The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
351
Figure 1. Example of classification through ML classifier
When combining the information provided by the traditional
sensors with the one supplied by the LIDAR sensor, an
additional issue comes up: the origin of the HIS and infrared
attributes are digital images with regularly distributed pixels
generally forming rectangular grids that may have different
resolutions among them and they may range over ground areas
not necessarily coinciding among themselves; in addition, the
LIDAR scatter plot containing the information about the
intensity level and the Z increment are irregular grids as a
consequence of the way in which the points were obtained.
Hence, in order to be able to carry out the classification using
all the available information, the first step is to give attributes
(HIS and infrared) to every point making up the scatter plot
obtained with the LIDAR sensor; in order to do that, given the
coordinates of a LIDAR point, we will locate the nearest pixel
to it in every image and we will assign the attributes of the
found pixel as attributes of the LIDAR point.
Figure 2. Detail of the LIDAR cloud combined with RGB
Inversely, given a particular pixel in any of the available images
(HIS and infrared), its LIDAR attributes (intensity level and Z
increment) will be search for; the three LIDAR points nearest to
the particular pixel are located through its coordinates; the
attributes obtained from those points are interpolated using the
Delaunay triangulation (Priego el al, 2004), and finally the
result will be assigned to the pixel. With this procedure
synthetic images may be created through the LIDAR attributes
(among other applications).
Figure 3. Detail of a synthetic image
4. MULTI-CRITERION ANALYSIS
Presently there exist different techniques to obtain DEM’s. This
paper focuses on the study of two main methodologies, namely
photogrammetry and LIDAR. Through the integration of these
two basic technologies a third option comes forth sharing
certain characteristics with the other two techniques. This third
option stems from the integration of the points in the scatter plot
from a LIDAR system, from the break lines fully outlining the
irregularities of the terrain.
It has been empirically verified that DTM generation through
airborne sensors has both advantages and inconveniences. The
overabundance of points allows, through a series of algorithms,
automating the classification of the points on the terrain and on
no-terrain. However this overabundance does not entirely make
up for the efficacy or quality of the end product that can be
achieved by the involvement of an operator. In certain areas of
the terrain, the lack of break lines defining it correctly will take
a certain degree of accuracy away from the representation
model automatically obtained by LIDAR sensors.
In this section a comparative study is made of the three already
defined methodologies, which would allow choosing the most
efficient one, the focal point being the three essential aspects in
a typical process of cartographic production: time, cost and
quality of the product. For that purpose a multi-criterion
analysis is carried out, thereby assessing these three aspects in
detail and leading into a general decision triangle which would
enable anybody to choose the best option according to
individual needs.
4.1 Indicators and independent analysis
In order to evaluate the three methodologies (photogrammetry
(manually edited automatic correlation with break line
integration), LIDAR (automated process of classification and
filtering) and the integration of LIDAR with photogrammetric
break lines, it is necessary to define a series of general
indicators to be independently assessed in each case. These
indicators allow assignment of an objective numerical value to
different basic characteristics of a cartographic production
process in relation to the three main aspects that allow choosing
a technique.
The following table shows the indicators used in the analysis
carried out for each methodology:
Altimétrie accuracy
Planimetric accuracy
Resolution
Type of model (TIN, GRID, etc.)
Grid spacing
Realistic model
Full model
Production time
On-flight storage
Real cost of equipment
In-flight data processing
Preparation of human resources
In-flight losses
Condition for picture taking
(orientation)
Calibration
Flight cost/difficulty
Terrain characteristics
Information about homogeneous
surfaces
Information about break
lines
Weather/time
Automatic 3D data
acquisition
Measurement redundancy
Semantic information
RDI
Information source
Table 2. List of indicators used in the analysis