Recent developments in hyperspectral images have heightened the need for advanced classification methods. To reach this goal, this paper proposed an improved spectral-spatial method for hyperspectral image classification. The proposed method mainly consists of three steps. First, four band selection strategies are proposed to utilize the statistical region merging (SRM) method to segment the hyperspectral image. The segmentation map is subsequently integrated with the pixel-wise classification method to classify the hyperspectral image. Finally, the final classification result is obtained using the decision fusion rule. Validation tests are performed to evaluate the performance of the proposed approach, and the results indicate that the new proposed approach outperforms the state-of-the-art methods.
Hyperspectral images are generally composed of hundreds to thousands of spectral bands. This rich spectral information can effectively distinguish different objects and physical materials and thus cause broad applications in the mineral detection, environment monitoring, and precision agriculture. The classification technology is currently the predominate method for analyzing hyperspectral images and has received much attention. Over the past decades, numerous pixel-wise classification methods, which only use spectral information, have been proposed to classify remote sensing images. In reviewing the literature, pixel-wise classification methods mainly include maximum-likelihood [
Although pixel-wise classification methods have been researched for years, the spatial information has still not been sufficiently investigated. Generally, the spatial information is important for classification accuracy that can cause decrease of the classifier performance if neglected, particularly for very high spatial resolution satellite images. Previous studies show that pixel-wise methods will sometimes produce classification maps that look noisy (also known as “salt and pepper” effects) if the image spatial information is not used [
Another spectral-spatial scheme includes the postclassification spatial information using a segmentation map. An approach for classifying high spatial resolution urban satellite imagery is based on the different segmentation results of various scales [
Based on the aforementioned analysis, this study presents a new spectral-spatial classification approach for hyperspectral images. The spatial information is obtained from the statistical region merging (SRM) [
The main contributions of this paper are two-folder: proposing a strategy for band selection from the hyperspectral image; proposing a method for spectral-spatial classification using SRM based on the designed band selection strategy.
The remainder of this paper is organized as follows. The spectral-spatial classification using grouping clustering is introduced in Section
The proposed spectral-spatial classification combines advances in SVM classification and SRM segmentation methods. The proposed method has three main steps, as summarized in Figure
Flowchart of the proposed method.
To segment the hyperspectral image, statistical region merging (SRM) [
The original SRM algorithm is used for segmenting color images that contain only three spectral channels and thus cannot directly segment hyperspectral images. Although a minimum heterogeneity rule based SRM method [
Once the
(1) (2) Generate a random number (3) Select the (4) Set (5) Repeat Step 2 to Step 4 until three spectral bands have been selected. (6) Apply SRM to segment the image composed by the selected three spectral bands.
The spectral-spatial classification is performed to postprocess pixel-wise SVM classification result after segmentation results obtained by SRM. In this study, the scheme [
The logic flow of the spectral-spatial classification [
In order to evaluate the performance of the proposed spectral-spatial classification approach, experiments on two hyperspectral images were carried out. The first experiment used a ROSIS image whereas an AVIRIS image was used in the second experiment. In this study, MATLAB with R2010b version was used as the coding environment on a PC that has Intel Core2Quad processor with 2.83-GHz clock speed.
The University of Pavia image is of an urban area recorded by the ROSIS-03 optical sensor, with an image size of 610 × 340 pixels. The image has a spatial resolution of 1.3 m per pixel and the number of spectral bands is 115, which ranges from 0.43 to 0.86
ROSIS image of University of Pavia. (a) False color image. (b) Corresponding reference map.
The supervised classification was firstly created by the multiclass SVM and without feature selection. Table
Class-specific accuracies in percentage for the ROSIS image by different classifiers.
Class | Samples | Method | |||||
---|---|---|---|---|---|---|---|
Train | Test | Pixel-wise SVM (%) | Three PCs (%) |
|
|
|
|
C1 | 252 | 567 | 90.30 | 93.83 | 98.57 |
|
96.17 |
C2 | 135 | 355 |
|
88.02 | 90.08 | 89.82 | 40.70 |
C3 | 720 | 1697 | 92.29 | 98.56 | 98.51 |
|
96.44 |
C4 | 1260 | 2961 | 92.82 | 98.41 |
|
|
89.54 |
C5 | 91 | 214 | 99.41 |
|
99.41 | 99.48 | 98.74 |
C6 | 198 | 463 | 70.00 | 78.70 | 79.33 | 80.13 |
|
C7 | 173 | 323 | 68.65 | 73.42 | 99.67 |
|
86.47 |
C8 | 644 | 1619 | 81.41 | 92.32 |
|
94.89 | 87.42 |
C9 | 513 | 1125 | 96.52 | 98.31 | 99.58 |
|
55.12 |
SRM results on the ROSIS image using different band selection strategies: (a) the first three PCs; (b) uniform distribution; (c)
University of Pavia image classification result: (a) SVM classification result; (b) the first three PCs; (c) majority vote result of uniform; (d) majority vote result of LDA; (e) majority vote result of Entropy.
The spectral-spatial classification was then performed after the segmentation maps were obtained. Figures
Table
Comparison of the SVM and the developed spectral-spatial classification method for University of Pavia image.
Method | OA (%) |
|
|
---|---|---|---|
Static | Pixel-wise SVM | 80.49 | 75.59 |
Three PCs | 86.81 | 83.20 | |
|
|||
Dynamic | Uniform | 88.81 | 85.18 |
LDA | 89.15 | 86.18 | |
Entropy |
|
|
To assess the impact of the presented algorithm on the results of hyperspectral image classification, a comparison was carried out among SVM plus majority vote method (SVMMV) [
Quantitative evaluation of different spectral-spatial classification methods on the ROSIS dataset.
Method | OA (%) |
|
---|---|---|
SVMMV [ |
85.42 | 81.30 |
DSM [ |
87.51 | 85.20 |
DSMw2 [ |
88.76 | 86.22 |
The proposed approach |
|
|
The Indiana Indian Pines hyperspectral image captured by the AVIRIS sensor on June 12, 1992, was used in the second experiment. The data and corresponding true ground data, as shown in Figure
Comparison of class-specific accuracies in percentage for the Indiana image by different methods.
Class | Samples | Method | |||||
---|---|---|---|---|---|---|---|
Train | Test | Pixel-wise SVM (%) | Three PCs (%) |
|
|
|
|
C1 | 422 | 1012 | 84.67 | 88.91 | 92.12 | 93.10 |
|
C2 | 252 | 582 | 73.91 | 72.54 | 93.65 | 94.96 |
|
C3 | 392 | 902 | 95.79 | 97.76 |
|
98.38 | 98.92 |
C4 | 150 | 347 | 92.70 | 93.36 | 95.98 |
|
|
C5 | 198 | 416 | 83.91 | 97.88 | 98.86 | 98.53 |
|
C6 | 232 | 515 | 96.39 | 89.96 | 98.53 |
|
99.06 |
C7 | 150 | 339 |
|
99.18 | 99.18 | 99.39 | 99.39 |
C8 | 277 | 691 | 71.53 | 75.83 |
|
78.10 | 78.31 |
C9 | 52 | 160 | 99.30 | 99.06 | 99.53 |
|
|
C10 | 64 | 170 | 64.44 |
|
75.64 | 79.06 | 78.21 |
C11 | 112 | 268 | 73.99 | 78.16 |
|
96.05 | 96.05 |
C12 | 760 | 1708 | 85.00 | 96.56 | 98.91 | 98.78 |
|
(a) AVIRIS image of Indian Pines (50, 27, and 17). (b) Corresponding reference map.
First, SRM based on four band selection strategies were applied to segment the hyperspectral image, as shown in Figure
Examples of SRM segmentation results using different band selection strategies: (a) the first three PCs; (b) uniform distribution; (c)
Indiana image classification result: (a) SVM classification result; (b) the first three PCA bands; (c) majority vote result of uniform; (d) majority vote result of LDA; (e) majority vote result of Entropy.
For the purpose of quantitative comprise, Table
Comparison of the SVM and the developed spectral-spatial classification method for Indiana image.
Method | OA (%) |
|
|
---|---|---|---|
Static | Pixel-wise SVM | 85.32 | 83.14 |
Three PCs | 90.41 | 88.97 | |
|
|||
Dynamic | Uniform |
|
|
LDA | 95.03 | 94.29 | |
Entropy | 95.05 | 94.32 |
In the second experiment, to assess the suitability of the proposed approach for the classification of hyperspectral images, a comparison analysis was carried out on three other methods (i.e., SVMMV, DSM, and DSMw2). As reported in Table
Quantitative evaluation of different spectral-spatial classification methods on the AVIRIS dataset.
Method | OA (%) |
|
---|---|---|
SVMMV [ |
93.78 | 92.88 |
DSM [ |
90.20 | 88.30 |
DSMw2 [ |
89.50 | 87.50 |
The proposed approach | 95.27 | 94.56 |
In the first experiment, band selection using
This figure shows the visual comparison of spectral-spatial classification of University of Pavia image using different band selection strategies: (a) the first three PCs; (b) uniform distribution; (c)
In the second experiment, the classification method based on
This figure gives a visual comparison of spectral-spatial classification for the Indiana image, using different band selection strategies: (a) the first three PCs; (b) uniform distribution; (c)
An advanced spectral-spatial classification method for classification of hyperspectral images, which combines advances of region-based segmentation and image fusion, has been proposed in this study. The proposed approach has been achieved by (a) integrating pixel-wise support vector machine (SVM) classification and statistical region merging (SRM) segmentation results; (b) multiclassification results fusion using majority voting. Four different band selection strategies have been studied to implement the SRM algorithm to segment the hyperspectral image. The proposed approach has two advantages:
In this study, the spatial information is derived from the region-based segmentation results, which suffers from two main drawbacks:
The authors declare that there is no conflict of interests regarding the publication of this paper.
The work presented in this paper is partly supported by Ministry of Science and Technology of China (Project nos. 2012BAJ15B04 and 2012AA12A305), the National Natural Science Foundation of China (41331175), and Ling Jun Ren Cai Project of National Administration of Surveying, Mapping and Geo-Information, China. The authors would like to thank Professor David A. Landgrebe from Purdue University for providing AVRIS datasets, Professor Paolo Gamba from University of Pavia for providing the ROSIS hyperspectral images, and Dr. Hua Zhang and Dr. Yi Liu from China University of Mining and Technology for their kind discussion and support on computer programming.