Ask Your Question

Building detection in satellite imagery

asked 2018-07-13 06:38:46 -0500

Hello everyone, I am trying to implement a building detector for satellite images using opencv (python). I include an example photo below. What I need to do is to be able to detect all available buildings (or roofs) in a picture and, if possible, calculate their surface. I am a complete novice in OpenCv and have only completed minor projects, so this seems quite difficult to me. My first thought is to use a HaarCascade Classifier, but I encounter a bit of a trouble in understanding the training method. I would be grateful to anyone who gives an opinion on my choice, or an alternative. Also, any helpful tutorials on how to train the classifier or calculating surface would be really appreciated.

image description

edit retag flag offensive close merge delete


Did you get the solution to this? I wanted to do the exact same thing but can't get the idea regarding where to begin.

Diptark Bose gravatar imageDiptark Bose ( 2019-02-13 05:36:26 -0500 )edit

@Diptark Bose please do not post answers here, if you have a question or a comment, thank you.

berak gravatar imageberak ( 2019-02-13 05:45:15 -0500 )edit

1 answer

Sort by ยป oldest newest most voted

answered 2019-04-17 05:30:18 -0500

Hello, I managed this to a degree of success, looking at the literature they don't suggest using this due to how much roofs vary. My outputs do classify houses however they also think trees are houses and car parking spaces, lots of things appear to have rectangular features. I used this link to train my classifier I think a larger training samples are needed.

edit flag offensive delete link more

Question Tools

1 follower


Asked: 2018-07-13 06:38:46 -0500

Seen: 2,363 times

Last updated: Jul 13 '18