AIRBORNE The Power of the POS/DG System Automatic sensor orientation using integrated intertial GPS By E. Lithopoulos, D.B. Reid, and Joe Hutton Differential GPS has greatly increased the productivity of aerial survey operations in recent years. With a sufficient number of satellites in view, DGPS is now capable of delivering position fixes rapidly and reliably to accuracies of a few centimeters using carrier phase techniques. This capability enables the surveyor to move quickly from point to point-maintaining positioning accuracies to within survey standards while operating "on the fly" (OTF). Still further gains in survey productivity can be realized by combining DGPS with precision inertial sensors. A system that combines DGPS with inertial offers full 6 degree-of-freedom, high-accuracy position and orientation solution (position + roll, pitch, heading), high rate outputs (200 samples per second), and continuity of data through GPS outages. These features provide the user with the capability to survey reliably from rapidly moving platforms. In addition, very high pointing accuracy can be achieved by combining carrier-phase differential GPS and precision inertial. With accurate orientation (angle) measurements available, high-resolution imaging sensors can be used to map the terrain remotely to survey-level accuracy. Large gains in surveying productivity can be realized in this way. Examples of such applications include high-altitude terrain mapping with scanning lasers and airborne photogrammetry using precision cameras. Conventional Photogrammetry and Aerotriangulation Conventional photogrammetry entails constructing an accurate terrain model or topographic map from airborne photography acquired using large-format precision film cameras. The basic approach is to acquire a series of photos of the terrain, with exposures timed such that each photo overlaps the previous by about 60% in the track direction. Typically, a grid pattern is flown so that photos taken on each line also overlap photos acquired on parallel adjacent lines by about 30%. Cross lines must also be flown to provide additional redundancy. As outlined below, stereo images (analog or digital) are constructed from the overlapping photos by using ground control points and common reference points appearing in the imagery to determine the correct orientation of each photo. Measurements of the position coordinates and elevations of ground features are derived from the stereo images. Maps are then generated from these data. In order to construct the stereo images and to reference these images to the Earth, the position and orientation parameters of the camera at the time of each exposure must first be determined. This is done using a procedure called aerotriangulation. Here, with ground control established at key locations in the survey area, the photos in which the ground control points appear are processed to determine the camera position and orientation parameters corresponding to these photos. In principle, this involves measuring the photo-coordinates of the control points and then solving a set of simultaneous equations with the position and orientation parameters as unknowns. Having established the position and orientation parameters for the controlled photos, the control is transferred from photo to photo by "bridging" through common points or features appearing in overlapping photos. The end result is a mosaic of photos whose relative orientations (with respect to each other) and absolute orientations (with respect to the Earth) are known. The process of construction of the photo mosaic is labor-intensive and time-consuming. In the final step, stereo pairs of photos are processed to measure the Earth-referenced coordinates of height contours and terrain features. Here, the Earth-referenced coordinates of the measured points are determined from the corresponding photo-coordinate measurements using triangulation and knowledge of the position and orientation parameters of each photo. Typical accuracies achieved through aerotriangulation are in the range of 1 part in 5,000 to 1 part in 10,000, depending on map classification. This corresponds to an on-the-ground positioning accuracy of 10centimeters to 20cm from a flying height of 1,000 meters. Expressed as an angle, 1:10,000 corresponds to an accuracy of about 20 arc seconds in the absolute orientation of the photos. The cost of performing aerotriangulation, including establishing ground control, is estimated as $30 to $50 per stereo model. Typically, a medium to large size airborne survey company will process from 5,000 to 10,000 stereo-models per year or more for a total cost of $150,000 to $500,000. System Description Applanix manufactures a line of integrated GPS inertial products designed for remote mapping, surveying, and precision motion compensation applications. This product line is referred to collectively as POS, standing for Position and Orientation System. A POS model specifically designed for airborne photogrammetry applications was introduced in 1997-referred to as POS/DG (Direct Georeferencing). POS/DG is a precision integrated GPS inertial system designed specifically to replace aerotriangulation in airborne photogrammetry. POS/DG provides this capability by measuring the position and orientation parameters of the aerial camera so that the relative and absolute orientations of the acquired photography can be determined to sufficient accuracy with minimal use of ground control and bridging points. The coordinates of ground points, height contours, etc., can then be measured directly from the stereophotography avoiding aerotriangulation altogether. Airborne Element The airborne element is made up of three main components: IMU The Inertial Measurement Unit (IMU), comprised of three accelerometers, three gyros, and signal processing electronics, outputs high-accuracy acceleration and angular rate measurements in digital form to the POS computer system. These measurements are processed to compute a full 6 degree-of-freedom position and orientation solution referenced to the IMU location, updated at 200Hz. The IMU is custom developed for airborne applications. It is a light-weight and compact unit that is hard-mounted directly to the aerial camera for maximum accuracy. POS/DG The GPS POS/DG is typically configured with a dual-frequency differential GPS receiver embedded in the POS computer system. Alternatively, POS can be interfaced to a user-supplied stand-alone receiver. PCS The POS Computer System (PCS) implements real-time processing, time alignment, data acquisition, and storage functions. The PCS offers simple two-button operation for on-off and data recording. Several hours of data can be recorded on the PCMCIA disk for post-processing. If required, display and control is provided by a custom application running under Windows on a laptop PC. Time synchronization with the camera is achieved with the time-of-exposure pulse from the camera acquired by the PCS via its discrete port. Post-Processing-POSPac The precise camera's position and orientation parameters are determined by post-processing of the recorded POS data. Post-processing is implemented in two stages. Forward Time In the forward pass, the inertial sensor data are processed using a strap-down inertial navigation algorithm to compute a complete, dynamically accurate 6 degree-of-freedom position and orientation solution for the camera. Update rates on the strap-down computations are at 200Hz, with position and orientation outputs referenced to the time of film exposure by time-alignment with the camera pulse discretes. The recorded GPS data are processed using carrier phase methods to obtain position fixes to about 5cm or 10cm accuracy at 1Hz. Positions logged from a dual frequency GPS base-station are used for differential corrections. The inertial and GPS data are blended using a Kalman filter configured typically with 22 to 35 states. The Kalman filter estimates the position, velocity, and attitude errors of the inertial solution and the residual sensor biases, scale factor, and alignment errors. These estimates are fed back to the inertial navigation computations to remove drift in computed position and orientation and to null the effects of the residual sensor errors. Backward Time The estimates of camera position and orientation obtained in forward time are derived using GPS and inertial measurements that occur prior to and at the time-of-validity of each estimate. Backward time processing incorporates GPS and inertial measurements into the computations that occur after the time-of-validity of each estimate so as to improve overall estimation accuracy. An optimal recursive smoother is used for this purpose. Accuracy improvements realized by smoothing can be very significant with improvement factors ranging from about 2 to 10 depending on the mission profile and dynamic environment. The WGS-84 reference ellipsoid is employed in forward and backward time processing of the data. On completion of backward processing, the computed camera position and orientation parameters are transformed from WGS-84 into the local survey coordinate system. This transformation includes conversion of the three orientation angles-roll, pitch, and heading-to the three rotations employed in photogrammetry. After this transformation, a file is output containing the six camera orientation parameters at the time of each exposure. This file is input to the photogrammetry process (digital or analog) to construct stereo models from the acquired imagery. Software that implements the functions described above is bundled with POS/DG. This application-called POSPac-runs on a PC under Windows 95/98/NT. It provides: data extraction tools (to extract data from the flight disk); graphical user interface; processing and display tools; optimal aided inertial navigator processing, both forward and backward time; differential code-phase and carrier-phase GPS processing (GRAFNAV from Waypoint); coordinate conversion from WGS 84 to local mapping frame; and plot file generation and graphical display. One hour of recorded data will typically take a little less than 1 hour to post-process with POSPac on a 266Mhz Pentium-II laptop with 64MB of RAM. Results Results of flight trials conducted using an RC30 camera are described here. In these tests, the POS/DG IMU was mounted on the RC30 camera and flown over the Ohio State test range. A total of 62 ground controls were used. Flying height was 2,400 feet. A pair of Ashtech Z-12 GPS receivers were used-one installed in the aircraft and the other as the base-station. "On-the-Ground"- POS/DG vs. Ground Control In this test, residual measurements from 62 ground control points were selected. POS/DG position and orientation data were used to directly compute the "on-the-ground" coordinates of the 62 surveyed reference points. Conclusions The flight trial results presented here demonstrate that POS/DG provides accuracies comparable to that of conventional aerotriangulation in measurement of camera position and orientation-position accuracy better than 20cm (RMS) per axis and orientation accuracy better than 30 arc seconds (RMS) per axis. An accuracy of 20cm to 35cm (RMS) per axis was achieved in directly georeferencing ground points from a flying height of 2,500 feet. This is well within photo-mapping standards defined for the majority of applications. POS/DG is now being flown operationally by a rapidly increasing number of users providing benefits through reduced operational costs and faster turn-around times. About the Author: E. Lithopoulos, D.B. Reid, and Joe Hutton work at Applanix Corporation in Toronto, Canada. Back |