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제6권4호
• /
• pp.1140-1152
• /
• 2012

A novel and universal method of video image text detection is proposed. A coarse-to-fine text detection method is implemented. Firstly, the spectral clustering (SC) method is adopted to coarsely detect text regions based on the stationary wavelet transform (SWT). In order to make full use of the information, multi-parameters kernel function which combining the features similarity information and spatial adjacency information is employed in the SC method. Secondly, 28 dimension classifying features are proposed and support vector machine (SVM) is implemented to classify text regions with non-text regions. Experimental results on video images show the encouraging performance of the proposed algorithm and classifying features.