TopoSys GmbH
Obere Stegwiesen 26
88400 Biberach
[email protected]
Introduction
To process an elevation model from the data acquired by an airborne laser scanner (ALS or Lidar) one needs to know four basic parameters (fig. 1):
- Distance from the sensor to the ground/object
- Sensor position
- Sensor orientation (attitude)
- Beam deflection
Figure 1: General Principle Of Airborne Laser Scanning (ALS)
The sensor systems mainly used differ very much in the technical parameter but the usability and reliability of the sensors and the derived products are closely linked to the mean of beam deflection and the scan pattern.
Laser beam deflection
IMU and GPS provide position and attitude of the sensor, but not the orientation of the laser beam. The method which is used to move the beam across the swath has got a great impact on the quality of the sensor system and the quality of the products as well as on mission restrictions and sensor efficiency.
Currently three concepts are used:
1) Oscillating mirror
The Laser beam is deflected by an oscillating mirror (fig. 2). Basic advantages are a very flexible adjustment of the required viewing angle. Basic disadvantages are
- Additional errors due to the mechanics, accelerations and wear-out
- Inhomogeneous scan pattern on ground
- Strong reciprocal relation between viewing angle and scan frequency
- Regular calibration required
Figure 2: Oscillating Mirror
This makes this concept not suitable for high precision Laser scanning but quite useful for small scale mapping. This concept is used by e.g. Optech and Leica.
2) Rotating mirror
The Laser beam is deflected by an rotating polygon mirror (fig. 3). Basic advantages are a flexible adjustment of the required viewing angle. Basic disadvantages are
- Additional error due to misalignment of the surfaces of the mirror and wear-out
- Inhomogeneous scan pattern on ground at wide scan angles
- No practical use of the theoretically possible wide scan angle due to loss of measurements at the border of the scan
- Regular calibration required
Figure 3: Rotating Mirror
This makes this concept also not suitable for high precision Laser scanning but quite useful for small scale mapping or for terrestrial scanners. This concept is used by e.g. Terrapoint or Riegl.
3) Fiber Based Scanner
The Laser beam is deflected by a fixed linear array of fibers (fig. 4). Basic advantages are
- Laser pulse rate is not linked to the viewing angle and flight height
- Dense and regular scan pattern
- No calibration required after factory setting
- Forward and side looking mounting possible
Basic disadvantage is the fixed viewing angle. This concept is perfectly suitable for high precision Laser scanning and only used by TopoSys.
Figure 2: Fiber Based ALS System
EDM and airborne Lidar scanning philosophy
Airborne laser-scanning might be considered to be comparable with common distance measurements by an Electronic Distance Measurement device (EDM) but there are a some essential differences which need to be taken into account.
In a distance measurement by an EDM the requirement of a multiple measurement is normally made. The measuring duration or number of measurements is oriented to the "quality" of the reflector. The less the reflecting object (reflector, house wall, tree trunk, etc.) is known, the longer the measuring duration has to be.
Increasing the number of measurements will make the final result more precise and more reliable, as the EDM averages over the total number of individual measurements.
Airborne laser-scanning does not allow the repetition of an individual measurement, as the sensor position and the beam direction change continuously. At airborne Laser scanning, each distance measurement is unique. It cannot be verified or improved by multiple measurements. The reliability of a single measurement with a scanning laser is hence significantly less than that of a measurement with EDM.
For EDM long experience exists and guidelines are well known how to measure precisely:
- No one would accept singular measurements for a reliable distance
- Multiple Measurements depending on stability of reflecting surface, calculated distance is an weighted average of all measurements
- With reflector:
- Required measurement duration about 1 sec
- Required number of measurements are between 100 and 1000
- Without reflector:
- Required measurement duration more than 1 sec
- Required number of measurements are more than 1000
The reliability of an elevation model hence has to be attained by other means. As with EDM, an ensemble has to be found permitting dependable information. This can be accomplished by very dense measurements where it may be assumed that hardly any differences in height have occurred in the immediate vicinity. The selection of the ensemble has to be made dynamically and be adapted to the respective conditions.
Scan Pattern
Figure 5 shows the typical scan pattern of a mirror based ALS system. Characteristics are:
- Narrow beam
- High lateral accuracy of individual measurement
- Wide measurement spacing
- No direct neighbors, so no context analysis possible
- Sensitive against random erroneous measurements
Figure 5: Mirror Based Scan Patter
Clearly the scan pattern does not follow the expertise of EDM and we need to ask ourselves why anyone does rely on such a singular distance measurement?
Figure 6 shows the typical scan pattern of a fiber based ALS system. Obviously the scan pattern is fully in line with the requirements of EDM. Characteristics are:
- Wide beam
- Medium lateral accuracy of individual measurement
- Narrow measurement spacing
- Large overlap, so precise context analysis possible
- Insensitive against random erroneous measurements
Figure 6: Fiber Based Scan Patter
Obviously, depending of the scan pattern and other underlying measurement principles like beam divergence, an ALS is suitable to fulfill certain applications or not. Choosing the right ALS will prevent from being disappointed with the result.
Edge Detection and Object Extraction
How well or how accurately do the multitude of single measurements describe a terrain or objects? The single measurements are more or less randomly distributed over a strip. In a homogeneous terrain, the measuring distance or distribution of the measurements is of subordinate importance. If, however, objects such as trenches, embankments, roof ridges are to be recorded, then corresponding demands have to be made on distribution and measurement spacing.
Figure 7 shows a scene of buildings and trees. The Lidar raw data of the fiber based scanner "Falcon II" are shown in colored elevation. Due to the wide beam, the high scan rate and the huge amount of overlapping measurements, first and last echo information is detected from all pulses which hit the edges. This allows to precisely detect edges and objects.
Figure 7: Edge And Object Detection Capability Of Fiber Based Scanner Systems
There is also a similar problem for terrain structures. Here again a very old rule applies: If a signal is scanned with equidistant increments, the step width must be less than half the smallest form one wishes to recognize. In electrical engineering this requirement is known as Shannon's theorem and has been used for over 50 years. If one wants to recognize the 1.0 m wide retaining wall, the sampling distance must not be larger than 0.5 m. What has been explained here in simplified form, applies analogously in 3D. If the measurement spacing becomes too large, then detailed structures will be lost. The lack of capability of acquiring break lines or of object detection is not common to airborne laser scanning by itself but to certain types of laser scanner systems.
Integrated Line Scanner
True-ortho images offer important properties for the use in GIS applications. They do not contain perspective artifacts of the shoot as known from conventional ortho images. Also, they are true in scale. Laser scanner systems have been extended by optical means mostly by loosely coupled digital framing cameras. Falcon II of TopoSys embarks a tightly mounted four channel line scanner and provides a very cost effective acquisition of RGB and CIR true ortho images.
Having an integrated sensor system like the Falcon II, image and elevation data can be acquired in parallel. Figure 8 and 9 show examples of 25cm true-ortho images which have been acquired by the Falcon II.
Figure 8: 25cm RGB And CIR True-Ortho Image Of Mannheim
Figure 9: 25cm RGB And CIR True-Ortho Image Of Mannheim
Conclusion
Reckoning from the first paid commissions for the acquisition of elevation models with airborne laser-scanning, this method is just 8 years "old". Even in our present-day fast-moving times, this is still a very infant stage.
Within these years airborne laser-scanning has developed into a very efficient tool and still has very high potential for development. Laser scanning nowadays is used in a wide range of applications like
- 3D city models and city planning
- Costal monitoring and erosion monitoring
- Flood protection and hydraulic simulations
- Monitoring of deposits and mines
- Power line mapping
- Corridor mapping
- Forest inventory and management
- Environmental protection
- Archeology
A reliable basic elevation model has to have a high density of measuring points to provide a sound basis for all these diverse applications.
It has been shown that by evaluating scan pattern and other technical parameters only fiber based sensor systems fulfill the most common principles of electronic distance measurement. While Lidar scanner systems based on mirror technology seems to have reached their top, the fiber scanners seemed to be almost at the very beginning of their development and offer huge potential to serve future needs of the market.
