l. Small to moderately sized airfield
2. Proximity to partners
3. Obstructions
4. Interest from airport administration
The city — Santa Barbara situated in a coastal valley between
the Santa Ynez Mountains and the Pacific Ocean. The airport is
located on west of the citv. Santa Barbara airport has three
runways (see Figure 1). One is in east-west direction; another
two are in the direction from southeast to northwest. At the
north part, there are hills with varying heights. There are only a
few large buildings around the airport, and most residential
buildings are both smaller and lower than forested clusters. The
airfield features are surveyed in field, including runway
polygon and end points, taxiway, touchdown zone, overrun stop
way, apron hardstand, buildings, roads, inland water areas,
airfield elevation points, and many obstructions. The lidar
dataset covers only the areas within the IHS. It consists of 11.3
million first-return lidar points covering an area of 26 km. The
terrain relief is from 0.31 m to 148.51 m. An aerial color digital
orthorectified quadrangle (DOQ) with a 1-m GSD covers the
whole airfield.
Figure 1. Study area — Santa Barbara Airport
3. METHODOLOGY
3.1 Overview
The first step of the project is to identify airfield obstructions
by extracting physical features from Lidar and complying with
NIMA's Airfield specification. As airborne Lidar systems have
the capability on acquiring digital surface models (DSMs) with
high accuracy, and transforming the datasets (scattered 3D
points) into grid-base range, the digital terrain models (DTMs)
can be generated by using Lidar Expert, which is a toolkit for
automated information extraction from lidar data (Hu and Tao,
2004a). Then, the digital non-terrain model (DNM) is produced
by subtracting the DTM from the lidar DSM (Hu and Tao,
2004b), and represents all these non-terrain objects including
vegetation, buildings and other man-made object, upon a flat
reference plane.
Since last-return lidar data is not provided, it is impossible to
reliably distinguish buildings from forested areas solely using
shape measures. So the potential obstructions are digitized in
2-D manually, and then Lidar Expert is used to extract their
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
heights from raw lidar data. After removing all the digitized
man-made objects from the DNM, we get the vegetation height
model (VHM), which is the representation of the remaining
objects such as trees and bushes.
Next all OIS surfaces are created according to NIMA Airfield
Initiative Document. As mentioned before, objects that protrude
the OIS or the highest one within one surface regard as
obstructions. In order to find all potential obstructions, one of
GIS spatial tools, the index model (Lo and Yeung, 2002), is
used to generate an obstruction map with three indication levels.
The highest one means all obstructions in this level must be
removed immediately for the purpose of protecting the National
Aviation Infrastructure. The medium means trees, buildings, or
other physical objects maybe potential obstructions and need
observe periodically. The low means all objects in this level are
safe to the airfield. Inside the index model, four OIS-related
factors are used: distance to the center line of each runway and
its extent, location of objects related to the OIS, type of objects,
and penetration related to the OIS. The weights of factors are
determined by Saaty Methods. After reclassifving and
calculation, the final obstruction map is generated.
3.2 Workflow
In Figure 2, a cartographic model (Michael, 2000) illustrates
the workflow of the project. The first part is lidar data
processing. It uses filters to generate the DSM, applies image
processing and interpolation algorithms to generate the DTM
and the DNM. The second part is to digitize buildings and
residential areas manually on the aerial DOQ. The forested
areas are obtained by subtracting buildings and residential areas
from the DNM. The third part is to create seven OIS surfaces,
and to digitize bridges, rivers, and roads inside the SVTW. The
fourth part is to do geometric correction and bilinear
interpolation for scanned topographical map. The new map is
used as airfield background. The fifth part is to merge all the
features together. Objects extending the OIS should be recorded
as airfield obstructions. To help airport managers make a
priority decision for obstructions, four new attributes are added.
After doing spatial analyses, deriving weights, the finally
classified obstructions are identified by raster calculation and
are visualized in 3D.
3.3 Data Processing
3.3.1 Generation of the DTM and DNM: The direct products
of Lidar scanning are DSMs that are formed by the point clouds
returned from the top of the Earth’s surface partly covered by
manmade or natural ground objects. At beginning, the raw
Lidar point clouds are filtered to discard outliers or blunders,
which have too low or too high elevation values or very large
intensity values that do not match their surroundings. A median
filter is adopted in this step, because it is useful for removing
noise from the original image, especially shot noise by which
individual pixel are corrupted or missing. It selects the central
value from the lowest to the highest (Jensen, 1996).
The final DTM is derived by applying a hierarchical terrain
recovery algorithm (Hu and Tao, 2004). The algorithm
identifies terrain points by finding local minima and other
topographic points, and recovers the terrain surface in a
coarse-to-fine manner. First, after screening the blunders, the
scattered 3-D points are transformed into a grid-based range
image by sclecting the point of lowest elevation in each grid.
Then, an image pyramid is generated. The top-level image is
hypothesized to be a coarse DTM if its grid size is larger than
Internatio.
the
refir
eacl
conc
r1
[T rb eu TS
TY"TI 0I") r7 Pme] ] r^^ T
In F
the |
SA
—
sw
Figu
lines
the I
beca
shov
Whe
DNN
