3.2 A Conceptual Feature Extraction Model 
Figure 1 depicts the proposed conceptual model for 
feature extraction. It consists of three components. The 
first component (on left-hand side) is comprised of prior 
knowledge models on spectral properties of the objects 
under study and their relationships, edge detection, 
thinning algorithms and geometric and photometric 
properties. The middle component consists of a set of 
procedures for controlling the extraction process. The 
third component contains information or knowledge that 
is specific to the image or application. They are to be 
used as secondary knowledge sets to help to build 
hypotheses for edge linking and boundary tracking. The 
numbers indicate the sequence of the control process. 
Image specific 
knowledge 
Image 
Processing 
Prior 
Knowledge 
models 
‘properties of: | 
: the objects : | 
    
   
| 
| 
| 
| 
| 
i 
| 
| 
Spectral ; i ; 
; 
| 
| 
| 
) 
| 
| 
1 
| 
  
i 
| 
| . 
opt CAS 
e n Image classification Image 
files 1 1 2 Cum, 
i| | 
N ; Digital 3 | 
  
  
  
  
  
  
  
  
  
  
  
  
  
| 
! I 
Image : 
! | Attributes 
UU dge 2 | | ofthe 
detecting ! ( Edge Detection & ) | objects 
ut 
:and thinning: ! thinning | | 
: algorithms : ! 3 | 
we awe wee. ee | 1 
| Spatial 
Edges topological 
structure 
dL em min." i 
"Geometric &: i A 
photometric Edge linking or 
; Properties : boundary tracking 
I 
| 
| 
7 
I 
1 
| | Objects outlines 
with attributes 
; 
| 
| 
| 
Figure 1. The Conceptual Model for Feature Extraction 
  
  
  
! 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
i 
| 
| 
| 
| 
! 
| 
| 
! 
! 
| 
This model incorporates knowledge from remote sensing, 
photogrammetry and computer vision fields. Further 
research is being undertaken to test and to improve the 
model. 
4. CONCLUSIONS 
In this paper, we have proposed an integrated approach 
for feature extraction for high resolution remote sensing 
data, that incorporate approaches derived from the fields 
of photogrammetry, computer vision and remote sensing. 
Essentially the approach combines traditional feature 
extraction methods and thematic classification, to 
produce an output that combines edges and shapes, with 
attribute information. Considerable work is still required 
to develop and test this approach and the authors’ 
research is continuing. 
5. REFERENCE 
Agouris P., and A. Stefanidis, 1996, Automatic 
Extraction of Man-made Objects from Digital Imagery, in 
ASPRS/ACSM Annual Convention & Exhibition, 
Technical Papers, Vol. 1: Remote Sensing & 
Photogrammetry, April 22-25, 1996, Baltimore, 
Maryland, pp. 179-187. 
Albertz, J. and G. Konig, 1991, The Advanced 
Sterophotogrammetric System of the TU Berlin, in 
Digital Photogrammetric Systems, Ebner H,, D, Fritsch 
and C, Heipke (Eds), Wichmann, Karlsruhe, pp.17-27. 
Aplin, P., P.M. Atkinson and P.J. Curran, 1997, Fine 
Spatial Resolution Satellite Sensors for the Next Decade, 
International Journal of Remote Sensing, 18(18), pp. 
3873-3881. 
Bennamoun, M., L. Hapgood, S. Mullens and G. Nicol, 
1997, A New Approach to Edge Detection, The Zero 
Crossing Hybrid Edge Detector, in Proceedings of the 
International Workshop IAIF'97: Image Analysis and 
Information Fusion, Pan H., M. Brooks, D. McMichael 
and G. Newsam (Eds), Cooperative Research Centre for 
Sensor Signal and Information Processing, 6-8 
November, 1997, Adelaide, Australia, pp.119-122 
Beneniktsson, J. and J.R. Seeinsson, 1997, Feature 
Extraction for Multisource Data Classification with 
Artificial Neural Networks, International Journal of 
Remote Sensing, 18(4), pp. 727-740. 
Canny, J.F., 1986, A Computational Approach to Edge 
Detection, IEEE Transactions on Pattern Analysis and 
Machine Intelligence, 8(6), pp. 679-698. 
Danson, F.M., S.E. Plummer and S.A. Briggs, 1995, 
Remote Sensing and the Information Extraction 
Problem, in Advances in Environment Remote Sensing, 
Danson F.M. and S.E. Plummer (Eds.), Johnwiley & 
Sons, New York, pp. 171-177. 
De Gunst, M., 1996, Knowledge-based Interpretation of 
Aerial Images for Updating of Road Maps, Publications 
on Geodesy, New Series Number 44, Netherlands 
Geodetic Commission, Delft, The Netherlands, 
Dong, Y., B.C. Forster and A. Milne, 1997, Segmentation 
of Radar Imagery Using Guaussian Markov Random 
Field Models and Wavelet Transform Techniques, in 
1997 IEEE International Geoscience and Remote Sensing 
248 Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 
g 5 I , 
Syr 
Sin 
Do 
Im: 
For 
Phe 
59‘ 
Fri 
Ph 
Ge 
Gn 
wit