REAL-TIME POINT CLOUD GENERATION IN ALS WITH SUB-DECIMETER ACCURACY: FIRST EXPERIENCE
J. Skalouda*, Y. Stebler a, P. Schaer a, P. Tomé b, R. Stengelec
Swiss Federal Institute of Technology (EPFL), a TOPO Lab, b ESPLAB, Station 18, 1015 Lausanne, Switzerland
c BSF Swissphoto AG, Dorfstrasse 53, 8105 Regensdorf-Watt, Switzerland
KEY WORDS: LiDAR, Real-time, Georeferencing, GPS/INS, RTK, Integration, Point Cloud
ABSTRACT:
The ALS system developed at EPFL has the real-time ability of integrating GPS/INS/LiDAR data for generating and analyzing laser point cloud directly in flight. The absolute accuracy of the resulting point cloud depends mainly on the quality of the trajectory, which is related to the type of GPS solution (e.g. absolute positioning, DGPS, RTK). To reach sub-decimeter accuracy for the point cloud in the real-time, an RTK-GPS solution is required. This requires establishing a communication link for the transmission of GPS corrections (or measurements). This article analyzes results of usability of RTK-GPS/ALS acquired in the frame of two flights, in the particular context of helicopter based missions. In the first flight, the transmission of the corrections has been carried out by a radio link, while in the other case, GPRS-based communication has been used. After describing the general system architecture the paper focuses on analyzing the critical aspects of achieving real-time ALS with sub-decimeter accuracy. The use of modern communication technologies and the exploitation of nation-wide reference GNSS networks are of primary interest. It is shown that the short outages in the GPS observations and the varying data throughput in the communication channel can be successfully surmounted by the process of GPS/INS integration. The results of these first experiments conclude that the integration of an RTK- GPS solution in the real-time GPS/INS integration and the subsequent point cloud generation provide a final point cloud with the quality that is sufficient for a wide range of applications. It also opens new opportunities for monitoring missions requiring short reaction time.
* Corresponding author. Tel: +41 21 693 27 53 / Fax: +41 21 693 57 40 / email: [email protected] 1. INTRODUCTION
1.1 Conventional Approach to ALS
Contrary to the terrestrial laser scanning (TLS), the conventional airborne laser scanning (ALS) generates the point cloud coordinates only in post-mission. There, the laser data is merged with the trajectory in a process that is called ”basic- processing”. For mapping tasks of higher accuracy, the trajectory estimate requires integrating inertial and satellite observations from rover and one or more base receivers. The base-receiver data is normally made available only after the flight and therefore the integrity of the carrier-phase differential positioning (CP-DGPS) can be obtained only then. If the quality of CP-DGPS is insufficient for periods longer than 10 to 30 seconds, there is a high probability that the quality of the GPS/INS integrated trajectory will also be insufficient during this interval. In (rare) cases, the resulting positioning error has constant influence during the flight-line. Then, its effect could be mitigated by the strip adjustment supposing there is a good overlap between the adjunct strips, e.g. (Filin and Vosselman, 2004; Pfeifer et al., 2005). In most cases, however, there is some fluctuation in the phase data observation, or in the satellite constellation, the reasons for which the above condition does not hold (Skaloud, 2006). The same is true in corridor- mapping, where the internal point-cloud accuracy cannot be judged from inter-strip discrepancies. In such cases, the remaining alternative for improving data quality is re-flying the mission, or at least part of it. This alternative is not only costly,
but maybe also not viable under some circumstances (e.g.
monitoring applications supporting decision making, short-data delivery, etc).
1.2 RTK Approach to ALS
As mentioned in (Skaloud, 2006) there is a number of possibilities for checking the quality of the satellite measurements in flight, most of them are coming from avionics.
Nevertheless, as the demand on trajectory accuracy in ALS applications is usually higher, the approaches pursued in avionics can only be regarded as complementary. The ultimate control for checking the eminence of the phase observations is performing CP-DGPS positioning in real-time. In surveying community, this approach is called Real-Time-Kinematics (RTK) and requires establishing and maintaining communication between the rover and the base receivers. This concept is certainly challenging for airborne rovers when relying on publicly available methods of mobile communication that are restrained in coverage (e.g. radio power or mobile- phone infrastructure) or continuity (i.e. dynamic allocation of service in radio-packed transmission as GPRS). This paper aims to demonstrate that, although demanding in realization, the RTK approach is feasible for helicopter-based ALS missions, at least in European-like milieu.
1.3 Paper Outline
To take the full benefit of the RTK positioning, the whole chain of basic-processing needs to be implemented on-line. Such implementations generally do not exist in commercial systems;
therefore, the focus will be on a specific system that offers such possibility (Schaer et al., 2008). After brief explanation of the system functionality with the on-line processing capability, the paper describes two types of communication links with respect to their advantages and shortcomings. Both approaches are then tested under real operations and the results are presented. The subsequent discussion focuses mainly on analyzing data accuracy (e.g. real-time point-cloud vs. post-processing) and solution availability (e.g. percentage of RTK-fixed solution).
The conclusions are drawn after that.
2. SYSTEM DESCRIPTION
Figure 1. Overview of the Scan2mapTM multi-modal architecture
The on-line georeferencing of airborne LiDAR data is practically nonexistent in commercial ALS. This section briefly presents a custom-made system which benefits, among others, of such functionality. The Scan2mapTM airborne mapping system has been built up upon its commercialized predecessor (Skaloud et al., 2006). It integrates multiple-frequency GNSS receivers with tactical-grade inertial measurement unit, mid- range airborne laser scanner and medium format digital camera.
The system is mainly adapted for helicopter based surveys, with the sensor head suspended on its side. The integration of all data streams is performed in a modular architecture that is linked via Ethernet communication. As shown in Figure 1, the data acquisition components are followed by two integrating modules: the GIINAV for real-time GPS/INS integration and the LIEOS for LiDAR exterior orientation. Finally, the supervisory and flight-management role is accomplished by the HELIPOS module. Within a flight-line, the module focuses on processing all measurements and generating laser point-cloud in real-time. Once the strip is terminated, its quality gets analyzed within the LIAN sub-module. There, the data could be also classified, and a coarse digital surface/terrain module produced (Schaer et al., 2008). The precision of the obtained laser point
cloud is directly related to the precision of the trajectory. The quality of the latter depends on the GPS positioning mode and is the best for the RTK with fixed-ambiguities. The possibility of achieving such type of positioning in flight is tested in the following sections.
3. COMMUNICATION TECHNOLOGY Communication links are required for the real-time transmission of GPS measurements or its corrections. The transmission possibilities of this information range from (geostationary) satellites to terrestrial wireless data transmission techniques.
The satellite based (e.g., SBAS) concept is limited to code- corrections, accuracy of which is not sufficient. The reasons related to bandwidth, interference, coverage or cost further limit the relatively wide possibilities to two choices: radio and cell- phone related technologies.
The transmission by radio is used in the traditional RTK applications (i.e. surveying). Its inconvenience for ALS is the limited range that is related to the (relatively low) transmission power. As the weight (essentially for power supply) is not critical here, the range can be increased using either ground repeaters or increased transmission power as long as the legal requirements are respected.
The second generation of the Global System for Mobile communications (GSM) is limited by its data rate of only 9.6 kbps. That corresponds approximately to 5 Hz of dual- frequency measurements from one reference station. The problems related to cell registration and hand-over are known to occur for fast moving carriers, such as aircraft. On the other hand, the General Packet Radio Service (GPRS), that is available on practically all GSM networks, does not suffer such setback and has four times larger bandwidth (i.e. four voice channels). The more recent UMTS technology can handle even higher data transfer rates; however, the transmission is usually handled by ‘bursts’ of packets and therefore has varying latency. This approach is therefore less suitable for RTK positioning than GPRS. Although the cell-network coverage decreases in rural regions, the coverage in European countries is good and constantly spreading.
4. TEST DESCRIPTION
Two flight tests were conducted on June 2008 and April 2009, respectively, in order to test the two RTK architectures previously mentioned. The first flight was flown in the Sion region of the Swiss Alps. There, a radio communication link was used between the base and the rover receivers. Two GPS receivers (rovers) were connected through a splitter to the same antenna on board the helicopter. One receiver (JAVAD Legacy) was configured in standalone mode enabling classical carrier- phase post-processing, while the other (TOPCON Hiper Pro) ran in the RTK mode. The radio modem was powered up by a serial port of the RTK-rover to which the corrections were transmitted. The radio antenna was mounted on the mapping sensor head pointing downwards. The second flight was also flown in the Swiss Alps, but in the region of Chur. This time, the satellite corrections were obtained from the swipos-NAV service of the Swiss Federal Office of Topography (Swisstopo).
This is a nation-wide service providing corrections from the Automated GNSS Network Switzerland (AGNES). The correction data was sent to the RTK rover receiver (JAVAD
Alpha-G2T) via internet using the NTRIP protocol (Grünig and Wild, 2005) and GPRS communication. The second receiver (JAVAD Legacy) connected to the same antenna was configured in standalone positioning mode. The virtual reference station (VRS) technique based on the AGNES network was employed. In such configuration, the processing center interpolates the data from several AGNES stations (figure 2) for the chosen VRS that is located in the proximity of the rover.
Figure 2. Overview of the virtual reference station operating mode
5. DISCUSSION
In this section, we will present and discuss some of the results obtained from the real-time approach to ALS. We will first compare the real-time trajectory solutions to those of post- processing which are considered as reference. The post- processed solution was obtained by a software which is frequently used in the ALS industry: The Waypoint’s GrafNav for the carrier-phase processing and Applanix’s Posproc for the GPS/INS integration.
5.1 Trajectory
The two upper plots in Figure 3 show an extract of the difference in trajectory estimation expressed in a local frame between the real-time solution (using RTK-GPS and GIINAV real-time GPS/INS integration) and the post-processed results for the second flight. As the differences in the East and North components are similar, only the latter is shown here. The upper horizontal red bars indicate the periods where the system was in a flight line (and thus the results are of higher importance). The obtained differences of the fixed RTK positions are mostly below 10 cm in the altimetry and planimetry, respectively. Nevertheless, the effects of float (red areas) and standalone (blue areas) solutions are demonstrated through sudden accuracy losses. These occur mainly during the transition phases of the flight where the GNSS signal reception is affected either by obstructions due to the environment (high mountains) or by the helicopter itself (large banking angle). It should be noticed that even in the post-processing step, the ambiguities could not be re-solved for a portion of the flight (the large red interval on the right-hand part of the plots). This
fact highlights the importance of monitoring the RTK ambiguity status within the flight as it identifies the potential problems in the carrier-phase post-processing step.
Figure 3. Difference in the computed position and attitude between the real-time solution using RTK-GPS and post-processed solution using CP-DGPS for the second flight (RTK by NTRIP/GPRS)
Similar results have been obtained in the first flight using radio communication link (Figure 4). There, the planimetry and altimetry variations are also below 10 cm. The existing bias in the vertical channel is very likely related to some imprecision of base station height. In this flight the proportion of the fixed solutions is significantly lower. This is caused by the reduced quality of the radio signal for some parts of the flight. The base line length is only one factor determining signal degradation.
The flight dynamics, and particularly its speed have strong influences on the communication link quality because of the radio antenna instability (Stebler, 2008).
Figure 4. Difference in the computed position and attitude between the real-time solution using RTK-GPS and post-processed solution using CP-DGPS for the first flight (RTK by radio)
Figure 5. Radio link quality expressed in percentage obtained during the first flight (RTK by radio)
Figure 5 shows that the bad radio signal occurs mainly during the transfer period of the flight and in the turns where the receiver antenna is no longer oriented vertically. On the contrary, the NTRIP/GPRS communication link was always available in the second flight. There, the losses of fixed ambiguities were mainly caused by GPS signal obstructions, the fact that was also noticeable during the carrier-phase post- processing step. The two lower plots of Figures 3 and 4 depict an extract of the differences in attitude. Since the roll and pitch differences are very similar, only the former ones are represented here. The differences are as expected for forward filtering with this type of the IMU (Litton LN-200). For both cases, the roll/pitch Root Mean Square (RMS) value is below 0.05°. The yaw RMS is naturally higher because of its dependence on the flight dynamics and alignment, but it remains below 0.10°.
The Tables 1 and 2 summarize the differences in the ALS point cloud estimates for the 3’000 and 13’000 epochs of data from the first and second flight, respectively. They show also the percentage of the observations obtained in each type of GPS solution as well as the mean differences and the standard deviation along the three axes. The XYZ column presents the RMS of the 3D vector magnitude. It can be seen that the
proportion of fixed and float solutions is higher in the NTRIP/GPRS configuration, highlighting the good availability of the GPRS network in the flown region. Although the data lengths are considerably different, the proportional distribution of the RTK-fixed solutions and the 3D discrepancies remain similar for both flights.
5.2 Laser point-cloud
The histograms in Figure 6 show the distribution of the laser point-cloud coordinate differences (RTK-ALS versus post- processed solution) obtained in the second flight. It can be shown by forward covariance propagation that the position errors influence directly the accuracy of the georeferenced point cloud (Glennie, 2006; Landtwing, 2005). According to that, the point coordinate errors along the three axes are similar to the GPS errors depicted in Figure 3: the planimetric and altimetric accuracies remain within a range of 10 cm (Table 3). For the first flight the distributions are very similar and therefore not shown here. It can be deduced that the GPS data outages of shorter duration or the isolated float and standalone positioning solutions are successfully surmounted through adequate stochastic modelling in the real-time GPS/INS integration.
Mean differences [m] Standard deviation [m] RMS [m]
Solution type Epochs [%] X Y Z X Y Z XYZ Standalone 15.7 0.01 0.03 -0.10 0.06 0.11 0.17 0.23
Float 0.0 - -
Fixed 84.3 0.01 0.02 All 100.0 0.01 0.02
- - - - -
-0.08 0.03 0.05 0.08 0.13 -0.09 0.04 0.06 0.10 0.15
rst flight (RTK by radio, based on 3’000 observations) Table 1. Summary of the ALS position estimates for the fi
Mean differences [m] Standard deviation [m] RMS [m]
Solution type Epochs [%] X Y Z X Y Z XYZ Standalone 5.2 0.08 0.07 0.17 0.53 0.22 0.16 0.62 Float 4.8 0.58 0.08 0.08 0.33 0.25 0.29 0.78
Fixed 90.1 -0.01 -0.01 0.06 0.08 0.05 0.06 0.13 All 100.0 0.03 0.00 0.07 0.20 0.09 0.10 0.25
Table 2. Summary of the ALS position estimates for the second flight (RTK by NTRIP/GPRS, based on 13’000 observations)
Figure 6. Histograms of difference in point-cloud coordinates computed in real-time and in post- processing for the second flight (RTK by NTRIP/GPRS, based on a sample of about 1’000’000 points)
6. CONCLUSION
In this paper we have first described the methodology for real- time generation of the laser point cloud from moving platforms.
Secondly, we have identified and tested two communication technologies suitable for airborne RTK positioning. Thirdly, we compared the real-time generated point cloud to that obtained by post-processing. From that comparison we concluded that whenever RTK-fixed solution is obtained, and that was 90% of the time, the respective differences in the laser point cloud coordinates are practically negligible (i.e. always smaller than 0.1 m). Although our testing was not exhaustive it served as a proof of concept for the new approach to ALS data collection, benefits of which are multiple:
• As the whole chain of the “basic data processing” gets automated, the method of airborne laser scanning becomes more economical and faster in production.
• The quality of all data sources can be controlled integrally and directly within the flight. This issue is particularly critical for the satellite phase observations.
• The technology is potentially viable for new type of monitoring missions requiring short reaction time.
REFERENCES
Filin, S. and Vosselman, G., 2004. Adjustment of Airborne Laser Altimetry Strips, ISPRS Congress. International Archives of Photogrammetry and Remote Sensing, Istanbul, Turkey, pp.
285-289.
Glennie, C.L., 2006. Rigourous 3D Error Analysis of Kinematic Scanning LIDAR Systems. Terrapoint USA Inc., The Woodlands.
Grünig, S. and Wild, U., 2005. Swipos über Internet. Neue Entwicklungen bei der Echtzeit-Positionierung. Geomatique Suisse, 3: 121-124.
Landtwing, S., 2005. Airborne Laser Scanning:
Genauigkeitsinformationen für den Auswertungsprozess, Swiss Federal Institute of Technology (ETH), Zürich, 83 pp.
Pfeifer, N., Oude, E. and Filin, S., 2005. Automatic tie elements detection for laser scanner strip adjustment. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 36(Part 3/W19).
Schaer, P., Skaloud, J. and Tome, P., 2008. Towards in-flight quality assessment of airborne laser scanning. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 37(B5): 851-856.
Skaloud, J., 2006. Reliability of Direct Georeferencing Phase 1:
An Overview of the Current Approaches and Possibilities, Checking and Improving of Digital Terrain Models / Reliability of Direct Georeferencing EuroSDR Official Publication 51.
Mean difference [m] Standard deviation [m] RMS [m]
X Y Z X Y Z XYZ
Flight 1 (radio) 0.04 0.01 -0.10 0.06 0.07 0.03 0.15 Flight 2 (NTRIP/GPRS) 0.05 0.04 0.02 0.11 0.08 0.08 0.17
Table 3. Summary of the point-cloud differences (real-time versus post-processed) for both flights
Skaloud, J., Vallet, J., Veyssiere, G. and Koelbl, O., 2006. An Eye for Landscape - Rapid Aerial Mapping with Handheld Sensors. GPS World(May): 7.
Stebler, Y., 2008. GPS/INS Integrity in Airborne Mapping.
Master Thesis, Swiss Federal Institute of Technology, Lausanne.
Wild, U., Grünig, S. and Hug, R., 2006. Teil 11: Automatisches GPS-netz Schweiz (AGNES): Stationsnetz und Positionierungsdienste. 978-3-302-10003-6.
