Keywords: SPOT 4 XI, Land cover, Automated classification, Color composite
Abstract The SPOT 4 satellite with short wave infrared band provides a new data source for environmental monitoring and natural resource management. The authors carried out research to develop a new methodology which can fully exploit the advantages of the short wave infrared band. Two issues will be reported in this paper: automated land cover classification and a new color composite model.
The conventional classification methods (supervised or unsupervised) are based on statistical models which use mean vectors, standard deviation and distances such as Euclidean or Mahalannobis as the major classifiers. Different land cover objects have different spectral reflectance properties that can be visualized as a spectral reflectance curve, so it is possible to use this curve as one of the principal measures for classification. The automated classification method developed by the authors uses this spectral reflectance curve along with other quantitative values such as band ratio and band differences for classification. The classification algorithm which is based on graphical analysis of the spectral reflectance curve (GASC) works well with LANDSAT TM data that has 7 spectral bands. SPOT 4 is equipped with a new short wave infrared band at 1.5 mm that provides higher spectral resolution and enhanced sensitivity for leaf moisture content and canopy structure. These improvement is essential for successful application of the GASC algorithm to SPOT 4 XI data in automated classification of land cover. In this paper the authors report on the preliminary results of automated classification using SPOT 4 XI data scene 277/329 acquired on April 24, 2000 near to Hochiminh City, Vietnam.
SPOT 4 XI with 4 spectral bands provides 24 different color composites using the RGB model. Each RGB color composite enhances certain land cover characteristics. However, none of them is capable to display information available in all 4 spectral bands. In this paper the authors report experiment to develop a color composite using all 4 spectral bands. This new color composite is based on data transformation from 4 dimensional conic vector space into 3 dimensional orthogonal space. The transformed components are converted to IHS and RGB space using common algorithms. The new color composite provides more information than any of the conventional ones. The visualized image is an excellent tool for vegetation study and water and infrastructure mapping.
I. Introduction
The SPOT 4 satellite has been launched successfully into orbit on Mar. 24, 1998. From that date the new sensor HRVIR provided new image data for natural resource management and environment monitoring. With new spectral band in short wave infrared region 1.5 - 1.7 mm the HRVIR sensor has broadened application of SPOT data because the SWIR band is particularly sensitive to soil moisture content, vegetation cover and leaf moisture content. The conventional methodology for processing and analysis of multi-spectral remote sensing data, of course, still can be used for SPOT 4 data. However, there is a potential of development of new technique which will help to fully exploit advantages of all four spectral bands of HRVIR sensor. In this paper the authors will present research results on automated classification of land cover and a new color composite model for SPOT 4 XI data. This methodology has been developed in the Environmental Remote Sensing Laboratory, Institute of Geography, Vietnam. SPOT 4 data has been provided by the Satellite Remote Sensing Laboratory, National Central University, Taiwan in the framework of Visiting Scientist Programme.
II. SPOT 4 XI DATA
Image data of SPOT 4 HRVIR is provided in two modes: XS - multispectral mode without SWIR and XI - multispectral mode with SWIR. Depending on processing level, different preprocessing is applied, however, the detector radiometric equalization (MTF enhancement and optional digital dynamic stretching)is always applied for SPOT 4 raw data. Because of variation of ground radiance condition HRVIR sensor applies several gain modes to achieve the best dynamic range of data. Absolute calibration coefficients can be retrieved in the header record of CAP format to compute equivalent radiance at the input of the HRVIR instrument. The gain mode is applied differently for different scenes and different bands of the same scene. This arrangement has caused saturation of image data for some highly reflected objects such as cloud, sand, construction and even bare soil. From this point of view one can expect proper usage of SPOT 4 XI data for interpretation or classification of objects which are not too dark or too bright. Absolute calibration coefficients of some SPOT 4 scenes are shown on Table 1. While gain coefficients for the first three bands are relatively low, band 4 has always very high value of gain coefficient. It is maybe the main reason for digital value saturation of highly reflected objects in band 4. For comparison, digital values of some land cover objects have been read out and shown on Table 2. When compare these values we can see that low reflectance objects
Table 1: Absolute calibration coefficients
| Scene number |
Gain/offset |
| Band 1 |
Band 2 |
Band 3 |
Band 4 |
277/329
2000/03/01 |
1.93500 / 0.0 |
2.28786 / 0.0 |
2.45268 / 0.0 |
13.31878 / 0.0 |
278/321
2000/04/22 |
1.29258 / 0.0 |
1.01000 / 0.0 |
1.08000 / 0.0 |
8.79000 / 0.0 |
278/320
2000/04/22 |
1.29258 / 0.0 |
1.01000 / 0.0 |
1.08000 / 0.0 |
8.79000 / 0.0 |
Table 2: Digital values of some land cover objects
| Objects
| Scene 277/329 |
Scene 278/321 |
| Band 1 |
Band 2 |
Band 3 |
Band 4 |
Band 1 |
Band 2 |
Band 3 |
Band 4 |
| Cloud |
254 |
254 |
254 |
254 |
254 |
254 |
168 |
207 |
| Sand |
215 |
254 |
199 |
254 |
133 |
150 |
74 |
139 |
| Bare soil |
170 |
198 |
133 |
254 |
178 |
214 |
112 |
186 |
| Turbid water |
96 |
96 |
33 |
17 |
94 |
87 |
20 |
28 |
| Clear water |
67 |
48 |
24 |
33 |
54 |
36 |
8 |
17 |
such as turbid or clear water are sensed correctly in dynamic range of one byte integer for both scenes 277/329 and 278/321. However, due to different gain mode some saturation occurred for bare soil and sand in scene 277/329 (high gain mode) while in the scene 278/321 (normal gain mode) they are still in right values. Cloud is always saturated in all gain modes. Readers should be noticed that the right dynamic range of SPOT 4 digital values is from 1 to 254. This fact should be taken into consideration in digital processing SPOT 4 data
