Introducing mapsnap: Automated Georeferencing for Historic Sanborn Insurance Maps
Over the past few months, a program named mapsnap has been developed to automatically georeference old insurance maps. This article explains the significance of these maps, how mapsnap functions, and the rationale behind undertaking this endeavor. Georeferenced maps can be explored on mapsnap.org. The author's passion for history, cities, and maps led them to the Sanborn Insurance Maps, which offer a unique perspective of urban spaces before major urban renewal projects of the mid-20th century.
To learn more about Sanborn maps, refer to the Library of Congress or watch a six-minute explainer video.
Sanborn maps, produced by the Sanborn Map Company from 1870 to 1960, provide a detailed block-by-block view of every city in the United States. Originally intended for insurance assessment, these maps have become valuable resources for understanding the history of urban landscapes. The New York Public Library (NYPL) and OldNYC have a photo of a Sanborn map block, which can be matched with the buildings in a historic photo to capture the atmosphere of that era.
The NYPL holds a collection of New York maps, and the Library of Congress (LoC) has the largest collection of Sanborn maps available online for free.
Georeferencing Sanborn maps involves aligning them with a modern map, which is a time-consuming process traditionally done manually. The OldInsuranceMaps.net (OIM) community specializes in georeferencing public-domain Sanborn maps. After georeferencing a few maps in upstate New York and Brooklyn, the author questioned whether a computer could perform this task. After further investigation, mapsnap was created to address this challenge.
Mapsnap utilizes Optical Character Recognition (OCR) to detect street labels on the Sanborn map. EasyOCR, a text recognition model from 2020, is employed for this purpose. This model is advantageous due to its ease of setup, high performance, and ability to detect text at any angle, crucial for maps with vertically or diagonally oriented streets. After detecting street labels, the program identifies candidate intersections to generate a fit, ultimately creating Ground Control Points (GCPs) to establish the georeference.
Rewritten by urgent.news; may contain errors — check the original.