Improved ability to model trees and other relatively disconnected, tall objects

[Radbert]

First off let me say that I am a huge fan of Photoscan.  I have also used other more expensive competing software packages to process UAV photo surveys that have more bells and whistles, but I find that Photoscan is highly robust and generates excellent output products, especially orthophotos with minimal distortions over vegetated surfaces.

I prefer to use Photoscan for most processing, but there is one application where it doesn't deliver quite as well in creating a desired product.  The application is creating detailed point cloud models of complex, often disconnected vegetated surfaces... mainly in the form of tall, thin vegetation structures like tree stems with small crowns.  I imagine that this limitation may not be due to something inherent in the PS algorithms, but perhaps in the manner that the densification step is executed or filtered (even by using very high photo overlap, mild filtering, or processing at different image scales, PS is often not able to capture these features quite as well).

Is it possible that certain processing parameters used in the PS densification step could be made accessible to better enable capturing these types of features?  For example, does PS require points in the dense cloud to be matched in a minimum of three reprojections, and if so, could there be an option to require only two?

Thanks for your consideration and I do acknowledge that vegetation modeling and mapping may be more of a specialized application for your software.