NYPL Labs Building Inspector: Extracting Data from Historic Maps

Slides for the talk given by Mauricio Giraldo Arteaga at OpenVisConf 2014 http://openvisconf.com/

Video here: https://www.youtube.com/watch?v=Oph1o3IZEFU

Errata (Apr 30, 2014): There's some (kind of cool) dispute on the bronze map. More info: http://exhibitions.nypl.org/treasures/items/show/163 and http://news.nationalgeographic.com/news/2013/08/130821-ostrich-globe-map-discovery-science-nation/

Transcript

1. mauricio giraldo arteaga @mgiraldo NYPL Labs
3. hola
4. my name is mauricio
6. research and circulating library system spanning 3 boroughs in NYC
8. NYPL Labs
10. i’m going to talk about maps
11. The Great Map Data Extraction
12. an adventure in three acts
13. prologue
14. The Lionel Pincus and Princess Firyal Map Division
15. photo by christopher cannon
21. 500,000+ maps 20,000+ books & atlases
27. year
28. street names year
29. use type street names year
30. use type street names name year
31. material use type street names name year
32. material use type street names name class year
33. material use type street names address name class year
34. material use type street names address floors name class year
35. material use type street names address floors name class year skylights
36. material use type street names address floors name class year skylights backyards
37. material use type street names address floors name class geo location year skylights backyards
38. footprint material use type street names address floors name class geo location year skylights backyards
39. footprint material use type street names address floors name class geo location year skylights backyards
40. we got these for several decades since the 1800s
41. data trapped in a legacy format
42. gimme all dems datas!
43. f**k yeah historical data!
44. citysdk.waag.org/buildings
45. NYU Stern / Imaginaria3D
46. NYU Stern / Imaginaria3D
47. maps.google.com
48. maps.google.com
50. maps.nypl.org/warper
53. hand-crafted, artisanal, locally-sourced data
54. 500,000+ maps 20,000+ books & atlases
55. 500,000+ maps 20,000+ books & atlases
56. 500,000+ maps 20,000+ books & atlases* *imagine how many pages an atlas has
58. this will take us a few millenia* *actual number taken out from a hat
62. geo-rectification or: “make it match Open Street Map”
65. *this is a simulation. actual process is intensive. consult your mathematician before trying
66. vectorization or: “draw the building shapes”
68. results from maps.nypl.org/warper
70. ~120k footprints produced in three years
    by staff and volunteers
71. there has to be a better way
72. act i: will there be polygons?
73. requests to geo companies went unanswered
75. can we automate this?
77. WTF!? @mgiraldo
78. what is a building?
80. completely enclosed by black lines
81. completely enclosed by black lines dashed lines are not walls
82. completely enclosed by black lines dashed lines are not walls > 20m2 (~180ft2)
83. completely enclosed by black lines dashed lines are not walls > 20m2 (~180ft2) < 3,000m2 (~27,000ft2)
84. completely enclosed by black lines dashed lines are not walls > 20m2 (~180ft2) < 3,000m2 (~27,000ft2) not paper-colored
85. process
86. github.com/NYPL/map-vectorizer
90. completely enclosed by black lines dashed lines are not walls > 20m2 (~180ft2) < 3,000m2 (~27,000ft2) not paper-colored
91. completely enclosed by black lines dashed lines are not walls > 20m2 (~180ft2) < 3,000m2 (~27,000ft2) not paper-colored
92. provide the best (possible) input image
97. differences in resampling cubic nearest neighbor
98. differences in resampling cubic nearest neighbor
99. make the image a binary bitmap or: “black and white”
102. polygonize or: “convert contiguous pixels to a single line shape”
104. ! gdal_polygonize.py test.tif -f "ESRI Shapefile" test.shp test
105. ! gdal_polygonize.py test.tif -f "ESRI Shapefile" test.shp test
106. ! gdal_polygonize.py test.tif -f "ESRI Shapefile" test.shp test
107. simplify* *for those polygons that we care about
108. completely enclosed by black lines dashed lines are not walls > 20m2 (~180ft2) < 3,000m2 (~27,000ft2) not paper-colored ✔ ✔
112. alpha shape convex hull with a sample point set cran.r-project.org/web/packages/alphahull/ *code basically stolen wholesale from rpubs.com/geospacedman/alphasimple
113. ﹡ ﹡ ﹡ ﹡ ﹡﹡ ﹡ ﹡ ﹡ ﹡ ﹡ ﹡
114. ﹡ ﹡ ﹡ ﹡ ﹡﹡ ﹡ ﹡ ﹡ ﹡ ﹡ ﹡
115. ﹡ ﹡ ﹡ ﹡ ﹡﹡ ﹡ ﹡ ﹡ ﹡ ﹡ ﹡
116. we need a set of points
118. pts = spsample(polygon, n=1000, type="hexagonal")
119. pts = spsample(polygon, n=1000, type="regular")
120. pts = spsample(polygon, n=1000, type="random")
121. now we alpha shaping
122. x.as = ashape(pts@coords, alpha=2.0)
123. there are other point reduction algorithms like Ramer-Douglas-Peucker or Whyatt Curve Simplification
124. separate the buildings from the chaff
125. completely enclosed by black lines dashed lines are not walls > 20m2 (~180ft2) < 3,000m2 (~27,000ft2) not paper-colored ✔ ✔ ✔ ✔
128. [218, 211, 209]
129. [218, 211, 209] paper [199, 179, 173], [179, 155, 157], [206, 193, 189], [199, 195, 163], [207, 204, 179], [195, 189, 154], [209, 203, 181], [255, 225, 40], [194, 198, 192], [161, 175, 190], [137, 174, 163], [166, 176, 172], [149, 156, 141] [205, 200, 186] not paper
133. this is good enough for our use case
137. ✔ ✔ ✔ ✔ ✔ completely enclosed by black lines dashed lines are not walls > 20m2 (~180ft2) < 3,000m2 (~27,000ft2) not paper-colored
138. computer-vision for attribute recognition *bonus quest
142. 66,056 footprints produced in one day for an 1859 atlas of manhattan
143. caveats:
     ! adjacency not enforced
     false positives/negatives
     buildings may also overlap
144. act ii: the vectorizer needs to prove itself
149. multiple inspections for each item and let consensus surface on its own
150. footprint validation or: “tell us what the computer got right or wrong“
151. are people willing to spend time checking building footprints? insurance atlases are not exactly the coolest type of maps
153. buildinginspector.nypl.org
154. github.com/NYPL/building-inspector
159. about a month later…
164. 420k+ flags* 70k+ unique polygons ! consensus: ~84% YES, 7% FIX, 9% NO *a “flag” is a YES/NO/FIX by one person for a given polygon
165. seems people are willing after all… we — our contributors
166. seems people are willing after all… we — our contributors
167. act iii: the return of the inspector
168. footprint material use type street names address floors name class geo location year skylights backyards
169. divide and conquer
170. footprint material use type street names address floors name class geo location year skylights backyards
171. three new tasks for now… we really want it all!
172. footprint material use type street names address floors name class geo location year skylights backyards
173. check
174. check YES
175. check YES address color
176. check YES FIX address color
177. check YES FIX address color fix
178. check YES FIX address color fix
179. check YES FIX address color fix *footprints marked as “NO” go to building heaven
180. check YES FIX address color fix *footprints marked as “NO” go to building heaven
181. fix
182. fix
183. address
184. address
185. classify color
186. classify color
187. resulting data available via an API in 100% recyclable GeoJSON
188. resulting data available via an API in 100% recyclable GeoJSON
190. not the end
191. gracias bocoup! mauricio giraldo arteaga @mgiraldo NYPL Labs images from giphy, wikimedia commons & nypl digital collections