** NOTE: Display Scaling is affecting the output of this program if you have scaling set other than 100%. ** Reccomended not to use this script unless output can be verified for now.
This is a self-contained python script that can be used to compute frontal area projections of STL meshes for use in analysis, particularly when doing comparisons of frontal area of automotive shapes for CFD purposes.
External Dependencies:
- VTK
- Numpy
The method I am using leverages the VTK rendering pipeline to take a numerical approach to the area calculation problem, using ray-tracing on a fixed raster resolution window to compute an approximation of the projected orthogonal area.
This method is ~30x-60x faster than the analytical approach and is inspired by the method that is used to measure frontal area of physical vehicles for windtunnel testing, which uses collimated laser light sources and photo mapping to measure the shadow of a vehicle in a darkroom.
Because the heavy lifting is being done by VTK, a compiled package that is
meant to load and view very large data and visualizations, this script can
handle medium sized exterior vehicle STL's in <1 second and can handle even very
large scans, 1GB+ files in about 30 seconds on an M1 Macbook.
Binary STLs load fine and the VTK render doesn't care about unified
normals, non-manifolds, etc. It tends to just work.
The accuracy for typical sedan sized test vehicles seems to be roughly ~0.0015 m^2 or better depending on the resolution used and the fit factor of vehicle to render frame.
The program has several command line options (see -h for help) to help get best results. It generates a PNG of the computed shadowmask by default and shows it in a preview, but this can be disabled via -nosave and -noshow. Default is projection in the X direction as is the case for most frontal areas of automotive models and default windowing and resolution is roughly correct for most vehicles.
The scale and resolution are adaptable, but the defaults are mostly setup to work with car- sized STL models in millimeters. The program has some safety checks to determine if the renderer is configured to work with an appropriate model size, and if not, it exits with error info.
To test, run:
python pyfrontal.py -x tests/test_mesh.stl
This should result in a Projected Area: 2.0 m^2
An example use case:
The resulting preview window:
