H.B. Qi1, Z.L. Li1, J. Chen2

1 - Department of Land Surveying and Geo-Informatics, Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong

2 - National Geomatics Center, Haidian District, Beijing, China


Nowadays the emphasis of GIS has turned from data producing to data updating. Ideally, a mapping organization should only need to update one detailed topographic dataset and then propagates these updates automatically to other less detailed datasets. But new database designs and new generalization methods (such as multiple representation/resolution database and incremental generalization) for this are still in study and have not been used in production. What’s more, as opposed to multiple representation databases, the fact is that existing datasets with different scales in National Mapping Agency are stored in different databases. Traditionally, they were updated respectively by using several kinds of data sources, such as surveys, aerial photographs or satellite images. However, it is a labor intensive, time and money consuming procedure. A promising method is to update larger scale dataset in the traditional way first, and then to update existing smaller scale datasets with the updated dataset. An important step of this method is automated detection of temporal changes between datasets of different scales and times, which is the topic of this paper. Considering that methods for detection of changes are more or less feature/theme dependent and settlement is one of the most active features in topographic database, we confine our research to settlement.

Detection of changes can be defined as the identification and location of discrepancies in the patterns of two temporal datasets. In this paper, discrepancies between datasets are classified into three categories according to their different causes, i.e. error-caused discrepancies, representation-caused discrepancies and change-caused discrepancies. The aim of this paper is to describe an automated method to detect these discrepancies and to distinguish change-caused discrepancy from the other two. In the context of data updating, the main idea of our method is that those discrepancies which may be due to different levels of abstraction by generalization or which is within the limitation of allowable errors are not considered to be temporal changes. In order to do so, discrepancies caused by generalization are analyzed first. After that, six types of discrepancies are identified according to the relationship between objects, and method for computation of degree of these discrepancies is given. With combination of possible cause analysis and the degree of discrepancies, some rules are formulated to supervise the execution of change detection. The new method is tested using real-life data. Results show that our method works well.