ISPRS Commission III, Vol.34, Part 3A „Photogrammetric Computer Vision“, Graz, 2002 
  
KNOWLEDGE BASED MULTITEMPORAL INTERPRETATION 
K. Pakzad 
Institute for Photogrammetry and GeoInformation, University of Hannover 
Nienburger Str. 1, 30167 Hannover, Germany - pakzad@ipi.uni-hannover.de 
Commission III, WG III/4 
KEY WORDS: Knowledge Base, Interpretation, Monitoring, Multitemporal, Change Detection, Remote Sensing, Vegetation 
ABSTRACT: 
This paper describes a procedure for an automatic multitemporal interpretation of vegetation areas, which uses both structural 
features and temporal knowledge. For the interpretation of vegetation areas the concept of manual interpretation by using 
interpretation keys was transformed into the automatic interpretation system. For interpretation of temporal changes an approach 
was used, which discretely describes temporal conditions of regions, and which transfers the most probable temporal changes of the 
given conditions as temporal knowledge into a state transition diagram, then using it for multitemporal interpretation. Based on these 
  
approaches a procedure for automatic multitemporal interpretation of industrially used moorland was successfully developed. 
Proceeding from an initial segmentation based on Geo-Data a rese 
gmentation and an interpretation of the segments is carried out for 
each investigated epoch. By using temporal knowledge it is possible to separate moor classes, which can only be detected in 
temporal order. The application of temporal knowledge and structural features enables the exclusive use of grey scale images for 
interpretation of vegetation areas. The results show that the presented procedure is suitable for multitemporal interpretation of 
moorland, and that it is able to distinguish additional moor classes compared to the approaches used so far. It is further applicable for 
a more robust multitemporal interpretation, and does not depend on colour images. 
1. INTRODUCTION 
The method, which has to be used for the interpretation of an 
image relies on the characteristic of the objects, which can be 
found in the scene. Different groups of objects can be 
distinguished. One contains landscape objects, which can be 
recognized by their characteristic unequivocal shapes. To 
recognize them, it is necessary to search for special shapes, 
perhaps in conjunction with a particular radiometry and texture. 
However, the shape of such objects is the most important 
characteristic. 
A second group of objects are those, which have no 
characteristic shape, but can also be recognized by their known 
roughly homogenous texture and radiometry. The landscape 
object forest is an example for such an object. 
Many objects can be described and recognized by the 
composition of object parts, which can be taken from objects of 
the described groups, and which have particular spatial relations 
to each other. Many artificial landscape objects can be 
described in this way. 
However, for interpretation of many vegetation areas the 
distinction from these two groups is not sufficient. Especially 
biotope areas hardly ever show particular shapes or a 
homogenous texture. Nowadays this group of objects and areas 
is manually interpreted. So-called interpretation keys are often 
used for interpretation. They describe characteristic features or 
structures for different objects or areas, which have to be found 
for assigning a special meaning to them. The interpretation keys 
can be divided into selection keys and elimination keys. The 
selection key provides several example images for every known 
class. Interpretation is done by comparison of the example 
images and the examined image parts. The elimination key 
systematically provides features and structures, which have to 
be found in the images. In a first step it describes features and 
A - 234 
structures for the separation of coarse classes. Then the 
description is refined step by step for more and more classes. 
The existing approaches for automatic interpretation of 
landscape objects primarily treat objects from the first two 
groups. The approaches for interpretation of objects and areas 
from the third group mostly use only multispectral classification 
methods. The result of a multispectral classification of such 
objects is an oversegmentation of those object areas, which 
would be interpreted as one area by a human operator. 
The strategy used in this work is based on the described manual 
approach, which applies interpretation keys and finds necessary 
features and structures for a class. Therefore the goal was on the 
one hand to transform the interpretation keys, which in 
Germany already exist for many feature classes (e.g. Von 
Drachenfels, 1994), into the interpretation system and use them 
for the automatic interpretation, and on the other hand to do this 
easily. 
These considerations lead to the following conditions: 
l. The system must enable to save and to use explicit 
knowledge. 
2. As the examined areas could be inhomogeneous, a 
multispectral classification is not suitable to find 
segment borders. 
3. It should be possible to automatically verify the 
interpretation keys in the images by using image 
processing operators. 
Condition 1 was fulfilled by using the knowledge based 
interpretation system AIDA (Automatic Image Data Analyser). 
AIDA uses an explicit knowledge representation by semantic 
nets. À short description will follow in section 2. 
Condition 2 was considered in the strategy by separation of the 
recognition of segment borders and the interpretation of the 
particular segments. A description of the extraction of segment 
borders follows in section 3, and of interpretation in section 6.