TWENTY-FIVE YEARS AND STILL COUNTING:

THE RASTER TO CAD SAGA!

An update on Scan to CAD Conversion

by Richard N. Stover

Prelude

After tilting at windmills for more years than I care to acknowledge, I have tired of the whole issue. But just today, someone again called me to get my sense of the size of the "raster to vector" market. "What market(?)" is always my short response but then as I ramble on about the various fragments that still exist, I've got to concede that a small niche market exists. Awake to reality, here are the facts.

Five years ago, I was invited to speak to the Intergraph User's Group on this subject and I used the above title, with "twenty years" at that time. Much of that presentation was then reduced to writing for a chapter in the CAD/CAM Handbook, 1996 edition, edited by Carl Machover and published by McGraw-Hill entitled "The Role of Scanners, Automated Data Capture, and Document Storage and Distribution." I invite anyone who would like to read the history of this technology and the background to this update, to pickup a copy of this book (ISBN 0-07-039375-3).

The Crux of the Problem Twenty-five years ago I presided over a company that developed software for raster to vector conversion. In the years since that time, many companies have developed and marketed such software. Converting raster to vector lines is not the problem. CAD systems need more than that, specifically an associated and properly connected (intelligent) database. How does one achieve this? By creating CAD compatible elements or entities with associated identifying information (metadata) and providing the proper connectivity information between all of the elements. And to boot, the connectivity information must provide the proper order of assembling the elements! It sounds relatively simple, but try to describe those steps in software, to cover all of the real world variations on drawings, and you've got an impossible task. Most drawings do not provide complete information on what is to be accomplished from that drawing, let alone interpret the information automatically.

What Approaches Do Work?

Having an engineer sit at a CAD workstation, interpreting the drawing while re-entering the information from a menu of CAD entities and commands, is the one of the proven ways of converting drawings. Where do you find engineers who will do quantities of such work? In offshore countries, especially India, a land of many English-speaking engineers with AutoCAD training. Many drawings have been converted at relatively low costs by conversion vendors who employ such offshore engineers. Another variation of CAD redrawing is to first scan the drawing and display the raster image on the computer screen. The operator then retraces the image using the CAD tools. This approach was first marketed as CAD Overlay about 10 years ago by Image Systems Technology, a company later acquired by Softdesk which was later acquired by Autodesk. Known as Heads-up Digitizing as opposed to "head bent over a drawing" digitizing, this technique is also used by the offshore facilities.

The next approach, not usually used to convert entire drawings, also uses scanning of drawings and more sophisticated raster editing that includes localized raster to vector conversion under the interactive control of an operator. By working on smaller portions of a drawing, it is possible to convert small drawings or smaller parts of a larger drawing which for certain applications may be sufficient. For example, an existing as-built, 2-D facility drawing may be modified with the changes made in vector and the remainder of the drawing left in raster. Here the object is to store a hybrid (raster and vector image) file with the option of inserting the vector changes into the CAD drawing file at some later time. Software vendors that provide such solutions include Informative Graphics Corp., GTX Corp., Spicer Corp., and Arbor Image.

Approaches That Work for Specific Drawing Types

In the CAD/CAM Handbook referenced above, I outlined 5 levels of intelligence associated with scanned drawings, the applications for each level, and the companies that supplied solutions for each at the time the article was prepared. (This will provide some perspective about the companies and the changing market.) The first level is the raster image (compressed or not) that is not converted but can be stored as a separate CAD layer for viewing purposes. The others are: Intelligence Level 2: Raster image with associated ASCII text and vectors (hybrid files). Companies supplying these solutions use methods including heads-up digitizing (covered under the previous heading), and selective interactive conversion (selected by line, object, area). Five years ago, vendors included Cimage, Formtek, Intergraph, ALPHAREL/Optigraphics (now Altris Software), GTX, Softdesk Imaging Group (now Autodesk), and Arbor Image. (Note from the current listing under the previous heading how vendor directions have changed based on a smaller, niche market.)

Intelligence Level 3: Connected Vectors with no attached intelligence. This is essentially a 2-D vector image that many CAD systems can accept for viewing purposes but not for integration into the CAD database. (Due to the limited use, no market has developed for this capability.)

Intelligence Level 4: Continuous Vector Strings With Tagged Values. This technique involves interactive/selective line following with the ability to enter and tag information to the line. Mapping applications, primarily contours and land-use information require this level of intelligence. Vector strings of contour information can be tagged with elevation values, for example. Companies providing this five years ago included: Scangraphics, Laser Scan Ltd., Intergraph, ALPHAREL/Optigraphics (now Altris Software), and Information and Graphic Systems (now Hitachi). (While these companies still exist, not all still supply this kind of product.)

Intelligence Level 5: Associated Intelligent Database. Intelligence level 5 works best on schematics and other one line diagrams such as maps, piping and instrumentation diagrams (P&IDs) (such as for nuclear power plants). While the solutions provided will use varying amounts of human interaction, the following companies provided solutions five years ago. AUDRE (now Vector Systems) provides a pattern matching solution. (Within the past year, this company was acquired, has taken a low profile and may be doing other things.) Coherent Radiation provides a solution which may use some pattern matching and vectorized rule bases. (Coherent was focused on building large GIS databases and has taken a low profile in recent years.)

The Electric Power Research Institute (EPRI) had funded GTX Corporation to develop a solution using vectorized rule bases for nuclear electric utilities. (EPRI is now offering such solutions for converting P&IDs for nuclear power plants.) Intergraph Corporation has developed a solution using vectorized rule bases for the French Telecom which may later be commercialized. (Intergraph has taken a low profile on this kind of product in recent years.)

Dimension 4, not on the list five years ago, is the latest company based in Houston, TX, that uses an approach developed at IBM-Norway to convert P&IDs and other one-line drawings for 3-D modeling applications. Dimension 4 is concentrating on the petrochemical and nuclear utility industries.

Where is this technology going?

Ten years ago, I was involved in a proprietary study that showed that a medium complexity E-size drawing would need to be converted at a total cost of no more than $100 for the Raster to CAD market to develop to any substantial size. At that time, such a drawing was costing $200-300. Since that time, the cost has fallen below the $100 level through the use of offshore facilities. Dimension 4 also claims to have achieved competitive costs on the kinds of drawings they can handle. Any developing market will depend on lower conversion costs to achieve any significant size, regardless of how the CAD database is achieved. In the meantime, more products and facilities are being designed directly on the CAD system, requiring less manual drawings to be converted. In short, I am still a market "bear" waiting for someone to prove me wrong.

Richard N. Stover is Editor and Publisher of Document Management Magazine, Senior Consultant for ERM Advisors, and serves as AIIM Staff Liaison for the EMMC. He can be reached at docmanage@aol.com.