City3D involves several parameters that work for most already tested buildings. Adjustments to the following two parameters might be necessary for the practical reconstruction from datasets with varying characteristics:
-
min_points
(for plane/roof extraction). This parameter defines the minimum number of points required to detect roof planes, which affects the granularity of plane extraction. A smaller value yields more planes capturing finer details and smaller structures but potentially with less confidence. However, decreasingmin_points
may lead to an excessive number of candidate faces (and thus an optimization problem that may not be solved within a reasonable time window). In this case, the user is advised to increasemin_points
to target results with simpler structures. -
pixel_size
(in height map generation). This parameter determines the resolution of the height map, which will affect the precision of line detection for footprint generation. The recommended values are typically in the range [0.15, 0.3]. A smaller value increases the resolution of the height map and thus results in more detailed reconstruction results. However, decreasingpixel_size
may lead to an excessive number of candidate faces (and thus an optimization problem that may not be solved within a reasonable time window). In this case, the user is advised to increasepixel_size
to target results with simpler structures.
For City3D, these parameters can be
adjusted interactively (see image below). You can check the Manual
box to access these parameters.
For CLI_Example_1 and CLI_Example_2, these parameters can be manually tuned within the source code, e.g.,
City3D/code/CLI_Example_1/main.cpp
Line 31 in 210325f
City3D/code/CLI_Example_1/main.cpp
Line 32 in 210325f