I have a task where I need to identify dwellings/homes on property lots. The data that I have is an aerial image of the town, cadastre/property boundaries, land tenure information and land-use.

I was thinking using the land-use data to identify the residential areas and use those areas to select the cadastre/property boundaries (which may or may not have a dwelling/home on it).

What I am looking for is way to minimise residential properties that are vacant lots. Checking each lot manually will take forever so I am looking for some advice that could assist e.g. software that could detect a buildings from imagery or links to similar problems and solutions.

I am currently using ArcGIS 10.2.

  • 1
    This sounds like Feature Extraction so I have added a tag for that to your question. Clicking on it will help filter questions to those that may be useful for you to review.
    – PolyGeo
    Jun 10, 2015 at 23:57
  • 2
    Good luck with that! You will need more data to accurately determine dwellings (LiDAR will work, infra red can also help) you can do a supervised classification to find areas that look like buildings but as roofing material differs in colour significantly you will need to do multiple iterations to find likely construction then trim down by area to remove 'noise' but really your data is insufficient for the task with any deal of accuracy. To incorporate LiDAR into this (classified or unclassified) be prepared to do a lot of coding. Jun 11, 2015 at 0:21
  • 2
    which area the building footprints might be already captured.
    – Mapperz
    Jun 11, 2015 at 0:47
  • 2
    While there are methods to automatically extract buildings from imagery, and you could assume that homes only exist on residential classed lots, and further assume that there is only one house per lot (apartment buildings? condos?) and that the largest must be the house, there's pretty much no way to tell from imagery what a building is without making a lot of assumptions. I did photo interpretation for a couple of years. Without ancillary sources, what you describe is pretty much impossible. Is footprints your goal, or just vacant/not? Because that's usually in the property records.
    – Chris W
    Jun 11, 2015 at 4:32
  • 1
    Significant portions of Australia have building (points and polygons) with reasonable accuracy. In local government areas like Gold Coast I can attest the buildings (generated from LiDAR/Orthophotograpy) are less than 2 years old and are 99.9% accurate... I would contact your LGA and see what spatial data they have, if it's old at least you can reduce the parcels you need to look at - buildings don't often disappear but new ones are frequently built. @ChrisW, you could join development applications against lots to determine which ones have been built on - if you can get that from your LGA. Jun 11, 2015 at 4:34

1 Answer 1


As commented by @RichardFairhurst:

Have a look at this post for a possible approach to creating building footprints - https://geonet.esri.com/thread/107721 It does require access to parcel data and/or building permit data to be really useful, but the results provide very good footprints. I work for a local government, so I have access to both kinds of data for my jurisdiction. We also have a website where this data can be downloaded, but not all jurisdictions do.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.