The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
317
As DTM extraction is the most popular and main fundamental
process in most of the application, so it is always important to
seek for new accurate and precise methods for DTM production
from Lidar data. Still there are some errors in almost all of the
DTM calculation approaches which need to be investigated that
affect the calculated height the objects above the terrain surface
in the normalized DSM. Appropriate interpolation method
should be chosen for surface reconstruction: Gridding or
Tinning. The surface reconstructed by grid method, passes from
all the points and resulting surface is smooth. TIN interpolation
method is unreliable in sparse and missing data cases. Most of
the current methodologies outlook is applying an isolate
treatment of the Lidar data but there are some tendencies for the
utilization of the structure based methodologies sensitive to the
shape and neighborhood orientation of the locally adjacent
points around the point of interest like 3D fuzzy-morphology
algorithms. In forest application only single coniferous trees
have been extracted that have more regular and simple shape
for modeling. Better regression models with dynamic
parameters considering topographic relation of the trees should
be proposed for tree canopy extraction of different type. As
objects are vague and represented as cloud point in Lidar range
data, fuzzy and neural network approaches are well suited for
this kind of data. For example, for roads, a fuzzy reasoning
system can be established in a way that if the points’ heights are
near to the DTM and have gentle slope and same intensity in
the optical image then pixels are attributed as road. As another
example, a neural based system could be trained to identify
different trees shape. Newly there are some efforts on finding a
direct relation between acquired optical image and the
corresponding Lidar data to avoid resampling which affects the
shape of the trees and objects. Future works would concentrate
on evolving available statistical information and multi-source
information into the decision-supported classification
algorithms. The effective use of Lidar intensity data has not
been established yet and as mentioned before it is limited in
classification or matching process of Lidar data and optical
images. As the Lidar intensity data is more representing the
surface characteristics, it can be used along with optical data for
better object classification. In more technical sentence, the
result of the classification of Lidar data and optical data is not
the same and a well-defined reasoning system is required to do
an intelligent overlapping process for object extraction.
Absolute orientation of low cost photogrammetric data using
the automatically extracted 3D linear features from the Lidar
intensity data could be another interesting topic which opens a
new horizon on robust true ortho-photo generation in flood
mapping or determining coastline border accurately. In 3D
man-made object’s surface reconstruction especially in building
extraction, which are built close enough to the water bodies,
still there are unsolved problems. In bathymetric Lidar, water
surface and bottom reconstruction are the most important
concerns and a lot of efforts have been done to achieve higher
accuracy and better spatial resolution. Still there are some
uncertainties related to penetration depth of the laser beam,
GPS/INS systematic errors, atmospheric effect, ground control
point errors, foot print size, mission planning, processing of
data and etc. and those need more investigations. Still coastal
shoreline monitoring is not completely solved. Water body
dynamic movements are simulated according to the acquired
Lidar data for the prediction purpose of the water flows.
Reliable Lidar data and better prediction models are required
for improved results. Finally some important recommendation
and future works are expressed by the authors as follows:
1. For better waterfront detection, it is better to use ground
data and non-ground data together, as sometimes there are
misclassified.
2. Proposing a method for better differentiating of ground
returns from building or vegetation and from atmospheric
aerosol for better overall accuracy.
3. Providing better visualization tool for better detection of
outlier of Lidar data and the ability of using video images
simultaneously.
4. Calibrating of Lidar data during the flight by some ground
calibration targets due to varying conditions of the
environment and internal calibration drift of the laser
scanner components.
5. Providing advanced feature extraction methods for
discrimination of abrupt elevation changes like: water
front, sea wall, cliffs, and coastal dunes.
6. Integration and fusion of Lidar sensor with other type of
images especially CASI images.
7. Designing an improved automatic methodology for GIS
data-base updating using extracted features from the Lidar
data to facilitate the time-consuming task of manual
editing.
8. Investigation on the accurate water surface wave
extraction especially in shallow areas to achieve better
estimation of the depth and laser beam traveling path in
the water body. One should consider the tidal effects on
surface waves. Also distance between laser shots must be
less than half of the surface wave height.
9. Using the slope information of the adjacent topography of
the river to use it as estimation parameters for river
bathymetry and water height in very shallow area.
10. Comparing depth derived from the Lidar data with other
accurate data like echo sounder data to study the possible
systematic errors or fusion of the data.
11. Studying the errors caused by the projection
transformations applied on the Lidar data.
3. CONCLUSIONS
Airborne lidar systems are rapidly developing and expanding in
new applications. In this report, we focused on different
applications of Lidar sensor in hydrology and oceanography
especially for the river and watershed management. A good
comparison has been demonstrated among all the available
sensors and Lidar sensor seems to be an efficient sensor for this
purpose. This shows that Lidar sensor provides efficient, rapid,
and low cost tool for hydrological application, especially for
coastal and river water management. But still there are some
weaknesses on the Lidar data segmentation, visualization, very
shallow depth measurement, water wave estimation and etc.
Integration of lidar with imaging sensors and better processing
algorithms would make a great development in obtaining more
realistic and accurate 3D models of the geospatial objects.
Please note that the above discussed issues are more the authors
personal point of view.
REFERENCE
Ackermann, F., 1999. Airborne laser scanning: present status
and future expectations. ISPRS Journal of Phogrammetry &
Remote Sensing, Vol. 54, pages: 64-67.