| CAD2data was created for a small land surveyor's
office. They needed a database system badly -- searching through
rows of filing cabinets was taking way too much time. When I began
to design a system for them, I initially looked at a conventional
database application, but then I wondered - who has the time to
key in all the existing data? I realized that all the data we need
was in principle contained in the drawings, if that data could
be extracted and parsed.
Our drawings were created on several software packages, by different
draftsmen with different styles - for example, some used title
blocks, some used title blocks inconsistently, and most used
no title blocks at all. However, there were some common features.
A simple example is the text "FIELD DATE:" followed
by a date (actually, not so simple: the heading and/or date may
be plain text or part of a title block; the drawing may be rotated
at an angle; the date may be in any one of a number of formats;
the whole thing may actually be hidden, junk text...).
So I started work on extracting text from drawings, initially
creating a simple linear file-search program (a bit like the
Windows search tool). This program has been in use for several
years now, providing the basis for the next level of functionality.
I decided we wanted a Web-based interface very early, for selfish
reasons -- I hate to deploy, and re-deploy, an application over
multiple PCs during the development process. This decision added
to the initial complexity, but has given the application a foundation
to grow into new areas, like e-commerce.
|
