VEGETATION COVER MAPPING 
USING HYBRID ANALYSIS OF IKONOS DATA 
Y. Hirose* *, M. Mori?, Y. Akamatsu* ,Y. Li b 
? Kokusai Kogyo Corporation., Hino Technical Center, 3-6-1 Asahigaoka, Hino, Tokyo 191-0065 Japan — 
(masaru_mori, yoko_hirose, yukio_akamatsu)@kkc.co.jp 
? Japan Space Imaging Corporation, 8-1 Yaesu 2-Chome, Chuo-Ku, Tokyo 104-0028 Japan — 
yunli@spaceimaging.co.jp 
Commission VII, WG VII/3 
KEY WORDS: Vegetation, High Resolution, /KONOS, Classification, Segmentation, Spectral, Texture, Remote Sensing 
c 
ABSTRACT: 
Detailed mapping of vegetation cover types is often required in various environmental study projects. Very high-resolution satellites, 
e.g. IKONOS, are expected to provide a new opportunity to make detailed vegetation cover maps efficiently for large study areas. 
This paper reports the methodology of the vegetation cover mapping using IKONOS data for a watershed management project in 
south-western Japan. In this study, we examined spectral, geometric, and textual properties of image objects extracted through both 
object-based and pixel-based image classification of IKONOS panchromatic, mulit-spectral, and pan-sharpen images of the study 
area to check whether vegetation cover types in the study area are distinguishable using IKONOS data. The result of the analysis 
shows that, in conjunction with spectral information, textural and geometric information of image objects extracted from IKONOS 
data provide useful information for detailed vegetation cover classification. Based on the analysis, we developed a multi-scale 
object-based image analysis method for vegetation cover mapping using IKONOS data. Applying the multi-scale analysis to 
IKONOS images, a vegetation cover map of the study area was obtained. 
Table 1. Classification scheme of the vegetation map of the river 
information data set and the summary of the classification result 
of an IKONOS image via maximum likelihood classification 
1. INTRODUCTION 
  
In Japan, river environment information such as spatial 
distributions of vegetation and wild life habitats along rivers has 
been periodically collected by the ministry of land, 
infrastructure, and transportation since 1990. Along with the 
recent increase in concern over the restoration of the natural 
environment in Japan, information contents of the river 
information data set is now being reviewed by the ministry so 
that it would be able to provide more useful information for 
realizing environmentally appropriate watershed management. 
The requirements for the new data sources of the river 
information data, which would be an alternative for aerial 
photo-interpretation and field survey, include cost-effectiveness, 
continuity, periodicity, and objectivity. Very high-resolution 
satellite images are expected be a useful information source for 
acquiring vegetation information along major rivers in Japan, 
which is part of the environmental data set. 
This paper reports the tentative study of the vegetation cover 
mapping from IKONOS data for a watershed management 
project in south-western Japan, assuming the future use of 
IKONOS data for the nationwide river information acquisition. 
2. CLASSIFICATION SCHEME 
Table 1 shows the classification scheme designed for the 
vegetation mapping in this study. The right column of the table 
summarizes the result of the tentative vegetation mapping from 
IKONOS data using the maximum-likelihood classifier. Based 
on the result, the vegetation mapping for the project seem to 
need both object-based and pixel-based classifier. 
286 
(+: Acceptable, *: Moderate, -: Unacceptable) 
  
Class 
Results of maximum likelihood 
classification of an IKONOS image 
  
Water plant 
+ Field check is necessary 
  
Grass 
+ Sometimes it is confusingly similar 
to deciduous trees 
  
Deciduous tree 
+ Sometimes it is confusingly similar 
to grasses 
  
Coniferous tree 
*Difficult to distinguish from bamboo 
forests 
  
  
  
  
Bamboo *Difficult to distinguish from 
coniferous forests 
Bush + Field check is necessary 
Bare land + Relatively easy to identify 
Orchard * Visual interpretation must be used 
in conjunction with maximum 
likelihood classification 
  
Vegetable field 
* Visual interpretation may be 
needed to ensure the classification 
results 
  
Rice field 
- Difficult to distinguish from 
agricultural fields without imagery 
taken at submerged period 
  
Manmade 
structure 
* Distinction from residential area IS 
problematic 
  
Residential area 
* Distinction from manmade 
structures is problematic 
  
  
Open water 
  
  
+ Relatively easy to distinguish 
  
  
Internatio 
Ven 
Ground re 
patch SIZ 
values a 
distinguis 
vegetatior 
maximum 
informati 
individual 
object-bas 
well as s; 
suitable 1 
vegetatior 
contiguou 
and label 
spatial re 
and textur 
In this s 
from acti 
of the Ni 
western | 
land use 
forests ai