| id |
ddss9452 |
| authors |
Koutamanis, Alexander |
| year |
1994 |
| title |
Recognition and Retrieval in Visual Architectural Databases |
| source |
Second Design and Decision Support Systems in Architecture & Urban Planning (Vaals, the Netherlands), August 15-19, 1994 |
| summary |
The development of visual architectural databases is heavily constrained by two technically, practically and conceptually intricate problems, input and retrieval. Input of visual images indifferent forms and from a variety of sources results into computer documents which can only be reproduced and disseminated. Any other use requires extensive annotation of the images with respect to indexing terms and other conceptual structures that make the images identifiable. The bulk of even modest visual databases and the complexity of the images and of the conceptual schemes means that interactive processing is labour-intensive and unreliable. Retrieval also relies on the same processes of annotation and indexing, which make possible the correlation of database contents with user queries. The paper presents the potential of automated recognition for inputting architectural floor plans into visual databases. An optically digitized image is segmented and each segment recognized as an instance of a building element (wall, door, window, etc.). The array ofrecognized elements is then controlled for recognition and segmentation errors. Further processing allows identification of spaces in the floor plan and of their interrelationships. The output of the process is a symbolic array that is much more compact than the original pixel array and also amenable to abstract and /or specific user queries, such as "How many doors are there in the floorplan" or "Which floor plans contain a double loaded corridor". These queries can be input verbally or graphically. Identification of building and spatial elements in a floor plan also allows use of vocabulary control in retrieval: user queries are checked against a thesaurus of architectural terms for accuracy and precision. The user is then presented with options for the improvement of the query before proceeding with identifying relevant entries in the database. Use ofvocabulary control as a search intermediary improves performance and reduces user frustration by making explicit the relevance of a query. |
| series |
DDSS |
| references |
Content-type: text/plain
|
| last changed |
2003/08/07 16:36 |
|
