Two main products were developed: the resource evaluator
and the CT scan image analyzer Preprocessing tools were added on both tools
in order to generate appropriate input data.
1. CT scan image analyzer
1.1. Preprocessing: the image acquisition and conversion
tool.
A graphical conversion and compression tool was developed,
preserving the full dynamic and resolution in the original CT scan images.
The storage volume dropped down from 240 Mb to nearly 60-70 Mb per Ct exam
without losing any significant data. Algorithms were than integrated in
a user friendly graphical interface with utility tools (image cropping,
administrative annotation...).
1.2. The Ct scan image analyzer and visualization tool.
Tissue detection: the results leads to labeled images.
- Single threshold and morpho-mathematical process according
to CT number value were written in order to detect bark, hardwood and sapwood
tissues. The detection is done by a script that can.be automatically applied
on the full CT scan exam.
- The pith localisation feature was developed using both
manually user placement and constraint interpolation or propagation from
distant slices.
- Branch detection is performed by a complex image processing
2D and 3D filtering process followed by a 3D region growing algorithm.
Results are quite good in terms of detection (all the branches, even the
very thin ones are detected) and computational time (less than 1 sec per
slice) .
3D visualisation: original CT images and labeled images are
used.
Functions of C2000 software were used to obtain 3d reconstruction
and visualizations (including slicing).
Geometrical output generation is performed from the labeled
images.
It mainly consist in a sector generator and various
control fields. Thus rays can be generated as well as all geometrical parameters
needed in the resource evaluation software.
2. Resource evaluator.
Due to many technical aspects, developments were not built
on the initial EPIFN software but completely re-written in C and C++ languages.
2.1. Preprocessing: the internal geometrical data generation
tool.
Single tree data organization is based on IFN inventory description.
It consists mainly in the tree identification, its age, its height, its
dbh at 1.3 m, its heights of the first dead branch, first green branch
and thirst green whorl.
An interactive editor was built in order to load, check,
correct, add and save tree data.
Geometrical internal data are generated from models defined
by ERQB INRA team and inventory tree descriptions.
The user is able to select different models, and have access
to the various parameters of each of them. Generated data can be visualized
with an Excel(TM) compatible tool and read in the resource evaluator.
2.2. The resource evaluation tool.
Principles.
Inputs are height versus age, diameter versus age and ring
number and branch distribution data. The visualisation tool is based on
an two window display system. In each of them, data can be compute and
visualised on an hierarchical way from the stand to the tree section level,
it allows so different description levels of the same tree or comparison
of the same property on two stands or on two single trees. Specific libraries
were developed in order to built a graphical components library. curves,
histograms, text, legends, titles, an interactive 2D and 3D mouse control
box. Contents of any graphical window can be converted in an postscript
file for printing.
Data visualisation.
On the stand level, heights and diameters, log grading distributions
as well as height and diameter curves versus age are implemented. On the
tree level, tree profile curve, cross section ring display with ring width
distribution can be displayed . An internal properties map visualisation
tool including ring width and infra density maps was also developed. An
interactive wire-frame 3D shape viewer of the stem with branches display
was designed. Single growth unit view allows closer look on whorls and
non circular cross section shapes.
Sawing simulation.
Log, board and veneer sawing pattern definition tools were
developed. The sawing simulations are fully 3D (boards are considered as
full solid volumes) and allows non circular shapes. Result of sawing simulations
(boards, veneer) can be displayed in 3D on the same way than stem geometry
and internal properties.