Remote sensing clustering analysis based on object-based interval modeling

doi:10.1016/j.cageo.2016.06.006

Computers & Geosciences

Volume 94, September 2016, Pages 131-139

https://doi.org/10.1016/j.cageo.2016.06.006 Get rights and content

Highlights

•
Accurate descriptions of object features are important in object-based clustering.
•
The interval-valued data model describes the object-based clustering prototype features more appropriately.
•
Increasing separability between classes will increase the classification accuracy.
•
Novel object-based adaptive interval-valued fuzzy clustering method makes higher flexibility and efficiency.
•
Applied to high-resolution, multiband, remote sensing data for land cover classification, can also be applied in wider areas.

Abstract

In object-based clustering, image data are segmented into objects (groups of pixels) and then clustered based on the objects' features. This method can be used to automatically classify high-resolution, remote sensing images, but requires accurate descriptions of object features. In this paper, we ascertain that interval-valued data model is appropriate for describing clustering prototype features. With this in mind, we developed an object-based interval modeling method for high-resolution, multiband, remote sensing data. We also designed an adaptive interval-valued fuzzy clustering method. We ran experiments utilizing images from the SPOT-5 satellite sensor, for the Pearl River Delta region and Beijing. The results indicate that the proposed algorithm considers both the anisotropy of the remote sensing data and the ambiguity of objects. Additionally, we present a new dissimilarity measure for interval vectors, which better separates the interval vectors generated by features of the segmentation units (objects). This approach effectively limits classification errors caused by spectral mixing between classes. Compared with the object-based unsupervised classification method proposed earlier, the proposed algorithm improves the classification accuracy without increasing computational complexity.

Introduction

Clustering analysis is a useful tool in remote sensing applications. However, there exists uncertainty in classifications of remotely sensed imagery. For example, there may be a series of uncertainties in the spectral signatures between classes and spectral variation within classes, because of the inherent uncertainty of remote sensing and the many sources of interference (Cheng et al., 2004). These uncertainties indicate that conventional, crisp clustering algorithms do not achieve a correct classification in most cases. Since the 1980s, fuzzy clustering has been extensively studied and successfully applied to remote sensing classification (Ibrahim et al., 2005, Schowengerdt, 2006, Ghosh et al., 2011). The most commonly utilized fuzzy clustering algorithm is the fuzzy c-means (FCM) algorithm. Many researchers have applied FCM to remotely sensed image analyses (Hasi et al., 2004, Xu et al., 2005, Qin and Xu, 2008, Yu et al., 2008, Joel et al., 2011, Bai and Zhao, 2013), and have achieved more satisfactory results than hard classification methods such as k-means and maximum likelihood classification. Standard FCM is based on image pixels, but high-resolution remote sensing images have smaller targets and more information, which leads to greater uncertainties than lower resolution images from the standpoint of land cover classification. Because more details in the high-resolution (more than 10 m) images often make it more difficult to describe a ground object. Therefore, as a pixel-based method, FCM cannot obtain the desired land cover classification results for high-resolution remote sensing images. Consequently, we expect a clustering algorithm that is more resistant to noise and can take advantage of more detailed information. Object-based classification methods for medium to high-resolution remote sensing images can provide a valid alternative to pixel-based methods (Geneletti and Gorte, 2003, Guo et al., 2007, Yang and Zhou, 2011). Yu et al. (2012) recently proposed an unsupervised classification method that adopts the object-based concept to automatically and effectively classify high-resolution remote sensing images. However, it is difficult to extract effective and stable features from the segmentation units, which directly affects the accuracy and stability of the classification results. For instance, the mean spectral signature is typically used to describe a segmentation unit, but this may not appropriately partition two different objects with the same mean value. Intervals are not only utilized to describe the uncertainties in the observed samples, but also utilized to represent a feature's uncertainty in the segmentation units. Partition clustering is a useful tool for interval-valued data analysis, and has been studied in depth in the literature. Adaptive and non-adaptive FCM clustering algorithms for intervals based on different dissimilarity metrics have been proposed (Gao et al., 1999, de Carvalho, 2007, Xie and Wang, 2012), but are not generally put into use. Moreover, de Carvalho (2007) verified that a clustering algorithm with adaptive distances based on a weighted version of the distances (such as the Hausdorff or city-block distances (de Carvalho, 2006)) outperformed the non-adaptive FCM algorithm. To obtain better results for high-resolution, remotely sensed image clustering analysis, we propose an object-based interval modeling method and an adaptive fuzzy clustering algorithm. We have also improved the distance metric for the intervals using the multiband character of remote sensing data. The structure of this paper is as follows. We introduce the interval-valued data model and dissimilarity metric for comparing the interval-valued data in Section 2, and discuss the proposed method in Section 3. Section 4 demonstrates the improved results by our new algorithms, and Section 5 presents our conclusions.

Section snippets

Definition of closed interval-valued data

The term “closed interval-valued data” means that the sample data are within a certain range. A more formal definition is (Moore, 1966): $Definition 1 \tilde{a} = [a^{-}, a^{+}] = {x \in ℜ : a^{-} \leq x \leq a^{+}},$ Where $a^{+} a n d a^{-}$ are real numbers representing the lower and upper bounds of the interval-valued data denoted by $\tilde{a}$ .Thus, $d = a^{+} - a^{-}$ is the width and $m e d = (a^{+} + a^{-}) / 2$ is the mid value of the subinterval, which are both important features of interval-valued data. Many connections between interval algebra and fuzzy theory have been

Methodology

We present an adaptive fuzzy clustering analysis scheme based on object-based interval modeling, to improve the unsupervised classification accuracy for high-resolution remote sensing images. The process flow is shown in Fig. 2. The method includes the following steps.

(1)
Segmentation and object interval modeling. The remotely sensed image is segmented to obtain a series of units with a spatial neighborhood and high homogeneity. The resulting data are utilized for interval-valued modeling.
(2)
Fuzzy

Experimental data

To illustrate the added values of the object-based AIV-FCM algorithm, we conducted land cover classification experiments for high-resolution remote sensing images. We selected three study areas. Two were typical areas with complex land cover in the Pearl River Delta area, an agriculture-focused region of Zhuhai, China, and an intensive cultivation region in the Shenwan district, Zhongshan, China. The third study area was located in Changping district, Beijing, China. We utilized SPOT5 satellite

Conclusion

The uncertainty of information and mixed pixels in remotely sensed image data restricts the effective discrimination of different ground objects. To address this problem, the FCM algorithm based on fuzzy set theory provides a powerful tool for soft classification. However, the added detail increases the difficulty and computational complexity when analyzing high-resolution images. The main contribution of this study is a new solution for the clustering analysis of high-resolution images. The

Acknowledgments

This work was sponsored by the National Natural Science Foundation of China [Grant nos. 11471045, 41272359, 61272364], the Specialized Research Fund for the Doctoral Program of Higher Education in China [Grant no. 20120003110032], the PhD Start-up Fund of Natural Science Foundation of Guangdong Province, China [Grant no. 2014A030310415] and the Fund of Guangdong province Education Bureau, China [Grant nos. 2013LYM_0102, 2014KQNCX240]. The authors would like to thank the reviewers for their

References (43)

F.A.T. de Carvalho
Fuzzy C-means clustering methods for symbolic interval data
Pattern Recognit. Lett.
(2007)
F.A.T. de Carvalho et al.
Adaptive Hausdorff distances and dynamic clustering of symbolic interval data
Pattern Recognit. Lett.
(2006)
F.A.T. de Carvalho et al.
Fuzzy K-means clustering algorithms for interval-valued data based on adaptive quadratic distances
Fuzz Set. Syst.
(2010)
F.A.T. de Carvalho et al.
Unsupervised pattern recognition models for mixed feature-type symbolic data
Pattern Recognit. Lett.
(2010)
R.M.C.R. de Souza et al.
Clustering of interval data based on city-block distances
Pattern Recognit. Lett.
(2004)
A. Ghosh et al.
Fuzzy clustering algorithms for unsupervised change detection in remote sensing images
Inf. Sci.
(2011)
A. Irpino et al.
Dynamic clustering of interval data using a Wasserstein-based distance
Pattern Recognit. Lett.
(2008)
A. Irpino et al.
Dynamic clustering of histogram data based on adaptive squared Wasserstein distances
Expert. Syst. Appl.
(2014)
H.X. Li et al.
Variable universe stable adaptive fuzzy control of a nonlinear system
Comput. Math. Appl.
(2002)
B. Sowmya et al.
Colour image segmentation using fuzzy clustering techniques and competitive neural network
Appl. Soft Comput.
(2011)

M. Xu et al.

Decision tree regression for soft classification of remote sensing data

Remote. Sens. Environ.

(2005)

Y. Bai et al.

HJ-1A satellite remote sensing data classification based on KPCA and FCM

Remote Sens. Land Resour.

(2013)

M. Chavent et al.

New clustering methods for interval data

Comput. Stat.

(2006)

J.C. Cheng et al.

The Uncertainty of Remote Sensing Data

(2004)

de Carvalho, F.A.T., 2006. "A Fuzzy Clustering Algorithm for Symbolic Interval Data Based on a Single Adaptive...

F.A.T. de Carvalho et al.

Dynamic Clustering of interval-valued data based on adaptive quadratic distances

Syst., Man., Cybern. (SMC)-Part A: Syst. Humans

(2009)

R.M.C.R. de Souza et al.

(2004)

T. Dietterich

Approximate statistical tests for comparing supervised classification learning algorithms

Neural Comput.

(1998)

A. Eiumnoh et al.

Band selection for dimension reduction in hyper spectral image using integrated informationgain and principal components analysis technique

Int. J. Mach. Learn. Comput.

(2012)

X.B. Gao et al.

A novel algorithm of FCM clustering for interval valued data

J. Xidian Univ. (Nat. Sci.)

(1999)

D. Geneletti et al.

A method for object-oriented land cover classification combining Landsat TM data and aerial photographs

Int. J. Remote. Sens.

(2003)

Cited by (16)

Adaptive enhanced interval type-2 possibilistic fuzzy local information clustering with dual-distance for land cover classification
2023, Engineering Applications of Artificial Intelligence
Land cover classification of remote sensing image is faced with uncertainties such as “significant difference in category density”, “the same object with different spectra”, and “different objects with the same spectrum”. Existing possibilistic fuzzy clustering-related method still cannot fully meet the interpretation requirements of remote sensing data. To accurately classify remote sensing image with complex geographical distribution, this paper proposes a robust interval type-2 dual-distance driven possibilistic fuzzy clustering motivated by interval-valued number for land cover classification. Firstly, interval-valued data model is established by using the local mean and variance of single-valued data. Secondly, Hausdorff distance and MW distance for interval-valued numbers are introduced to realize the maximum separability of overlapping categories in interval-valued data and generate interval uncertainty sets. To further enhance its robustness, weighted local information factors are constructed by making full use of the generated fuzzy membership and possibilistic typicality, and a novel robust interval type-2 possibilistic fuzzy C-means clustering is proposed. Finally, the adaptive type reduction method is introduced, meanwhile the adaptive expansion of interval-valued data model is realized by using the adaptive contraction expansion control factor. After the iterative algorithm convergences, the accurate classification of covered objects in remote sensing images is realized. Experimental results indicate that the classification performance of the proposed algorithm is better than that of existing interval type-2 fuzzy clustering algorithm and its variants, and it is more suitable for the interpretation of remote sensing images in the actual environments.
A novel interval-valued data driven type-2 possibilistic local information c-means clustering for land cover classification
2022, International Journal of Approximate Reasoning
Citation Excerpt :
(4) Adaptation of interval-valued data and interval type-2 possibilistic sets In the iterative process of single weighting exponent interval type-2 possibilistic clustering with local information for land cover classification, this paper gives a kind of adaptive method for interval-valued data modeling and the type reduction strategy of possibilistic typicality value in view of clustering compactness measure [37]. To verify the superiority of this proposed algorithm, many remote sensing images are tested.
In the process of land cover classification, existing fuzzy clustering is not enough to describe the high-order fuzzy uncertainty, while possibilistic clustering has serious parameter dependence and cluster consistency, which makes them unable to effectively deal with the phenomena of “different objects with the same spectrum” and “different objects with the same spectrum”. Hence, this paper proposes a robust type-2 possibilistic C-means clustering with local information and interval-valued data model for remote sensing land cover classification. Firstly, according to the local neighborhood variance, remote sensing information is modeled as interval-valued data. Secondly, existing possibilistic C-means clustering is modified and an enhanced possibilistic C-means clustering is obtained. Once again, it is used to clustering interval-valued data and a single weighting exponent type-2 possibilistic C-means clustering with double distance measures is constructed. Finally, to further improve the robustness of clustering method, local information is embedded into the objective function of this enhanced type-2 possibilistic C-means clustering and a novel robust possibilistic clustering-related algorithm for remote sensing information classification is proposed. Experimental results show that the proposed algorithm outperforms existing state of the art type-2 clustering-related algorithms, and is of great significance to the interpretation of remote sensing images.
Remote sensing image classification based on semi-supervised adaptive interval type-2 fuzzy c-means algorithm
2019, Computers and Geosciences
Citation Excerpt :
The ultimate goal of remote sensing image classification is to divide the remote sensing image into multiple regions (pixel groups). However, the inherent uncertainty of remote sensing image data and its ubiquitous phenomenon of the “same object with different spectra” (He et al., 2016a,b; Xu et al., 2015) often leads to the ill-posedness of the model used in the algorithm. It further influences the accuracy of the classification results and restricts the development and application of remote sensing technology.
Because of the uncertainty in remote sensing images and the ill-posedness of the problem, it is difficult for traditional unsupervised classification algorithms to create an accurate classification model. In contrast, pattern recognition methods based on fuzzy set theory, such as fuzzy c-means clustering, can manage the fuzziness of data effectively. Of these methods, the type-2 fuzzy c-means algorithm is better able to control uncertainty. Furthermore, semi-supervised training can use prior knowledge to deal with ill-posedness, and hence is more suitable. Therefore, we propose a novel classification method based the semi-supervised adaptive interval type-2 fuzzy c-means algorithm (SS-AIT2FCM). First, by integrating the semi-supervised approach, an evolutional fuzzy weight index m is proposed that improves the robustness and well-posedness of the model used in the clustering algorithm. This makes the algorithm suitable for remote sensing images with severe spectral aliasing, large coverage areas, and abundant features. In addition, soft constraint supervision is performed using a small number of labeled samples, which optimizes the iterative process of the algorithm and determines the optimal set of features for the data. This further reduces the ill-posedness of the model itself. The experimental data consist of three study areas: SPOT5 imagery from Big Hengqin Island, Guangdong, China, and the Summer Palace, Beijing, China, as well as TM imagery from Hengqin Island. Compared with several state-of-the-art fuzzy classification algorithms, our algorithm improves classification accuracy by more than 5% overall and obtains clearer boundaries in remote sensing images with serious mixed pixels. Moreover, it is able to suppress the phenomenon of isomorphic spectra.
Enhanced IT2FCM algorithm using object-based triangular fuzzy set modeling for remote-sensing clustering
2018, Computers and Geosciences
Citation Excerpt :
The visual results in right corner of Fig. 8(h) and (i) indicate that these two pixel based clustering misclassify mariculture area and woodland classes. The visual results of this experiment and recent work (He et al., 2016) have verified that the object-based IV-FCM clustering method yields better performance than other previous methods. Thus, we mainly compare the proposed TFSV-IT2FCM and objected IV-FCM clustering methods in the next two experiments.
Object-based fuzzy clustering method has been widely used to remote-sensing clustering analysis. Mean and interval spectral signatures are typically used to describe an object's features. However, accurately distinguishing two objects with the same mean or interval values and different internal distributions is difficult. Focus on this problem, we developed triangular fuzzy set modeling to describe objects and designed an interval distance metric to measure the dissimilarities between triangular fuzzy sets. Furthermore, using the variation of fuzzifier (two fuzzifiers) to construct interval type-2 fuzzy c-means (IT2FCM) clustering methods, which are sensitive to the choice of fuzzifier, has uncertainty and subjectivity. Thus, an enhanced IT2FCM clustering algorithm that directly adopts the interval distance metric rather than the variation of fuzzifier is proposed for high-resolution remote-sensing clustering. We performed land-cover classification experiments for three study areas by utilizing remote-sensing images from the SPOT-5 and Gaofen-2 satellite sensor, which spatial resolution are approximately 10 m and 1 m, respectively. Visual and numerical results, including Kappa coefficients and the confusion matrix, were utilized to verify the classification results. The experimental results indicated that triangular fuzzy set modeling is appropriate for extracting features from ground objects; moreover, it limits the classification errors caused by same objects with different spectral features. Compared with the object-based interval-valued fuzzy c-means (IV-FCM) method reported in the literature, the proposed algorithm results in improved classification quality and accuracy.
A critical survey of GEOBIA methods for forest image detection and classification
2023, Geocarto International
A Novel Single Fuzzifier Interval Type-2 Fuzzy C-Means Clustering with Local Information for Land-Cover Segmentation
2021, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing

View all citing articles on Scopus

View full text

Case studyRemote sensing clustering analysis based on object-based interval modeling

Highlights

Abstract

Introduction

Section snippets

Definition of closed interval-valued data

Methodology

Experimental data

Conclusion

Acknowledgments

Pattern Recognit. Lett.

Pattern Recognit. Lett.

Fuzz Set. Syst.

Pattern Recognit. Lett.

Pattern Recognit. Lett.

Inf. Sci.

Pattern Recognit. Lett.

Expert. Syst. Appl.

Comput. Math. Appl.

Appl. Soft Comput.

Remote. Sens. Environ.

HJ-1A satellite remote sensing data classification based on KPCA and FCM

Remote Sens. Land Resour.

New clustering methods for interval data

Comput. Stat.

The Uncertainty of Remote Sensing Data

Dynamic Clustering of interval-valued data based on adaptive quadratic distances

Syst., Man., Cybern. (SMC)-Part A: Syst. Humans

Approximate statistical tests for comparing supervised classification learning algorithms

Neural Comput.

Band selection for dimension reduction in hyper spectral image using integrated informationgain and principal components analysis technique

Int. J. Mach. Learn. Comput.

A novel algorithm of FCM clustering for interval valued data

J. Xidian Univ. (Nat. Sci.)

A method for object-oriented land cover classification combining Landsat TM data and aerial photographs

Int. J. Remote. Sens.

Case study
Remote sensing clustering analysis based on object-based interval modeling