Graphical Analysis of Spectral Reflectance Curve
Nguyen Dinh Duong
Institute of Geography ,
National Center for Natural science and technology of Vitnam
Nghiado , Tuliem , Hanoi, Vietnam
Fax: 84-4-8361192, 84-4-8352483
E-mail : duong@igg.ac.vn
Abstract
In the paper the author describes a proposal of an algorithm for automatic classification of multispectral data set . this algorithm called as GASC is being developed under the NASDA ADEOS -II GLI Research Announcement of classification of sex 250m channels of future GLI sensor. Unlike the traditional classification methods that are based on either supervised or unsupervised ( clustering ) principles this algorithm will use graphical invariant such as shape of spectral reflectance curve, angles, area of geometric entities to define features of each land cover object. In case of using Normalised radiance this algorithm offers possibility to automate classification that is very important in processing of huge volume of data observed by sensor Global imager ( GLI) on board of future ADEOS-II satellite. The target of the paper is to explain the nature of the algorithm and to demonstrate preliminary result using LANSAT TM data.
Introduction
Nowadays Earth remote sensing ahs become increasingly developed and widely utilized due to advancement of technology and the need Earth environment monitoring form space . while the third generation of remote sensing satellite that is featured by very high spectral resolution ( up to 3 or 2 meters ) is already on the horizon and open a new phase of Earth observation in detail the need of global earth observation does not lose its significance but increases from day to day. This urgencies driven by recent learning of environmental degradation caused by growing population and over-exploitation of natural resources. In order to understand what is happening on the earth it is necessary to collect major geophysical parameters that are important for understanding the Earth's environment. The future ADEOS-II program with Global Imager sensor ( GLI) is one of missions targeted to fulfil this aim . it was designed to contribute important information to understanding the carbon cycle, estimating primary biomass production, understanding the energy and hydrological cycles and understanding the change in surface processes due to global warming ( Sciences on GLI Mission -NASDA). GLI sensor on board ADEOS-II with 36 spectral channels covering the spectral range from 0.38 to 12 micron with 1600 km swath and 1km and 250m spatial resolution will generate huge information volume( about 394 MB per scene level -1 A product with 250 m resolution and 129 MB per scene for level-1 A product with 1 km resolution- Fujitsu Report of level 1 and Products Description ) its processing and analysis will be time consuming and requires very big computer resource, in this paper the author submit and algorithm for automatic vegetation and land cover classification using six 250m channels of GLI sensor. This algorithm is developed and tested using a data set simulated by LANDSAT TM data. This is preliminary result of research that is being undertaken in the framework of NASDA ADEOS-II Research Announcement .
Graphical analysis of spectral reflectance curve
The nature of GASC algorithm is to find out spectral invariant that will help to easily classify land cover objects according to their spectral reflectance characteristics. This method assumes that different land cover objects has different spectral reflectance pattern. This pattern should be stable for certain remote sensing senor with fixed observation spectral channel composition. The six 250m spectral channels of GLI sensor are equivalent to six channels of well known LANDSAT TM sensor excluding the thermal one. Given a pixel vector.
P ( b
1, b
2, b
3, b
4, b
5, b
6)
b
1 Normalised atmospherically corrected digital count of channel 1
b
2 Normalised atmospherically corrected digital count of channel 2
b
3 Normalised atmospherically corrected digital count of channel 3
b
4 Normalised atmospherically corrected digital count of channel 4
b
5 Normalised atmospherically corrected digital count of channel 5
b
6 Normalised atmospherically corrected digital count of channel 6
then is spectral reflectance curve can be constructed as follows :
Figure 1: Definition of spectral reflectance curve
