Generating 3D point clouds in forested areas

[stpu]

Hi

I am trying to extract a 3D point-cloud in a forested area from 60%-30% overlap UltraCam images. My purpose is to get an accurate representation of the forest canopy, which means a 3D point-cloud that accounts for the high height variation present in a forest.

I have been producing some dense point-clouds with Photoscan, however it seems that the canopy always comes out smoothed, while I need to be able to reconstruct a model that shows also the tree tops and identify clearly the edges of the forest and the forest gaps.

Does anyone know how these objectives could be achieved? are there some parameters that can optimize my results?

Thanks in advance for the help.

[James]

Any point that you need to be included in your reconstructed model needs to be visible in at least 3 different images i.e. for a single straight flight line over flat ground you need >=66% overlap in the direction of travel.

With undulating surfaces where low areas are easily occluded by high areas you would need more photos to ensure you get your '3 shots' of the lower levels which could be occluded as soon as the camera has moved only a small distance, otherwise you will only capture the top most bits, depending on the density of cover, hence the smoothed result.

There may be other reasons too such as lower light levels the deeper you look, or movement in the trees, I'm not a forest person!

[Bruno Andrieu]

Hi

What is the pixel size, compared to a leaf of a ramet diameter ? I would think you need pixels to be very small compared to these elements.

can you post some images ?

Bruno


PS I am trying 3D reconstruction of wheat canopies, with pixel size of about 0,2 mm