ArcGIS Pro 3.7 has been out for quite a while now. When I was reviewing the new features, I noticed a very useful addition to the toolbox: Extract Scanned Lines and Extract Scanned Polygons. These tools can automatically extract line and polygon features such as contours, rivers, building outlines, and land use from scanned rasters. Recently I happened to have a task of extracting vector data for town and street boundaries, so I will use this as an example to introduce the workflow.
P.S.: All data used in this article are for demonstration purposes only. They do not represent any official production data and are shared solely to illustrate the approach.
Data Acquisition
There are not many official channels for town and street boundary data, and they tend to be scattered. I found one useful source: Standard Maps. In addition to the national standard map service, many provinces and cities also have their own standard map services—for example, Shanghai’s:
https://shanghai.tianditu.gov.cn/map/views/standardMap.html
Taking the administrative division diagram of Chongming District as an example, download this standard map.
Map Georeferencing
Note that this map is not georeferenced; it is only an image. We need to turn it into a map with geographic coordinates. For the georeferencing process, refer to: QGIS Map Georeferencing Tutorial (with AI Georeferencing).
I performed the georeferencing using QGIS because ArcGIS on my computer was busy processing data at the time. ArcGIS Pro would also work fine—just use whatever you are comfortable with.
Map Binarization
After loading the map into ArcGIS Pro, you cannot immediately use the Extract Scanned Lines and Extract Scanned Polygons tools. These tools require a binarized image as input, so you need to binarize the previously georeferenced image first.
Go to Imagery -- Raster Functions -- Binary Threshold, select the layer, and click “Create new layer”.
The resulting image may appear completely black—don’t worry, that is correct.
Extracting Scanned Lines
Why not use Extract Scanned Polygons directly? With my particular data, extracting polygons directly did not yield very good results; you can try it yourself—many polygons would fail to be identified. Open the toolbox and find the Extract Scanned Lines tool.
After setting the layer, set the foreground value to 1 and run the tool. The generated result still has some minor issues—for example, text labels are recognized as lines, so some manual cleaning is required.
Feature to Polygon
After deleting the unwanted data, you can directly convert the lines to polygons using the Feature to Polygon tool. The final result is shown below:
Summary
The whole process is quite simple, isn’t it? There is still room for optimization. For instance, since I only needed to extract boundaries, I could have used Photoshop or Illustrator to remove all the text from the image before georeferencing. That would have saved me from manually cleaning up the data later.
Finally, after so many version updates, this release finally brings more than just a version number bump. As AI capabilities improve in the future, it will be worth upgrading to the latest version when possible.