The Story of Map Junction: A Passion for Maps

Map Junction is the result of one man’s passion for maps and another’s skills for software and algorithms. For years, William Warner, the founder of Avid Technology, Inc. and Wildfire Communications, Inc. had a fascination with maps, especially maps of Boston. Boston had undergone huge changes over the last two hundred years, and Warner wanted to know exactly what changes, and where. He figured that if maps could be scanned at high resolution, and then accurately aligned to each other, one might be able to build an Internet-based tool that would let anyone view and overlay these historic maps.

In the Spring of 2000, Warner started working with Greg Cockroft of Agog, Inc. Warner and Cockroft had a 15-year history of working together. In 1988, Cockroft created the ground-breaking video compression for Avid. Later, he created the first high-volume mass-market speech recognizer that today serves as the core of Wildfire, an electronic secretary that works over the phone.

Warner’s company, Warner Research, LLC, commissioned the work with Agog. The partnership’s first result was Maps Over Time, a system for comparing historic maps.(www.mapsovertime.com). Warner kept repeating a goal for the system: To view and compare old, new and future maps through a single system.

This single-system goal is important because today, there is usually a separate system for each type of maps. One for computer generated Geographic Information System (GIS) maps, another for image-based maps, and typically no system for comparison of images, let alone the ability to combine all types together.


How Well Would Transforming Old Maps Work?

We started with a collection old maps. Some were scanned from Warner’s own collection, such as a 40-map 1890 Bromley Atlas of Boston. Others came from the high-resolution scans offered by the Library of Congress on their web site. Still others came from a variety of internet sites offering, for example, historic topographic maps of New England. Importantly, high-quality digital orthophotos from the US Geological Survey (USGS) were available to serve as an accurate base.

The most detailed old maps of Boston were two from the Library of Congress – 1775 Boston and Environs – a map covering the entire area of Boston, and 1775 Town of Boston – a map that showed every street and marked its name.

But when one tried to stretch the map to fit today’s Boston using a single transform, only three points would line up – the three that you picked. Otherwise, streets would be way off of the current location, and the alignment only showed that while map-makers in 1777 were pretty good, they were still far from accurate.

The solution was to find multiple points that corresponded to each other. The end of Hanover Street in today’s North End is just the same today. So is the end of Long Wharf. So is the location of the Old State House. With many point of alignment, Greg developed a system for dividing each map into triangles and transforming each triangle separately using an algorithm called piecewise-affine. This would let us stretch and shift the old map so it more accurately depicted the locations of streets in those days.

The result was remarkable. Old maps made in 1777 could be corrected and aligned without looking “warped”. These two maps seemed to have gentle errors that were easily corrected, and they now align remarkably well over today’s street pattern, much of which is still unchanged after 224 years. For maps as far back as 1826, only minor correction was needed, but it turned out most maps needed at least a little nudging to come into the kind of alignment that lets you really make comparisons.


Swallowing the Elephant; Or Sometimes Just Looking at Its Toes

Anyone who has used Adobe Photoshop knows that working with huge images is an extremely slow process. Typically, just opening the file, if it’s a big one, takes a long time, then it takes time to process the file, and still more time to save the file back out. With files in Maps Over Time easily reaching 500M bytes in their uncompressed form, the process of transforming a map could become unbearably slow, even on a fast computer.

To solve this, Agog created two technologies. First, a viewer that could surgically extract just one little part of an enormous compressed JPEG image file, without decompressing the file. This lets anyone look around at the map, efficiently getting just the part they want sent to their browser. JPEG was selected because it is supported in all browsers.

Some mapping solutions solve this by storing images in a proprietary format and then recompressing the images to JPEG. This puts a large load on the map server.

Second, Agog built a processing engine that could transform these huge image files, again by keeping them mostly on disk during processing. The result is a system that can take 25 M byte compressed map that decompresses to 125M bytes, transform it with 50 separate triangles and make it ready for viewing in ten minutes.


Tickeling the Elephant Through a Long String

Maps Over Time is completely Internet-based. Images are uploaded to the Maps Over Time Server. From there they are transformed using a small, efficient applet called Transapplet.

While the images may be huge, the data needed to modify them is actually quite small. All we need to do is view the base map and the map to be transformed. Alignment points are stored locally within Transapplet. After the first three points, transapplet can even calculate the transforms locally to find out where it thinks new points will show up on the base map.


The Boston Atlas

In 2000, Martin Von Wyss, Digital Cartographer for the City of Boston, approached Warner Research and Agog to see if Maps Over Time could be extended to meet their needs. Boston needed to overlay a range of GIS maps from a variety of sources, they wanted fast viewing of aerial photos, and they need fast printing right from the browser.

Agog transformed Maps Over Time into a full GIS and image display system, with capabilities for past, present and future maps. Agog built a sophisticated printing system that allows accurate, scaled maps to printed using any local Windows-compatible printer.

The Boston Atlas is now serving maps to users both with its host department, the Boston Redevelopment Authority, and for the general public.


Extending the System to 3D

Once the Boston Atlas was operational, Agog began exploring how 3D data could be combined with 2D information. Using 3D terrain data available as 30-meter Digital Elevation Models (DEM) from the USGS, Agog extended the server and the applet for 3D operation. Now users can view any data in 2D as well as in 3D. Building footprints can be extruded based on building height. 3D data can be seen through integrated links to local OpenGL hardware on PCs or Macs. Or 3D data can be output to any VRML viewer. Data can be output as VRML for input into many architectural CAD packages.


Support for Standards

MapJunction provides integrated support for many standards. The server is OpenGIS non-compliant. This means that existing OpenGIS applications and viewers can view and download information from a MapJunction Server.

MapJunction provides a range of outputs so users can incorporate its maps into other applications.