Abstract:
Abstract
Construction images are potentially valuable sources of information. However, there are several problems
related to its retrieval mainly relating to understanding of the semantics of the corresponding content. Additionally, the
information is. very much unstructured in nature. These problems have been solved by the development of concepts
enabling the content to be represented in structured form. The framework proposed for this is described. It is based on
the use of domain-specific concepts to retrieve image content and contextual information to improve content
understanding. A mathematical framework is also developed based on a Bayesian approach to make the implementation
possible for construction image classification. This robust framework enables fusion of related domain knowledge with
new information in a versatile manner. An example demonstrates how it is possible to combine information from both
available image data and domain knowledge to understand construction image content.
