CONTEXT CAPTURE

RE: CONTEXT CAPTURE

Oto — Fri, 18 Oct 2019 09:20:49 GMT

It is best to generate seed points and then remove those seedpoints which are on buildings or vegetation and then try to extract ground. Or other option is to add seedpoints manually by snapping them to ground. Maybe some previous Survey can be used as seed points for ground.

RE: CONTEXT CAPTURE

Steve Bury — Fri, 18 Oct 2019 09:14:50 GMT

Oto,

Regarding the ground extraction tool in Editor, I have tried several parameters for 'terrain variance' and 'max triangle edge' which range from 2-7 and 0.5-1.0 respectively. I also tried to adjust 'the smallest side of the largest infrastructure' many times. However, the result is so far away from correct. It seems like it is struggling with distinguishing ground and buildings.

Have I ignored something really important, or at the current stage the Ground extraction tool is just as good as the ability of Editor handling very large mesh?

RE: CONTEXT CAPTURE

Oto — Fri, 18 Jan 2019 10:11:07 GMT

You have to use Contextcapture Editor and use ground extraction tool.

Seems in future it could be automatic detection of objects - https://www.youtube.com/watch?v=SMCRzvsxlZc&t=6s

Some examples from Mapillary :

https://blog.mapillary.com/update/2017/12/08/massive-memory-savings-for-training-modern-deep-learning-architectures.html

https://blog.mapillary.com/update/2016/09/27/semantic-segmentation-object-recognition.html