The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bib. Beijing 2008 
726 
technique is then applied to classify the AVIRIS image. The 
classified pixels for the road and water classes are used to 
generate road and shoreline vector layers, respectively, via a 
line extraction process. Results showed an average positional 
accuracy of about five meters and an average detection rate of 
93%. The results demonstrate that integrating laser and optical 
data can provide high quality coastal information. 
2. DATASET DESCRIPTION 
The dataset used was collected over the coastal area of Ocean 
City, Maryland. Ocean City is an urban area that consists of 
roads, high-rise buildings, and residential buildings. In addition, 
along the east coast are a number of sandy beaches while 
harbours and docks are found on the west coast. The dataset 
was collected by the International Society of Photogrammetry 
and Remote Sensing (ISPRS) Commission III, Working Group 
5, (Csatho and Schenk, 1998). The dataset includes large-scale 
aerial images, laser altimetry data, and hyperspectral images. 
The large-scale aerial photographs are used as the source for 
ground control points (GCPs) and as the ground truth data for 
evaluation. The laser altimetry and the hyperspectral data are 
used for the coastal mapping process. 
The aerial photographs consist of a pair of 1:5000 aerial images 
acquired using an RC20 camera operated independently by 
NGS. The images are provided in a digital form at a resolution 
of 12.5pm. Well-distributed GCPs, surveyed using the 
differential global positioning system (DGPS), are used to 
compute the interior and exterior orientation parameters of the 
images. A manual feature extraction process is then performed 
in order to provide the ground truth data. 
The LIDAR-based DEM is acquired using an airborne 
topographic mapper (ATM) laser system. The ATM is a conical 
scanning laser altimeter developed by NASA for precise 
measurement of surface elevation changes. Laser elevation data 
is acquired as a point cloud that is used to drive the required 
DEM in three main steps. First, the flight path is reconstructed 
using the DGPS and the inertial navigation system (INS) 
techniques mounted with the ATM system on the aircraft. 
Secondly, the 3D coordinates for each laser pulse intersection 
with the ground are computed using the laser travelling time 
and the reconstructed flight path. The LIDAR data is generated 
as a high-density point cloud with an average spacing of one 
laser point per one square meter. Finally, a post-processing step 
is used for filtering out the data outliers and generating the 
DEM and other products. The final data is provided as a one- 
meter DEM projected in the universal traverse mercator (UTM) 
projection with the WGS84 used as the reference ellipsoid. The 
vertical accuracy of the LIDAR based DEM is about 10 
centimetres, (Ackerman, 1999). Twenty-eight ground 
checkpoints, measured in the reference orthophoto, are used to 
evaluate the horizontal accuracy of the LIDAR based DEM. 
The average of the root mean square errors (RMS) for the 28 
checkpoints is 1.08 meter. 
The AVIRIS hyperspectral image is obtained using the AVIRIS 
scanner from the jet propulsion laboratory (JPL) that was 
installed on the NGS aircraft. The AVIRIS instrument contains 
224 different detectors, each with a spectral bandwidth of 
approximately 10 nanometers (nm), allowing it to cover the 
entire range between 380nm and 2500nm. The ground 
resolution of the AVIRIS image is 3.8 meters. The image is 
provided in raw and column pixel units with approximate 
geographic coordinates (latitude and longitude) for the start and 
the end points of the flight line. Therefore, the AVIRIS image 
was first registered before it is used the coastal mapping 
application. 
3. REGISTERING OF AVIRIS IMAGE 
Several experiments are conducted to rectify the AVIRIS image 
using the 2D projective transformation model, Equation 1. 
Three experiments are performed using different sets of GCPs 
and an independent set of 25 ground checkpoints. For each 
experiment the RMS is used to evaluate the results. The true 
ground coordinates of the GCPs and checkpoints are measured 
from the stereo images using tradition photogrammetric 
techniques. After computing the 2D transformation parameters 
using the ground and image coordinates of the GPCs, the 
computed parameters are used to calculate the ground 
coordinates of the checkpoints. These coordinates are compared 
with the measured ground coordinates, from the stereo images, 
and their RMS is computed. Results show that the average RMS 
is about five meters. Figure 1 shows the AVIRIS imagery 
before and after the rectification process using 10 GCPs. 
p .... a E + a 2 j + a 3 
cji + c 2 j + l 
N ... bfi + b 2 j + b 3 
Cji + C 2 j + 1 
where E and N = true East and North coordinates, 
i and j = AVIRIS imagery pixel coordinates, 
a , a , a , b , b , b ,c,c = projective transformation 
1 2 3 1 2 3 1 2 
model parameters. 
Figure 1. AVIRIS image before and after rectification using 10 
GCPs 
4. COASTAL MAPPING 
The coastal mapping process is divided to two parts. In the first 
part, the LIDAR DEM is used to generate a vector layer for 
buildings. This task includes the following steps: DEM 
filtering, DEM segmentation, region classification, and region