New approach in using eRTK-GPS for direct georeferencing of aerial images in a GIS application

New approach in using eRTK-GPS for direct georeferencing of aerial images in a GIS application

Lyle, Stacey D

ABSTRACT: The purpose of this paper is to describe the research conducted in the development and preliminary testing of a real-time georeferencing method for a small-format digital aerial camera system using new methodologies in extended Real-Time Kinematic Global Positioning System (eRTK-GPS) solutions. This system is being developed to supply the U.S. Department of Agriculture Farm Service Agency (USDA FSA) with aerial digital images that would replace the current 35 mm slide film method and facilitate the development of a geographic information system (GIS) for crop compliance support. The direct georeferencing method was developed by the University of Georgia and Texas A&M University-Corpus Christi in fulfillment of the National Aeronautical Space Agency’s (NASA) contract for developing new technologies and methodologies in remote sensing. The research included development of eRTK-GPS in an aerial platform where differential solutions 35 kilometers from the base-station generate two-centimeter and five-centimeter accuracy horizontally and vertically from the digital camera exposure station’s location. A real-time georeferencing system results in a world file with affine transformation parameters without the need for post-processing GPS or ground control points. Flight tests were conducted in Georgia and Texas with different wireless communication systems to assess the robustness of the solutions on an aerial platform.

Introduction

The U.S. Department of Agriculture Farm Service Agency (USDA FSA) seeks to implement a geographic information system (GIS) to connect and manage crops for crop compliance monitoring under crop support programs. The USDA has determined that the 35 mm film slides traditionally used for aerial compliance are no longer compatible with the digital format needed to establish a geographic information system (GIS) for land ownership. Research suggests that a small-format digital camera system can be utilized for acquiring new compliance imagery (Davis 2002), as this new methodology creates the ability to maintain the image completely in a digital format.

The USDA expects to efficiently convert line data extracted from aerial photography into a digital format by using a GIS, which provides a framework for linking tabular data to their spatial dimensions. The line data are boundaries of common land units (CLU) of common owner, tenant, and producer associations, and the function of the CLU in the GIS is to define farmers and land relationships spatially (USDA FSA 1998). Seven digitizing centers have been established in several states to begin the process of developing common land units. The CLUs represent various combinations of land cover and land use. The boundaries and placement of these units in a GIS is based on these combinations. However, the land-cover and land-use demarcation lines must also coincide with the ownership lines recorded in the cadastre. Aerial images are used to visually delineate boundaries in USDA FSA cadastral layers.

The current method of utilizing 35 mm slide film requires a considerable amount of time to process and rectify, measure acreage, and digitize the land ownership. The wait time for film development is approximately two weeks. The slides are indexed by a slide projector oriented toward a flat-table digitizer (Figure 1). Images of target areas are often shot with smaller scales to capture the target site in one slide because the navigation of targeting areas is often performed from chart navigation methods. Digital navigation could provide an instant improvement with supported over-site verification of the target area within the aircraft at approach. Distance cross-course values as well as height correction from eRTK-GPS can improve target image capture. Having GIS overlay functionality in the navigation software will allow the CLU to instantly be integrated with the image to determine if the correct crop area was collected.

The objective of this research is to develop a system that utilizes the capture of digital data for direct georeferencing in real time based on the extended Real Time Kinematic Global Positioning System (eRTK-GPS) solutions with on-the-fly image georeferencing. According to Wolf and Dewitt (2000, p. 189) georeferencing is often referred to as rectification; however, “the term rectification is reserved for the process of removing effects of tilt from an aerial photograph.” Direct georeferencing has until recently been synonymous with direct rectification. In the context of this research, direct georeferencing will refer to the use of an orthogonal transformation account for shift, scale, and rotation.

The direct georeferencing solution is being developed with USDA and FSA and county aid offices and farmers to enable better determination of new crop lines for crop rotations, seed corn and plots, irrigation pivot management, specialty crop acreage, conservation work, and crop insurance (Bailey and Meador 2002). A GIS is a tool of precision farming and it assists in organizing the entire farming process. The spatial information database that is GIS is an organized collection of computers, hardware, software, geographic data, aerial imagery, maps, and personnel designed to efficiently capture, store, update, manipulate, analyze, and display all forms of geographically referenced information (Clark 1997).

Other methods have been deployed that utilize a combination of GPS and Inertial Navigation System (INS). Improvements to GPS/INS solutions have been forthcoming with new concepts of direct georeferencing, such as the concept of bridging INS and GPS by means of Kaiman filters (Skaloud 1999; Mostafa et al. 1998; Cramer and Haala 1999; Skaloud 2002). This solution allows the INS to aid in the ambiguity solution; therefore, extending the range from base station to rover without concern for lost carrier-phase ambiguity cycle count. Direct georeferencing uses GPS for position and INS for attitude determination at the time of image exposure. The position and attitude are correlated to the parameters of the exterior orientation.

The Photography Field office of the USDA FSA evaluated a number of digital camera systems for use by area offices (Davis 2002). Digital systems currently being reviewed still demand post processing GPS, but they do have the ability to shorten the time of obtaining the final product. This research proposes to extend the current concepts determined by the USDA and FSA to real-time data processing of the GPS solution by reducing the need for some post-processing steps.

To meet the objectives of this research, several key technologies and methodologies must be developed. Utilizing eRTK-GPS in an aerial platform requires an advanced communication system. The accuracy achieved with eRTK-GPS in an aerial platform must be evaluated. The exterior orientation parameters such as tilt and heading must be analyzed for this application. An algorithm was developed to directly georeference an image by writing a georeference world file with affine transformation parameters. This research will be discussed in future publications.

According to the USDA FSA, georeferencing (rubber-sheeting, warping) is sufficient for the accuracy level needed for the aerial compliance program (Davis 2002, p.9). Orthorectification would lead to higher accuracies than required for the crop compliance application. Instead, a georeferencing method was developed where only the scale, position, and yaw are taken into consideration.

The camera was mounted on a leveling device to reduce the tilt. Tilt from roll and tip was found to cause only a slight increase in the positional accuracy. At 2000-meter flying height with 10 minutes of tilt the 1/4″ CCD was found to have six meters of positional error. A six-meter by six-meter error would equate to 0.009-acre error that is well below the +/- 0.1-acre measure limit of the USDA FSA GIS application (Williams 2002).

eRTK-GPS for Direct Georeferencing

Basic GPS receivers with discontinued selective availability are capable of a 10-meter autonomous position with a single frequency L1 C/A coded receiver. At least four satellites in a good geometry generating a low dilution of precision (DOP) must be tracked to obtain a 3-D solution. This can be achieved easily without the need for differential corrections. Stand-alone positional GPS has been investigated as a means of collecting ground control points (GCPs) for use in the rectification of aerial images. The use of autonomous GPS in an aircraft has been attempted in several applications where ground control cannot be obtained. The ground control points are correlated to pixels, and the pixel size is based on the type of digital charged couple device (CCD) chip used and the flying height. As seen in Figure 2, the ground resolution of each pixel decreases with flying height for a 1/4″ CCD digital camera.

The development of the GIS is completed with verification and digitization using USGS 7.5 Minute Digital Ortho Quarter-Quadrangles (DOQQ) images. The DOQQs can be used to further improve the georeferencing by warping the small-format aerial images, as proposed by the USDA FSA (Davis 2002).

Differential GPS uses two types of methods to improve GPS positional tolerance. The first method is a code or pseudo-range solution. Differential-code solutions utilize base stations placed on known GPS coordinates locations to remove systematic and random errors from satellite signals. Differential code receivers have the capability of generating sub-meter positional accuracy at one standard deviation, based on the course/acquisition code or the precise code of the GPS. This accuracy can be achieved with both single- and dual-frequency receivers tracking the L1 and L2 frequency signals of the GPS. The process of differential GPS requires that the base-station and rover-station data be post-processed together to determine the errors and generate corrections.

This method is successful and is employed by the USDA Forest Service in tree acreage determination (USDA FS 1996). However, there are two disadvantages to using post-processing differential code solutions for direct georeferencing. One is that it is not possible to determine sub-meter positional tolerance until after the flight is completed and data are post-processed. While this can be amended with utilizing real-time differential GPS there still remains the possibility that a two or three standard deviation solution could be generated setting the linear accuracy above the requirement of the GIS. For many applications and GIS work in agriculture, sub-meter GPS accuracy is adequate for mapping and georeferencing data.

The second method of differential GPS is called differential carrier-phase. Carrier-phase differential GPS surveying systems provide centimeter accuracy based on a rapid search to determine the carrier-phase ambiguity or total number of unknowns, from the satellite to the receiver. This can be done in a kinematic process where a static base station collects data concurrently with a kinematic rover. The data are post-processed by combining the base and rover data and generating a centimeter solution.

Applications using eRTK-GPS solutions receive base-station information in a dynamic mode. Currently, modems and standard land communication methods pose a challenge in testing the eRTK-GPS. Typical UHF and Spread Spectrum radios have not been usable at high communication rates because of overland obstruction propagation. Wireless cellular devices such as GSM, CDMA, and G3 allow for the sharing of data at high rates with little to no signal loss. With wireless technology, communication solutions and permanent reference station array solutions are effective within a 35-km radius while maintaining carrier-phase ambiguity of two centimeters horizontal and twice that vertical. Tests show that at 35 kilometers from the base-station, results are obtained within two minutes of being observed and have a root-mean-square error (RMSE) of one centimeters horizontally and two centimeter vertically at two sigma (Lyle and Woods 2002).

The ability to obtain long-range solution is vital to the FSA GIS application, given the need to acquire images over large areas in remote agricultural counties. With this type of a scenario, an eRTK-GPS position could be directly used to compute the exterior location of the exposure station of a digital frame camera. Testing this concept will be a key element in developing an automatic georeferencing system for the USDA FSA crop compliance GIS application (Davis 2002).

Recently, eRTK-GPS solutions have been applied in wide-area spatial reference frame networks (such as the Texas Department of Transportation) to support multiple uses. Precision agriculture and construction are utilizing eRTK-GPS solutions to automate machinery by controlling steering, harvesting, earth movement, and real-time mapping activities. Technology growth in the commercial market and opportunities for free-of-charge correction in state-operated wide-area networks will generate more available resources for the USDA FSA utilization of this methodology.

Equipment Configuration

Small-format digital camera systems use CCDs to record the reflectance values with small light energy sensors. Most commercial chips are 1/3″1/4″ in size. A CCD has a standard pixel array that makes converting into raster format straightforward. Several small-format digital camera systems were evaluated and tested for this research. Kodak DC290, which has a digital frame of a 1901 x 1212 pixel array or 2.3 mega pixels generating an uncompressed file of 6 Mb, was selected. Figure 3 shows an example of an aerial image shot with the Kodak DC290.

The pixel size of the CCD is approximately 4.2 microns, or 4.2 x 10^sup -6^ m; the CCD array is thus 0.0080 meters by 0.0051 meters. Digital image data flowed from the camera to a laptop computer with an USB version 2.0 port on the notebook. As a backup, images were also left on the flash card memory of the camera.

Communication between the PC and the digital video camera can use the IEEE 1394 standard, often referred to as Firewire Technology. IEEE 1394 is defined as a serial data transfer protocol and interconnection system that provides the same services as modern IEEE-standard parallel buses, but at a much lower cost (Kulkarne et al. 2002). IEEE 1394 was created to allow universal interconnectivity, eliminating the need for many different input/output connectors. With IEEE 1394, one can connect 63 devices. Digital camcorders, scanners, printers, hard disk audio recorders, videoconferencing cameras, and disk drives all share a common bus connection, not only to an optional host computer but to each other as well. The IEEE 1394 standard defines three signaling rates, which, in precise terms, are: 98.304, 196.608 and 393.216 Mbps (megabits per second). These rates are referred to in the 1394 standard as S100, S200 and S400. In addition, the IEEE 1394 is “hot-pluggable,” meaning that devices can be added and removed without restarting the connected devices or PC. Table 1 provides a comparison chart between USB and the IEEE 1394 standard (Kulkarne et al. 2002).

The eRTK-GPS system tested comprised of Leica SR530 receiver and a Micro-Pulse airborne antenna mounted in a Cessna 152. The Leica SR530 was used for the base station with an AT302 chock-ring antenna to reduce multi-path. Data were broadcast from the base to the rover eRTK-GPS unit.

Several communication systems were tested for the eRTK-GPS solution. A ground test was made previously with the Sprint PCS wireless web at 19,200-baud rate. Airborne tests with the Sprint PCS were inconclusive, since PCS was moved to a G3 format, which resulted in communication connection difficulties. The aircraft and base would connect but they would stay locked for short periods only.

An airborne system test conducted with Verizon wireless communications confirmed that the system was unable to maintain a lock for longer periods of time, which might have been due to the cellular link jumping from tower to tower in rover locations above multiple visible towers. A third test was conducted with a UHF 35 Watt Pacific Crest radio; flying above the radio antenna enabled long-range data communication. Raw data and base station coordinates were logged for post-process verification while real-time data collection was in process.

The Kodak DC290 digital camera was mounted on the wing strut of the aircraft and the antenna was mounted directly above on the wing. An offset calculation of eccentricity was made from the antenna to the exterior body frame of the camera. The point was previously located on the camera in a laboratory calibration, and the phase center of the antenna was determined from literature published by Leica Geosystems. After the aircraft was manually leveled, a reflectorless total station was used to measure the offset eccentricity in 3-D. The values were entered directly into the GPS unit and applied to the offset computation. Based on previous tests by the USDA FSA with small-format digital cameras it was determined that no tilt sensor would be utilized for this GIS application. Using a 2-D affine transformation, a world file was written for each image.

Using digital imagery in GIS applications of eRTK-GPS for crop compliance will allow for speedier georeferencing without the need for complex tilt sensors. A tilt sensor could be attached to the camera and linked to the GPS sensor, so as to determine the binary string of the Easting and Northing, orthometric elevation, swing/yaw (kappa) from GPS, and roll (omega) with pitch (phi) from the digital tilt sensor. The USDA FSA transformed the Easting and Northing coordinates into a grid coordinate system that matched the CLU layer in the GIS. The elevation values were orthometric elevation values derived from a geoidal model loaded in the eRTK-GPS unit and applied on-the-fly. All eRTK-GPS positional data flowed to the computer using a National Marine Electronics Association (NMEA) LLK format string. A laptop computer with Windows 2000 and the control software was utilized.

A system calibration of the GPS and the small-frame digital camera were needed to assess systematic errors. Based on direct georeferencing of the exposure station for rectification, no aero-triangulation or block adjustment was performed. Latency of time intervals between image grab or exposure, and GPS logged time was noted and applied. Figure 4 illustrates the eRTK-GPS exposure station positions with a single, directly georeferenced image on a DOQQ.

Aerial eRTK-GPS System Results

The directly georeferenced image, which captured 26-kilometers from the base station, was compared to six GCPs from ground GPS survey. Results showed that the small-format digital image RMSE was + or -12 meters of that of the ground GCPs. A comparison of the direct georeferenced affine transformation with the georectified solution using the six GGPs (Table 2) suggests that the 12-meter difference could be attributed to measurements being taken at full one second while the camera image expositions were between the GPS measurements. Alternatively, it could be due positional error in the DOQQ where the GCPs were obtained. This positional error would not effect the acreage calculation because shifting the image to the proper location would not change the acreage of the tract. Direct georeferencing is ideal for overlaying the image on the crop tract being checked for compliance. For the USDA FSA, the most difficult aspect of aerial compliance monitoring is tying the georeferenced images to ground truth, i.e., the tracts of landowners and tenants who are in a crop compliance program.

A base station was set up approximately 30-kilometers from the airport where the aircraft started and within approximately 26-kilometers of a site the USDA FSA was checking for compliance. Raw carrier-phase and code data were recorded at a one second epoch rate in the aircraft and at the eRTK-GPS base-station. A National Geodetic Survey (NGS) Continuously Operating Reference Station (CORS) operating at the Southern Polytechnic State University Surveying Research Center in Marietta, Georgia, was configured to collect concurrent data at a one-second rate 40 kilometers north of the target area for the post-processing comparison. Results are shown in Table 3; Figure 5 shows the flight paths of the eRTK-GPS solution and the multi-station GPS networked solution. On a 10-km-long baseline, the RMSE values for position and elevation are approximately three and nine centimeters, respectively. On the longest baseline (27,154 meters), the real-time positional error was three centimeters in position and elevation as compared to the networked solution from two independent base stations (Table 3). Figure 6 shows the distance of the baseline in reference to time, and Figures 7-9 provide the residuals of the eRTK-GPS versus the post-processing multiple baseline solution. The post-processing solution is considered the most probable value to which the residual of the eRTK-GPS is computed.

The high accuracy of the results confirms the beneficial applicability of direct georeferencing to imagery obtained with small-format digital cameras and eRTK-GPS. No loss of carrier phase ambiguity was encountered during flight. The maximum error occurred as the aircraft was descending to land and radio communication was lost. An error of less than 0.5 meters was recorded when the aircraft banked over the target area. The images that were collected and georeferenced suffered from vibration from the mounting of the camera on the wing.

Conclusion

This research confirms that directly georeferenced eRTK-GPS can meet the minimum accuracy tolerance set by the USDA FSA for crop compliance GIS applications. The new approach removes the need for time-consuming post-processing of GPS data, transforms coordinates to a grid system, and applies them to a geoidal model with minimum delay. The new approach with small-format digital cameras is expected to shorten the urne required to collect digital data in support of the USDA FSA GlS implementation for crop compliance monitoring and increase the accuracy of digitization and thus the quality of the GIS. The system was expected to, and did, achieve centimeter level spatial accuracy in a real-time aerial platform of the exposure station as compared to post processing DGPS. Future publications and research will discuss improvements in mounting and image capturing. Given the good results obtained for direct georeferencing of aerial imagery, more research is recommended to achieve increased accuracy in concrete applications of this approach.

ACKNOWLEDGEMENT

This project is partially supported by a NASA-ERA grant, contract # ncc5-517. The author would like to express his sincere appreciation to everyone who has contributed valuable ideas and constructive suggestions to this paper.

References

Bailey, D., and L. Meador. 2002. Compliance, USDA FSA Illinois, [http: //www.fsa.usda.gov/il/page9.htinl; March 06, 2002].

Clark, R. 1997. Practices and potential: Assessing an agriculture revolution in progress. [http://www.precision(arming/ features/0497Clark.html; June 05, 200.1.

Cramer, M., and N. Haala. 1999. Direct exterior orientation of airborne sensors. Journal GIM International 13:46-9.

Davis, D. 2002. Using 35 mm digital cameras in the Aerial Compliance Program. Aerial Photography Field Office Internal Whitepaper, USDA FSA. [http: //www.apfo.usda.gov/whitepapers.html; May 5,2002].

Kulkarne, R., R. Bachnak, S. Lyle, C. Steidley. 2002. Design and development of an airborne multi-spectral imaging system. Algorithms and Technologies for Multispectral, Hyperspeclral, and Ultraspectral Imagery VII. 3425: 540-6.

Lyle, S., and L. Woods. 2002. Texas Department of Transportation Real-time DCPS Spatial Reference Frame. In: The Procedings of the 15th International Technical Meeting of the Satellite Divisions of The Institute of Navigation IONGPS 2002. Institute of Navigation, Portland, Oregon.

Mostafa, M. M., K.P., Schwarz, and R Gong. 1998. GPS/ INS integrated navigation system in support of digital image georeferencing. In: The Proceeding, Institute of Navigation 54th Annual Meeting, 01-03 June, Denver, Colorado, pp. 435-44.

Skaloud, J. 1999. Optimizing georeferencing of airborne survey systems by INS/ DGPS. Ph.D. dissertation, University of Calgary.

Skaloud, J. 2002. Direct georeferencing in photogrammetry. Photogrammetric Engineering & Remote Sensing 68(3): 207-10.

USDA FS. 1996. Summary report on FHP’s 1996 aerial photo missions using the Kodak CIR Digital Still Camera. [http://www.fs.fed.us/foresthealth/ technology/publications/96cirdigcam_ rpt/cirpaper.htm; March, 3, 2002].

USDA FSA. 1998. Common land unit. FSA Handbook, 8-CM, United States Department of Agriculture, Farm Service Agency, Washington, D.C. 20250.

Williams, K. 2002. United States Department of Agriculture Facts: Price Support. White Paper Publication. [http://www.apfo.usda.gov/whitepapers/ fsa_imagery.pdf; May 5, 2002.]

Wolf, P.R., and B.A. Dewitt. 2000. Elements of photogrammetry, with Application in GIS, 3rd ed. McGraw Hill, Boston, Massachusetts.

Stacey D. Lyle, RPLS

Stacey D. Lyle, Ph.D., RPLS, Assistant Professor of Geographic Information Science, Department of Computing and Mathematical Sciences, Texas A&M University-Corpus Christi6300 Ocean Drive, Corpus Christi, Texas 78412 Tel: (361) 825-3712; Fax: 361-825-5848. E-mail: .

Copyright American Congress on Surveying and Mapping Sep 2003

Provided by ProQuest Information and Learning Company. All rights Reserved