Analysis of Tasseled Cap Transformation Features in Winter Land Cover Using Fuyo-1 OPS Data
Classification Techniques
To reduce the variation of illumination with aspect and slope inclination in the image of mountainous forest, the SAM algorithm was employed using the tasseled cap components. Owing to the dependence of greenness on biophysical property of vegetation components must be less sensitive to the problem of variable illumination in the image classification than the original spectral bands, as presented in Carpenter et al. (1997). Here, Meyers (1995) were performed to obtain the components.
In the case of this study, the SAM belongs to a supervised classification technique that determines the spectral signature similarity between the representative spectral mean variables calculated from the training field pixels and the spectral variables derived from each pixel in the image through the spectral angle differences (angular distance in radians and/or steradians) between their vector directions in three dimensional space (see Fig.1)
Fig. 1. Diagram of two-and three-dimensional concept of the SAM classification. Darkly shaded pixels should fall near the origin, the dark point.
Results and Discussion
For the study area the extracted component values from the average SWIR band type of OPS tasseled-cap coefficients are shown in table 1 and their multivariate statistics are presented in table 2.
Table 1. Unvariate statistics and eigen values for the taseled cap transformation images
| |
Brightness |
greenness |
Wetness |
| Average SWIR Band |
Mean |
214.097 |
-27.683 |
-17.256 |
| Std. Dev. |
54.404 |
18.225 |
22.176 |
| Eig.val. |
3229.885 |
365.044 |
188.830 |
Table 2. Eighenvectors of the features derived from OPS tasseled cap transformation coefficients with the average SWIR band. The correlation matrix values are listed in the parentheses.
| |
Brightness |
Greenness |
Wetness |
| Brightness |
-0.954
(1.000) |
-0.111 |
2.278 |
| Greenness |
0.169
(-0.505) |
0.567 |
0.806 |
| Wetness |
0.247
(-0.581) |
-0.816
(0.112) |
-0.522
(1.000) |
In comparing the visual discrimination of the tasseled cap transformation images, they are generally connstent with color tone and displayed area for each feature. Therefore, the brightness, greenness and wenters derived only from the confiscation with the average SWIR band were used for each pixel in the classification stage. Table 3 shows the variation of the main value of the brightness and greenness transforms in the 8 classified categories. The highest greenness value is observed in coniferous forests on the sun-exposed aspects, while the brightness values decrease in the order of assets, and shaded forest. Therefore both transforms are related to shadow and canopy closure. The accuracy of the classification results obtained by using the 480 observation samples( a sample size of 2*2 pixel) with 50-pixel rater size, since the categories of the sample are identified above. Here the maximum like hood algorithm could provide the overall classification accuracy better than the SAM algorithm. Within the mountainous area the site-specific misclassified classes ( i.e., river, urban, classification have 5.8658% whereas those by SAM classification contain 7.4374%, with the exception of the unclassified class (Table4).
Table 3 Mean values and standard deviation of the brightness (B) and greenness (G) transforms in different land cover types
| |
Bare Soil |
Shaded Forest |
River |
Conifer (sun- exposed) |
Hardwood(Sun- exposed) |
Urban |
W.arable Farmland |
W. Fallow land |
B
Mean Std. Dev. |
32.82
31.96 |
107.54
13.64 |
106.61
6.82 |
194.48
10.67 |
217.22
24.26 |
218.73
20.01 |
251.56
14.60 |
318.95
17.49 |
G
Mean Std. dev. |
-62.25
9.04 |
-32.62
3.57 |
-52.97
2.46 |
10.97
6.51 |
-23.55
6.58 |
-59.45
5.82 |
32.34
20.82 |
-40.70
4.01 |
Table 4 Land cover percentage of each class derived from SAM maximum likehood classification and observation sample analysis in the mountainous area selected by mask technique.
| Class label |
SAM(%) |
Maximum Likelihood(%) |
Observation |
| Unclassified |
2.2666 |
|
|
| Bare soil |
14.8928 |
1.0423 |
5.2789 |
| Shaded forest |
2.4199 |
7.8059 |
10.5578 |
| River(water) |
0.3163 |
0.0038 |
|
| Conifer(Sun-exposed) |
9.5366 |
15.1611 |
16.0397 |
| Hardwood(Sun-exposed) |
12.7838 |
19.4622 |
17.4609 |
| Urban(Housing) |
2.6969 |
0.4041 |
|
| Winter arable farmland |
0.449 |
5.5107 |
|
| Winter fallow land |
3.9754 |
0.3072 |
|
| Total |
49.3373 |
49.3373 |
49.3373 |
