Radek Fiala, Karel Jedlička, Lucie Potřebová: Comparison of Geodatabase Terrain Pyramiding Methods for Airborne Laser Scanning Data (poster), 9th International Symposium GIS Ostrava, VŠB – Technical Univerzity of Ostrava, from 23rd to 25th January 2012
Designing IA for AI - Information Architecture Conference 2024
Comparison of Geodatabase Terrain Pyramiding Methods for Airborne Laser Scanning Data
1. COMPARISON OF GEODATABASE TERRAIN PYRAMIDING
METHODS FOR AIRBORNE LASER SCANNING DATA
Radek FIALA, Karel JEDLIČKA, Lucie POTŘEBOVÁ
Department of Mathematics, Faculty of Applied Sciences, University of West Bohemia,
Univerzitní 22, 306 14, Plzeň, Czech Republic www.gis.zcu.cz
fialar@kma.zcu.cz, smrcek@kma.zcu.cz, potreluc@students.zcu.cz
TEST DATA AND PROCESSING METHODOLOGY forest
Ÿ km test area (includes open terrain, bulit-up and forested area)
1×1 water
Unclassified airborne laser scanning (ALS) data (2,102,754 points) acqui-
Ÿ cultivated
land bushes
red within the Project of new hypsometry generation of the Czech Republic Detail (70 × 50 m)
Building of terrain pyramids using z-tolerance and window size methods
Ÿ
Evaluating errors of individual pyramid level compared with primary ALS
Ÿ
Ortophoto of the test area (1 × 1 km)
data using volume criterion
WINDOW SIZE PYRAMID LEVELS Z-TOLERANCE
Window size pyramiding method divides
Ÿ Ÿ pyramiding method based on z-tole-
The
the area of interest into regular planar rance controls the vertical accuracy of
square windows (tiles) of specified ex- each pyramid level relative to the full-
level 0
tent. Then, just one point (or two points -resolution data. The z-tolerance para-
in a case of min/max method) from meter expresses the maximum allowed
each window are picked up as points to
vertical difference between removed
higher pyramid level. The selection can
point and its footprint height in newly
be based on point minimum height,
created terrain (if the difference is big-
maximum height, mean height or mini-
ger than z-tolerance, the point has to
window size 2 × 2 m
mum and maximum height.
z-tolerance 0.5 m
remain in the generalized pyramid level).
Ÿ is no significant difference in value
There
level 1
of random errors comparing any of win- Increasing values of z-tolerance leads
Ÿ
dow size methods (mean, min, max, min/ to slowly decreasing number of points in
max). Dependence of systematic error on a pyramid level. Probably, setting the
selected window size method is obvious. parameter of z-tolerance high enough
Ÿ point count of a particular pyramid
The to generalize forested and built-up areas
level created by any of Window size met- would lead to a significant reduction of
window size 4 × 4 m
hods is very close to number of windows point count. In case of bare ground data
z-tolerance 1.0 m
(twice as much in case of min/max); the the results will be probably significantly
thinning method (mid level was used) different.
level 2
seems not to be worthwhile.
level 0 level 0
level 1 level 1
window size 8 × 8 m
level 2 level 2
z-tolerance 2.0 m
level 3 level 3
level 3
Detail of TINs formed from points in pyramid levels Detail of TINs formed from points in pyramid levels
1, 2 and 3 selected by window size method (mean) 1, 2 and 3 selected by z-tolerance method
COMPARISON RESULTS
Ÿ count of pyramid levels created by the z-tolerance and window size
Point Z-tolerance method provides significantly better results using comparable
Ÿ
methods are somewhat incomparable. When doubling the z-tolerance va- point count than any of window size methods.
lue, the point count decreases approximately by 30 % in resulting pyramid Ÿ comparing values of systematic and random error, pyramid levels
When
layers. For example, z-tolerance of 32 m leads to pyramid layer containing with comparable point count should be used. Because of this, comparison
39,686 points. The window size methods reduce point count by 75 % in of only the first levels can provide meaningful results. The z-tolerance
every next level, which correspond to z-tolerance ratio of about 15 between method provides significantly better results using comparable point count
levels. From this point of view, the z-tolerance method is not very suitable than any of window size methods. However, the low point count
for unclassified data. Similar results can be expected for a DSM data. In reduction of z-tolerance method for higher pyramid levels should
case of bare ground data the results could be probably significantly different. be taken into account.