11. QSM Models and Tree Assortment - VUKOZ-OEL/3d-forest-classic GitHub Wiki

The Quantitative Structure Models (QSM) menu offers methods to obtain information about branches and stem (volume, length, branching order) and methods for tree analysis (other quantitative parameters of trees). The QSM models may be employed if a tree position and height are known.

Tree Reconstruction

Tree reconstruction is the first step of the QSM analysis. The input parameters are the tree of interest, voxel size, and multiplicator that defines the close neighbor, i.e. maximal distance of used voxels. The reconstruction starts by dividing the whole tree into voxels. The points that are inside the voxel, as well as points and voxels that are close neighbor (voxel size * multiplicator), are recognized for each voxel. The stem reconstruction starts by neighbor search at the lowest voxel. All neighbors connected in one piece into the homogeneous part are considered as segments. If neighbors create two or more groups of non-connected voxels they are classified as possible branches. The segment is closed at that place and the analysis continues up to the tree length. Finally, the information about voxels, ordering, i.e. parent, and child succession of segments, is known for each segment. Once the whole tree is divided into the segments, the tree reconstruction itself follows. Firstly all segments without children are said to be the ending segments. Parents of each of the ending segments are followed to the lowest voxel and their lengths are computed. The stem is the longest sequence of segments. These segments are merged into a single segment and all its children are assigned. The reconstruction of branches of the first order follows with similar semantics until it reaches the stem segment. After all first-order branches are reconstructed, branches of second, third, etc. order follow until all segments are assigned to the tree.

Figure: A – unprocessed tree point cloud, B – results of tree reconstruction: stem (red) and branches (green to blue according to branching order), C – cylindrical model of a tree, D – wood assortment.

QSM model

After the tree reconstruction, the parametric model of the tree can be evaluated. First of all the tree of interest has to be specified. Since the cylinder estimation is done by randomized hough transformation (as for DBH) the number of iterations has to be set then. The size of the estimated cylinders defines the height of the cylinders used for reconstruction. Reconstruction can be further restricted to the branches up to given order, given length, or given diameter. The last option is the stem profile evaluation checkbox. This result can differ from those given by the algorithms offered in the “tree menu”. The final results of the QSM reconstruction appear as connected cylinders for each branch with an attribute table containing information about volume, length, and order if each fitted cylinder.

Tree assortment

The final step is to estimate possible wood qualities (as the base for timber value estimates) according to Czech industry standards for wood assortment. The only input is the trees of interest. The evaluated attributes of the tree are convergence, skewness, diameter, length, and number of connected branches. The algorithm classifies the whole tree represented by cylinders into quality classes, higher quality classes are prioritized.

Table: Example of tree criterium for the classification of wood quality.

Class	Convergence	Skewness	Diameter size	Length	Branches
1.	up to 1 cm	1,5 cm/m	45 cm	3 m	absent
2.	up to 1 cm	2 cm/m	28 cm	3 m	up to 3cm
3.		3 cm/m	20 cm	2,5 m	up to 4 cm
4.		6 cm/m	7 cm	2 m
5.		10 cm/m	7 cm	1 m
6.