IMPROVED CLASSIFICATION OF SPOT MULTI-SPECTRAL IMAGES 
FOR LAND-COVER TYPES EVALUATION ASSISTED BY 
DIGITAL ELEVATION MODEL (DEM) AND AERIAL PHOTOGRAPHS, 
A CASE STUDY 
Saeid Noori Bushehri and Nooshin Khorsandian 
Department of GIS, Department of Photogrammetry, 
Assistant Director of GIS Dept., Remote Sensing Specialist 
National Cartographic Center (N.C.C) 
Tehran, Iran 
Commission VII, Working Group 2 
KEY WORDS: Remote Sensing, L.and-Use, Classification, Image, SPOT, Multispectral, Spectral. 
ABSTRACT: 
The SPOT scene available for field work site, annually used by PHM3 & CAR3 students of ITC, was classified. The 
classified part is our area of interest for improved classification, whose existing classified map has been used as the 
ground truth. In conjunction with this, topographic data has been collected from mosaicked orthophotos which are 
subsequently digitized and overlaid to the composite SPOT image. The so formed topographic network superimposed 
to the landuse parcels visible on the satellite images is basically our tool for classification improvement as per case 
study title. 
Assessment carried out to compare and contrast the conventional against improved classifications depicts that there is 
a great role played by introduction of both mosaicked orthophoto from scanned aerial photo imageries and DEM 
since the area in question is mountainous and of various landuse. 
1. INTRODUCTION 
The great advantage of having data available digitally is 
that it can be processed by computer either for machine 
assisted information extraction or for embellishment 
before an image product is formed. Remotely sensed 
image data of the earth's surface acquired from either 
aircraft or spacecraft platforms is readily available in 
digital format; specially the data is composed of discrete 
picture elements or pixels and radiometrically it is 
quantized into discrete brightness levels. Even data that 
is not recorded in digital form initially can be converted 
into discrete data by use of digitizing equipment such as 
scanners. 
One of the applications of remotely sensed digital data, 
specially satellite imagery, is using them in /and cover 
classification. There is two methods. namely photo 
interpretation and quantitative analysis. Photo 
interpretation is done by expenenced operators to 
identify ground features and overall land cover. 
Successful interpretation needs high quality images and 
expert operators. Since there is a need of supervision of 
human, it is too difficult and/or impractical to work in 
pixel level. We have to use computer to evaluate the 
satellite imagery with several bands in pixel level and 
with high radiometric resolution for accurate results of 
analysing. Computer assisted interpretation of remotely 
534 
sensed data is called quantitative analysis. In this 
method, at first land cover types or spectral classes ate 
defined by the operators (users of the images), then by 
using of ground truth (data collected at the feld 
operations) sample land cover types are identified in 
limited parts of images. In fact, we try to teach the 
computer the nature of different land cover types and 
their spectral specifications (training). In the next step, 
the trained computer starts to identify land cover types 
in the rest of images, namely a label is assigned to each 
pixel due to its spectral value. This kind of classification 
is called c/assic or conventional classification. 
Present article is the result of a case study of authors at 
the end of a course, "Integrated Mapping and 
Geoinformation Production 3 (IGP3)," in 1993 at ITC, 
The Netherlands. The aim of the case study was to 
improve the conventional classification method by use 
of scanned aerial photos and existing DEM (Digital 
Elevation Model) of area. 
Geometric distortions in satellite images can be related 
to number of factors, including 
(1) the rotation of the earth during image 
acquisition, 
(ii) the wide field of view of some sensors, 
(iii) the curvature of the earth, 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996 
  
D (C) o» CC Noel UG Tra 
a DD N 
El f| 4]. HI d$) Ln —)0 Coke Rt "Le T 
I 83