| GISdevelopment.net ---> AARS ---> ACRS 2000 ---> Poster Session 1 |
Supervised Classification
of Multi-Temporal Remote Sensing Images
Chi-Farn Chen Yueh-Tan Li
Center for Space and Remote Sensing Research
National Central University
Chungli Li, Taiwan
Tel: (886-)3-4227151-7624 Fax: (886-)3-4254908
E-mail:cfchen@csrsr.ncu.edu.tw
Chi-Farn Chen Yueh-Tan Li
Center for Space and Remote Sensing Research
National Central University
Chungli Li, Taiwan
Tel: (886-)3-4227151-7624 Fax: (886-)3-4254908
E-mail:cfchen@csrsr.ncu.edu.tw
Keywords: multi-temporal remote sensing images, supervised fuzzy classification
Abstract
This paper presents a supervised approach for classifying multi-temporal remote sensing images. One major disadvantage to use the supervised classification of multi-temporal data is that each image is required to select its training data, even the images cover the same area. An attempt to use the fuzzy training method to avoid repeated selection of training data in each image is proposed here. Theoretically, the fuzzy training method is able to deal with the problem of mixture of training classes, therefore, the classification map generated from first-period image can be automatically becoming the fuzzy training sites for second-period image. The proposed approach is tested by a series of simulated multi-temporal images. The results indicate that the method presented here has great potential to extend to the practical applications.
1. Introduction
In processing of multi-temporal remote sensing images, accurate and convenient classification is among one of difficult tasks in practical applications (Baber, 1985). This study aims to use supervised algorithm to classify multi-temporal images. One major procedure of using supervised algorithm is the collection of training data that is relatively time-consuming and labor-intensive (Lillesand, 2000). Furthermore, the image responses may change due to variability in time and space within the multi-temporal images (Richards, 1993). Therefore, the problem that faces the supervised classification of multi-temporal images is that the training data has to be repeatedly selected for each image within the multi-temporal remote sensing data (Schowengerdt,1997). For this reason, a concept is proposed: choosing training set and finishing classification in first-period image, then the training data of the following period images would be automatically generated from the first-period classification image. The most difficult part of the study is that the class positions, numbers, and contents may change in the following-period images. Thus it would generate the complicated mixture of the training classes when it attempts to automatically select the training data for second-period image. This study uses the fuzzy training method (Wang, 1990) to overcome the mixing problem of training classes. Basically, the fuzzy approach allows the heterogeneity to exist within the training sites and may contain mixing classes in the training data. Consequently, the change of class positions and contents in second-period images can be tackled by the characteristics of the fuzzy training data, while the detection of change of class numbers leave a key problem in the process of automated selection of training data. This problem is studied and the solutions are obtained from a series of analysis of fuzzy means and covariance matrix of the training data. A series of simulated data are tested, and the results indicate that the proposed fuzzy training method has the potential to automatically classify multi-temporal remote sensing images.
2. Method
The following section 2.1 discusses the fuzzy training method and the section 2.2 present the multi-temporal supervised classification.
2.1 Fuzzy training
Basically, the fuzzy training method is the training procedure of supervised fuzzy classification. In this procedure, the conventional mean and covariance parameters of training data are represented as a fuzzy set. The following two equations (Equ.1 and Equ.2) describe the fuzzy parameters of the training data:
where µc* is the fuzzy mean of training class c, åc* is the fuzzy covariance of training class c, xi is the vector value of pixel i, fc(xi) is the membership of pixel xi to training class c, n is the total number of pixels of the training data. In order to find the fuzzy mean (Equ.1) and fuzzy covariance (Equ.2) of every training class, it must know the membership of pixel xi to training class c first. In this study, the membership function is defined based on the conventional maximum likelihood classification algorithm with fuzzy mean and fuzzy covariance.
where
fc(xi) is the membership of pixel xi to class c, Pc*(xi) is the maximum likelihood probability of pixel xi to class c, m is the number of classes, n is the number of the bands. These equations would ultimately produce the fuzzy mean and fuzzy covariance for each training class. Consequently, the membership values would be used to describe the mixing classes in every training site.
2.2. The supervised fuzzy classification of multi-temporal images
As long as the necessary fuzzy mean and fuzzy covariance of each training class are calculated from the procedures described at the previous section 2.1, a fuzzy supervised classification can be implemented at first-period image of multi-temporal data. Accordingly, the membership values of each pixel calculated from Equ.3 can be used to generate a classification map. Then the classification map is overlaid with second-period image, and the positions of the classes can be used as the reference to collect the training data. However, the problem always arises when second-period image has some variations in the class positions, contents, and numbers. The fuzzy training method with the capability to mix the training classes, in fact, would successfully manage the variation of the class positions and contents as long as the class numbers stays the same. It appears that the class numbers of second-period image has to be decided before the fuzzy supervised classification can be applied to second-period image. The values of fuzzy covariance and fuzzy mean of training data actually provide some answers to the problem of the class numbers. Accordingly, the threshold techniques are used to obtain the change of the class numbers. A high fuzzy covariance threshold will suggest an increase of the class numbers, while a high distance threshold between different fuzzy means will indicate a decrease of the class numbers. Therefore, with the class numbers known, the fuzzy training and classification will bring back to train and classify second-period image.
3. Test Data and the Results
The class variation of second-period image basically can be grouped into class positions, contents, and numbers. Their possible combination would be summarized to five cases, which is described as follows.
A series of images to simulate above five cases is generated for testing the proposed method. The following is the testing results and their discussions.
Case 1:
Fig.1 (a) simulated 1st-period image; (b) 2nd-period image (case 1); (c) classification image
Case 1 represents the situation with 'no-change' in both class numbers and class contents, but 'change' in class positions. The testing results are showing in Figure 1. The visual inspection (Fig.1(b) and (c)) and 96% overall accuracy indicate a successful classification.
Case 2
Fig.2 (a) simulated 1st-period image; (b) 2nd-period image (case 2); (c) classification image
Case 2 represents the situation with 'no-change' in both class numbers and class positions, but 'change' in class contents. The testing results are showing in Figure 2. The visual inspection (Fig.2(b) and (c)) and 98% overall accuracy indicate a successful classification.
Case 3
Fig.3 (a) simulated 1st-period image; (b) 2nd-period image (case 3); (c) classification image
Case 2 represents the situation with 'no-change' in class numbers, but 'change' in both class contents and class positions. The testing results are showing in Figure 3. The visual inspection (Fig.3(b) and (c)) and 99% overall accuracy indicate a successful classification.
Case 4
Fig.4 (a) simulated 1st-period image; (b) 2nd-period image (case 4); (c) classification image
Case 4 represents the situation with 'increase' in class numbers, and 'change' in both class contents and class positions. The testing results are showing in Figure 4. The visual inspection (Fig.4(b) and (c)) and 97% overall accuracy indicate a successful classification.
Case 5
Fig.5 (a) simulated 1st-period image; (b) 2nd-period image (case 5); (c) classification image
Case 5 represents the situation with 'decrease' in class numbers, and 'change' in both class contents and class positions. The testing results are showing in Figure 5. The visual inspection (Fig.5(b) and (c)) and 99% overall accuracy indicate a successful classification.
4. Conclusion
This study proposes a supervised fuzzy classification for multi-temporal remote sensing images. The class variation between multi-temporal images normally requires the selection of training data for every image when performing classification. An automatic procedure to generate the training data for multi-temporal image classification is presented here. The procedure combines the classification map from first-period image and the fuzzy training method to automatically collect the training data for second-period image. A series of simulated images are created for testing the proposed method. Their results suggest that the practical application of the method to the multi-temporal remote sensing images can be expected.
References
Abstract
This paper presents a supervised approach for classifying multi-temporal remote sensing images. One major disadvantage to use the supervised classification of multi-temporal data is that each image is required to select its training data, even the images cover the same area. An attempt to use the fuzzy training method to avoid repeated selection of training data in each image is proposed here. Theoretically, the fuzzy training method is able to deal with the problem of mixture of training classes, therefore, the classification map generated from first-period image can be automatically becoming the fuzzy training sites for second-period image. The proposed approach is tested by a series of simulated multi-temporal images. The results indicate that the method presented here has great potential to extend to the practical applications.
1. Introduction
In processing of multi-temporal remote sensing images, accurate and convenient classification is among one of difficult tasks in practical applications (Baber, 1985). This study aims to use supervised algorithm to classify multi-temporal images. One major procedure of using supervised algorithm is the collection of training data that is relatively time-consuming and labor-intensive (Lillesand, 2000). Furthermore, the image responses may change due to variability in time and space within the multi-temporal images (Richards, 1993). Therefore, the problem that faces the supervised classification of multi-temporal images is that the training data has to be repeatedly selected for each image within the multi-temporal remote sensing data (Schowengerdt,1997). For this reason, a concept is proposed: choosing training set and finishing classification in first-period image, then the training data of the following period images would be automatically generated from the first-period classification image. The most difficult part of the study is that the class positions, numbers, and contents may change in the following-period images. Thus it would generate the complicated mixture of the training classes when it attempts to automatically select the training data for second-period image. This study uses the fuzzy training method (Wang, 1990) to overcome the mixing problem of training classes. Basically, the fuzzy approach allows the heterogeneity to exist within the training sites and may contain mixing classes in the training data. Consequently, the change of class positions and contents in second-period images can be tackled by the characteristics of the fuzzy training data, while the detection of change of class numbers leave a key problem in the process of automated selection of training data. This problem is studied and the solutions are obtained from a series of analysis of fuzzy means and covariance matrix of the training data. A series of simulated data are tested, and the results indicate that the proposed fuzzy training method has the potential to automatically classify multi-temporal remote sensing images.
2. Method
The following section 2.1 discusses the fuzzy training method and the section 2.2 present the multi-temporal supervised classification.
2.1 Fuzzy training
Basically, the fuzzy training method is the training procedure of supervised fuzzy classification. In this procedure, the conventional mean and covariance parameters of training data are represented as a fuzzy set. The following two equations (Equ.1 and Equ.2) describe the fuzzy parameters of the training data:
where µc* is the fuzzy mean of training class c, åc* is the fuzzy covariance of training class c, xi is the vector value of pixel i, fc(xi) is the membership of pixel xi to training class c, n is the total number of pixels of the training data. In order to find the fuzzy mean (Equ.1) and fuzzy covariance (Equ.2) of every training class, it must know the membership of pixel xi to training class c first. In this study, the membership function is defined based on the conventional maximum likelihood classification algorithm with fuzzy mean and fuzzy covariance.
where
fc(xi) is the membership of pixel xi to class c, Pc*(xi) is the maximum likelihood probability of pixel xi to class c, m is the number of classes, n is the number of the bands. These equations would ultimately produce the fuzzy mean and fuzzy covariance for each training class. Consequently, the membership values would be used to describe the mixing classes in every training site.
2.2. The supervised fuzzy classification of multi-temporal images
As long as the necessary fuzzy mean and fuzzy covariance of each training class are calculated from the procedures described at the previous section 2.1, a fuzzy supervised classification can be implemented at first-period image of multi-temporal data. Accordingly, the membership values of each pixel calculated from Equ.3 can be used to generate a classification map. Then the classification map is overlaid with second-period image, and the positions of the classes can be used as the reference to collect the training data. However, the problem always arises when second-period image has some variations in the class positions, contents, and numbers. The fuzzy training method with the capability to mix the training classes, in fact, would successfully manage the variation of the class positions and contents as long as the class numbers stays the same. It appears that the class numbers of second-period image has to be decided before the fuzzy supervised classification can be applied to second-period image. The values of fuzzy covariance and fuzzy mean of training data actually provide some answers to the problem of the class numbers. Accordingly, the threshold techniques are used to obtain the change of the class numbers. A high fuzzy covariance threshold will suggest an increase of the class numbers, while a high distance threshold between different fuzzy means will indicate a decrease of the class numbers. Therefore, with the class numbers known, the fuzzy training and classification will bring back to train and classify second-period image.
3. Test Data and the Results
The class variation of second-period image basically can be grouped into class positions, contents, and numbers. Their possible combination would be summarized to five cases, which is described as follows.
| class variation | Class numbers | Class contents | Class positions |
| Case 1 | no-change | no-change | change |
| Case 2 | no-change | change | no-change |
| Case 3 | no-change | change | change |
| Case 4 | increase | change | change |
| Case 5 | decrease | change | change |
A series of images to simulate above five cases is generated for testing the proposed method. The following is the testing results and their discussions.
Case 1:
 |
||
| (a) | (b) | (c) |
Fig.1 (a) simulated 1st-period image; (b) 2nd-period image (case 1); (c) classification image
Case 1 represents the situation with 'no-change' in both class numbers and class contents, but 'change' in class positions. The testing results are showing in Figure 1. The visual inspection (Fig.1(b) and (c)) and 96% overall accuracy indicate a successful classification.
Case 2
 |
||
| (a) | (b) | (c) |
Fig.2 (a) simulated 1st-period image; (b) 2nd-period image (case 2); (c) classification image
Case 2 represents the situation with 'no-change' in both class numbers and class positions, but 'change' in class contents. The testing results are showing in Figure 2. The visual inspection (Fig.2(b) and (c)) and 98% overall accuracy indicate a successful classification.
Case 3
 |
||
| (a) | (b) | (c) |
Fig.3 (a) simulated 1st-period image; (b) 2nd-period image (case 3); (c) classification image
Case 2 represents the situation with 'no-change' in class numbers, but 'change' in both class contents and class positions. The testing results are showing in Figure 3. The visual inspection (Fig.3(b) and (c)) and 99% overall accuracy indicate a successful classification.
Case 4
 |
||
| (a) | (b) | (c) |
Fig.4 (a) simulated 1st-period image; (b) 2nd-period image (case 4); (c) classification image
Case 4 represents the situation with 'increase' in class numbers, and 'change' in both class contents and class positions. The testing results are showing in Figure 4. The visual inspection (Fig.4(b) and (c)) and 97% overall accuracy indicate a successful classification.
Case 5
 |
||
| (a) | (b) | (c) |
Fig.5 (a) simulated 1st-period image; (b) 2nd-period image (case 5); (c) classification image
Case 5 represents the situation with 'decrease' in class numbers, and 'change' in both class contents and class positions. The testing results are showing in Figure 5. The visual inspection (Fig.5(b) and (c)) and 99% overall accuracy indicate a successful classification.
4. Conclusion
This study proposes a supervised fuzzy classification for multi-temporal remote sensing images. The class variation between multi-temporal images normally requires the selection of training data for every image when performing classification. An automatic procedure to generate the training data for multi-temporal image classification is presented here. The procedure combines the classification map from first-period image and the fuzzy training method to automatically collect the training data for second-period image. A series of simulated images are created for testing the proposed method. Their results suggest that the practical application of the method to the multi-temporal remote sensing images can be expected.
References
- Baber, M. L., D. Wood, and R. A. McBride, 1985. Classification of Corn and Soybeans Using Multitemporal Thematic Mapper Data. Remote Sensing of Environment, Vol. 16, pp.175-181.
- Lillesand, T. M., and R. W. Kiefer, 2000. Remote Sensing and Image Interpretation. 4th edition, John Wiley & Sons, Inc,.
- Richards, J. A., 1993. Remote Sensing Digital ImageAnalysis: An Introduction. 2nd Ed., Springer-Verlag Berlin Heidelberg.
- Schowengerdt, R. A., 1997. Remote Sensing: Models and Methods for Image Processing . 2nd Edition, Academic Press, pp. 522.
- Wang, F., 1990. Fuzzy Supervised Classification of Remote Sensing Images. IEEE Trans. on Geoscience and Remote Sensing, GE-28(2), pp. 194-201.