Imagine putting on a virtual reality headset and ‘walking’ through the long-gone neighborhoods of a city. You can see streets and buildings as they were decades ago.
That possibility is very real now that researchers have developed a method to create a 3D digital model of the historic district using machine learning and historical maps from Sanborn Fire Insurance.
But digital models are more than just a novelty, they will give researchers the resources to conduct research that was previously nearly impossible, such as estimating the economic losses caused by the demolition of historic neighborhoods. .
“The story here is that we can now unlock the wealth of data embedded in the Sanborn fire atlas,” said study co-author Harvey Miller, professor of geography at Ohio State University. said.
“This enables a whole new approach to urban history research that was unimaginable before machine learning. It’s a game changer.”
The study was published today (June 28, 2023) in the journal. pro swan.
The study begins with the Sanborn Map, created in 1919 to help fire insurance companies assess liability in approximately 12,000 US cities and towns.th and 20th for centuries. Miller, director of the Ohio Center for Urban and Regional Analysis (CURA), said updates are often regular in large cities.
The problem for researchers, at least until maps were digitized, was that trying to manually collect usable data from these maps was tedious and time consuming. A digital version is now available from the Library of Congress.
Study co-author Yue Lin, an Ohio State University geography Ph.D. developed. residence, business, etc.
“The data we get from the Sanborn map gives us a very good idea of what the buildings are like,” Lin said.
Researchers tested machine learning techniques in two adjacent neighborhoods on the Near East side of Columbus, Ohio. These neighborhoods were largely destroyed in the 1960s due to the construction of Interstate 70.
One of the districts, Hanford Village, was developed in 1946 to house black veterans returning from World War II.
“The GI bill gave veterans money to buy homes, but the money could only be used for new construction,” said study co-author Gelika Logan, CURA’s assistance coordinator. “I mean, most of the houses were lost on the highway shortly after they were built.”
Another area studied was the Driving Park, which also had a thriving black community until Interstate 70 split it in two.
The researchers used 13 Sanborn maps of the two districts made in 1961, just before I-70 was built. Machine learning techniques have allowed us to extract data from maps and create digital models.
Comparing data to date from the Sanford map, a total of 380 buildings including 286 homes, 86 garages, 5 apartments and 3 shops were demolished in two districts for highway construction. was shown.
Analysis of the results showed that the machine learning model was highly accurate in recreating the information contained in the map. It is about 90% accurate for building footprints and construction materials.
“The accuracy of it was impressive. It really gave me a visual sense of what these neighborhoods were like that I wouldn’t have been able to do otherwise,” Miller said.
“With this project, we want to get to the point where we can give people virtual reality headsets and let them walk around the city as they did in 1960, 1940, or maybe 1881.”
Using the machine-learning techniques developed for this study, Miller said researchers could develop similar 3D models for nearly all of the 12,000 cities and towns that have Sanborn maps.
This will enable researchers to recreate areas lost not only to natural disasters such as floods, but also to urban regeneration, population decline, and other types of change.
Because Sanborn’s maps contain information about the companies that occupied specific buildings, researchers could recreate digital districts to reduce the economic impact of buildings lost to urban renewal or other factors. impact can be determined. Another possibility would be to study how replacing houses with solar-absorbing highways would affect urban heat island effects.
“There are many different types of research possible, and this will be a very useful resource for urban historians and other researchers of all kinds,” Miller said.
“Being able to create these 3D digital models and reconstruct buildings adds so much more than what charts, graphs, tables and traditional maps can display. There is so much possibility.”
