Automating roof orientation detection from satellite imagery?

Question

A while ago I did a neighbourhood roof orientation survey for solar panel deployment, by visually checking each roof and marking individually whether they were south or almost south-facing.

As this is very time-consuming, I would like to find a way to automate this, but don't know how this would be possible by merely analysing roof shadows.

I can script a little but have never written any plug-ins.

What other elements do I need to include?

We are on a very low budget, so seek any suggestions how to do this using open source tools.

@globalavocado, what's the size of your study area? Also, do you have an accompanying DEM of the area? — elrobis, Dec 05 '11 at 14:20
What kind of data do you have? Do you need to identify roofs automatically or has that been done? Can you obtain photos at two different times of day (near midday)? — whuber, Dec 05 '11 at 15:25
@elrobis the neighbourhood area is about 500m by 1.25km. I don't have a DEM yet, but I need to do one at some point. — globalavocado, Dec 05 '11 at 16:54
@whuber Yes, I still need to identify half of them and for that I would like to automate. So far I've digitised the relevant rooves as polygons using aerial imagery from Bing. If I can't find any better quality images, I would still be interested to hear about the method you are suggesting, looking at different times of the day. — globalavocado, Dec 05 '11 at 17:07
for a similar need (getting main roof orientation) I used opencarto java lib. But I work from vector data (not satelitte imagery). — simo, Dec 05 '11 at 19:05

score 5 · Answer 1 · answered Dec 05 '11 at 18:21

5

Is is possible that LIDAR has been flown recently in the area? You can extract buildings this way... LIDAR would most likely be too expensive to fly yourself, probably at least 8-12k for an area that size.

http://knol.google.com/k/aerial-extraction-of-roof-surfaces-for-solar-analysis#

Found that article, may be of some help.

answered Dec 05 '11 at 18:21

Thad

935
1
12
26

+1 nice article. @globalavocado, are you sure the size of your study area merits writing a program to perform the analysis? Given the size of the study area, I suspect you can get your answer faster by combing over the aerial manually. For a small area, it may take longer to write the program. Of course, if you do this frequently and with larger areas, writing a program becomes much more justified. – elrobis Dec 05 '11 at 18:58
@Thad thanks for the link, hope I'll be able to follow all that! – globalavocado Dec 05 '11 at 22:55
@elrobis it is a small area but I'll have to do more in the future. I would also like to expand my knowledge of how to approach this challenge. – globalavocado Dec 05 '11 at 22:58
1

I'm not sure, but it looks like youll have a hard time extracting the angle of the roof from 2-d imagery and that you need some sort of 3-d data. Post back if you get anything to work, I would be interested to know what you work out. – Thad Dec 06 '11 at 13:09
Just found this as well. http://www.cybercity3d.com/ – Thad Dec 06 '11 at 14:15
@Thad it's useful to know that I would actually need to use 3D data in order to automate this, something I can bring up with respect to funding. – globalavocado Dec 07 '11 at 11:27

Automating roof orientation detection from satellite imagery?

1 Answers1

Linked