Full text: XVIIIth Congress (Part B7)

  
  
A WINTER COVER CLASSIFICATION OF LOWER NAKDONG RIVER REGION 
USING JERS-1 OPS DATA 
Choen Kim 
Associcate Professor 
Department of Forest Resources 
Kookmin University 
Seoul 136-702, KOREA 
Commission VII, Working Group 1 
KEY WORDS : Land Cover Classification, Spectral Angle Mapper, Maximum Likelihood Classification, Vegetation Indexes, 
Bidirectional Reflectance Effects, Geometric Illumination Conditions, JERS-1 OPS Multispectral Data 
ABSTRACT : 
Spaceborne multispectral data have been used cost-effectively for monitoring of land surfaces. It is the purpose of this study to 
evaluate new methods for land cover classification of satellite imagery influenced by illumination conditions due to terrain 
irregularities. Low solar elevations cause strong shadow effects in mountainous terrain. In order to examine the effect of 
bidirectional reflection on the reflected radiance response of the JERS-1 OPS bands, the NDVI and TVI were analyzed for 
differential shading. These geometric illumination effects hinder accurate land use classification. This paper shows how the effects 
of differential shading on the classification of vegetation could be reduced through the use of the Spectral Angle Mapper(SAM) 
algorithm. This method proved better than layered classification logic or rule-based expert systems and fuzzy classification with 
nonparametric priors. In addition, maximum likelihood classification was used to distinguish between two different water 
properties (freshwater and salt water) on the images. These areas were not influenced by topographic effects. 
The spectral angle mapper classification method produced seven land cover classes in the lower Nakdong River region : forest, 
urban, winter agricultural area, reed vegetation, barren(rock & sands), river(freshwater) and sea (salt water), and an unclassified 
category which indicated shadow areas in the mountainous forests. 
1. INTRODUCTION 
Until recently, the incorporation of ancillary data and 
fuzzy logic have been used for improving the classification 
accuracy of remotely sensed data. However, these methods are 
applied only to the specific information extraction, require a 
more complex hierarchical process than conventional 
supervised classification method, and are more expensive 
(Wang & Civco, 1992 ; Maselli et al., 1995 ; Jensen, 1996). 
The Spectral Angle Mapper (SAM) method allows single- 
step matching of pixel spectra to reference spectra in n- 
dimensional spectral space. The effects of shadow on the final 
classification image can essentially be eliminated using this 
technique. 
Moreover, this method can also be applied to 
discrimination of vegetated surfaces which reflect incident 
solar radiation anisotropically. The recently observed 
directionality of reflectance in vegetation could be divided into 
forward, nadir and back scatter directions (Deering et al., 1992 
; Hanan et al., 1995). 
These considerations led to the initiation of a research 
project to determine if multispectral data acquired by the 
Japanese Earth Resource Satellite-1 Optical Sensor (JERS-1 
OPS) could be used for classifying land-use and vegetative 
cover type in the lower Nakdong River region. It is anticipated 
that JERS-1 OPS data will provide information on regional 
land cover changes for environmental and ecological 
monitoring, 
  
n He will be a visiting professor at the Department of Forest 
Resources at the University of Idaho, Moscow, ID 83844- 
1133, USA, until August 1996. 
The objectives of this paper are : 
1) To develop a winter land cover classification using the 
Visible 
and Near-Infrared (VNIR) bands of JERS-1 OPS. 
2) To analyze the vegetation indexes (NDVI and TVI) for 
geometric illumination conditions influenced by topography 
(slope, aspect and elevation) and shadows (solar elevation 
and azimuth angle). 
3) To assess the accuracy of the SAM classification in 
comparison with the maximum likelihood method. 
4) To identify tidal spread in Nakdong River, and the 
boundaries of the freshwater in the coast. 
2. STUDY AREA AND DATA 
The study area includes the western part of Pusan, and is 
situated at approximately 35? 12° North latitude, 128° 52° East 
longitude. It is approximately 8027 hectares in size. The area 
is divided into a steep mountain portion a delta plain portion, 
and a river and coastal sea portion. The last portion has the 
characteristics of upstream tidal currents and dispersion of the 
Nakdong River water. The flat portion of the area consist 
mostly of rice fields in summer and dry field agricultural 
activities (including Vinyl(Green) house farming) in winter, 
urban buildings zone, and a wetland with reed vegetation. 
Within the area, the test sites which comprise the 
homogeneous forest canopies of mountainous Morundai (A) 
and the homogeneous hanging reeds near Shinho village (B) 
were specially selected for analysis of the bidirectional effects. 
The dominant overstory tree species in site A is pine 
(Pinus tunbergii Parl.), and the understory trees are oaks 
(Quercus acutissima Carruth, Quercus aliena Bl. and Quercus 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996 
 
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.