FUSION OF HYPERSPECTRAL IMAGES AND LIDAR-BASED DEMS 
FOR COASTAL MAPPING 
A. Elaksher 
Cairo University, 12 abader st. alzitoon cairo, EGYPT - ahmedelaksher@yahoo.com 
KEYWORDS: L1DAR, Hyperspectral, Rectification, Classification, Image processing. 
ABSTRACT: 
Coastal mapping is essential for a number of applications such as coastal resource management, coastal environmental protection, 
and coastal development and planning. Coastal mapping has been carried out using a wide range of techniques such as ground 
surveying and aerial mapping. Recently, satellite images, active sensor elevation models, and multispectral and hyperspectral images 
have also been used in coastal mapping. The integration of two or more of these datasets can provide more reliable coastal 
information. This paper presents an alternative technique for coastal mapping using an AVIRIS image and a LIDAR-based DEM. 
The DEM is used to generate building cues that are converted to building polygons. Building pixels are then removed from the 
AVIRIS image, and a supervised classification is performed to generate road and shoreline classes. A number of image processing 
techniques are used to victories road and shoreline pixels. The geometric accuracy and the completeness of the results are evaluated. 
The average positional accuracy for the building, road, and shoreline layers are 2.3, 5.7 and 7.2 meters, with 93.2%, 91.3%, and 
95.2% detection rates respectively. The results demonstrate the potential of using LIDAR-based DEMs to detect building cues and 
remove their corresponding pixels from the classification process. Thus, integrating laser and optical data can provide high quality 
coastal geospatial information. 
1. INTRODUCTION 
Coastal areas represent important and diverse parts of the Earth 
along shorelines. They consist of recreational beaches, 
residential regions, industrial areas, and harbors. High quality 
geospatial coastal information is essential for coastal resource 
management, coastal environmental protection, and coastal 
development and planning. Acquiring such information has 
been carried out using aerial photographs and ground surveying. 
These techniques have many advantages, such as flexible 
scheduling, easy-to-change configuration, and high quality 
mapping results. However, they are expensive and needs special 
logistics and processing procedures. 
Recently, satellite-imaging systems have improved their image 
resolutions and opened the era for high resolution mapping 
from space. Pitts et al. (2004) used a series of five Landsat TM 
images, spanning from 1984 to 1997, for coral reefs mapping. 
Chang et al. (1999) used SPOT images for shoreline mapping 
and change detection. Drzyzga and Arbogast (2002) used 
IKONOS images to monitor and map coastal landscape changes 
during spring and late summer seasons for three sites along 
Lake Michigan and one site along Lake Huron. Di et al. (2003) 
used the multispectral IKONOS images for initial shoreline 
generation. The initial shoreline was then refined using the 
panchromatic IKONOS images. 
Active remote sensing techniques, LIDAR and synthetic 
aperture radar (SAR), have been tested and evaluated in a wide 
variety of coastal applications. Cook (2003) used LIDAR for 
shoreline mapping and change detection in Florida coast. Tuell 
(1998) used SAR for shoreline mapping and change detection in 
a remote area along the Alaska shoreline as part of the national 
geodetic survey (NGS) effort to evaluate the potential of several 
mapping technologies. Brzank and Heipke (2006) studied the 
extraction of land and water areas from laser scanner datasets. 
Flight strips were processed separately using a fuzzy logic 
approach that classify the data into water and non-water classes. 
The integration of active remote sensing technologies with 
other datasets makes it possible to provide reliable and 
automatic solutions for coastal mapping and change detection. 
Gibeaut et al. [8] used LIDAR data and historical aerial 
photographs to study the Gulf of Mexico shoreline changes. Lee 
and Shan (2003) combined LIDAR data and IKONOS images 
for coastal mapping. They reported an average detection rate 
89.3% without using the LIDAR data and 93% using LIDAR 
data. Bartels et al. (2007) presented a rule-based approach for 
improving classification accuracy obtained in a supervised 
maximum likelihood classification process using 
simultaneously recorded co-registered bands such as high 
resolution LIDAR first, last echo and intensity data, aerial and 
near infra-red photos. The results show that merging these 
datasets improve the quality of the classification process. 
This research demonstrates a new framework for automatic 
coastal mapping using the airborne visible/infrared imaging 
spectrometer (AVIRIS) and the light detection and ranging 
(LIDAR) systems. While AVIRIS images measure the spectral 
reflectance of the ground, operating with a wavelength band 
from 0.380pm to 2.500pm of the electromagnetic spectrum, 
LIDAR data are geometric range measurements, operating with 
a wavelength of about 1000pm. Therefore, the combination of 
these two measurements provides accurate geometric and 
spectral information about the ground, which could be used to 
produce high quality topographic maps. 
The LIDAR-based digital elevation model (DEM) and the 
AVIRIS image are processed and registered using a number of 
ground control points (GCPs) obtained from a pair of aerial 
images. The DEM is then filtered and segmented to generate 
building cues. These cues are converted to building polygons 
and used to generate a polygon layer for buildings. Due to the 
low resolution of the AVIRIS data and the diversity of the 
building roof materials, it is difficult to classify buildings as one 
class. Therefore, the building polygons are superimposed on the 
AVIRIS image. All buildings pixels in the AVIRIS image are 
725 excluded from the classification. A supervised classification