Role of Tie Points in Integrated Sensor Orientation for Photogrammetric Map Compilation

Direct measurement of exterior orientation parameters has been a challenge in photogrammetry for many years. Direct sensor orientation using a calibrated GPS/INS system can potentially eliminate the need for ground control points and aerial tiiangulation, and consequently, result in a great reduction in the cost and time of aerial photogrammetry. Previous studies have shown that, compared to conventional aerial tiiangulation, direct sensor orientation yields larger errors in the image and object space. It has also been shown that including a number of tie points within an integrated orientation approach can result in a reduction of errors in the image space. In this paper, the influence of the number and distribution of tie points on integrated orientation is investigated. Experiments with various numbers of tie points regularly as well as randomly distributed are presented. Results indicate that an increase in the number of tie points up to one point per model results in a considerable reduc­ tion of the errors in the image space, Introduction Ground control survey and aerial triangulation are the most costly and time-consuming stages in photogrammetric mapping projects. Direct measurement of exterior orientation parameters by a thoroughly calibrated GPS/INS system can potentially eliminate the need for ground control points and aerial triangulation, and consequently, result in a great reduction in the cost and time of aerial photogrammetry (Khoshelham and Eslami, 2007; Mostafa and Schwarz, 2001; Skaloud, 1999; Toth, 2002), With the direct measurement of the position and attitude of the camera perspective center at exposure times, the object-space coordinates of the image points can be computed using a least-squares forward intersection. This method is referred to as direct sensor orientation (Cramer and Stallmann, 2001; Yastikli and Jacobsen, 2005a). While direct sensor orientation is a promising method that can potentially reduce the photogrammetric mapping process to photography and stereo plotting, in practice, the range of accuracies obtained iising this method is generally Optical and Laser Remote Sensing Section, Department of Earth Observation and Space Systems, Delft University of Technology, Delft, The Netherlands, and formerly with the Contra of Excellence in Geomatics Engineering and Disaster Management, Department of Surveying and Geomatics Engineering, University of Tehran, Iran ([email protected]). lower than diat of conventional photogrammetry. Previous experiments with commercially available GPS/INS systems have shown that direct sensor orientation in the scale of 1:5 000 reaches accuracies that are two to three times lower when compared to the results of conventional aerial triangu­ lation (Heipke et al., 2002; Khoshelham et al., 2007). An alternative approach to determining sensor orienta­ tion parameters and transforming the image-space coordi­ nates to the object space is integrated sensor orientation (Ip, 2005; Jacobsen, 2004). In this approach, tie points in overlapping images contribute to the refinement of the directly measured exterior orientation parameters through a bundle adjustment, It has been shown that the introduc­ tion of tie points leads to a considerable improvement of the accuracy In the image space (Heipke et al., 2002). In effect, improved accuracy in the image space means lower y-parallax in the stereo-models, which is of great signifi­ cance in photogrammetric map compilation. It is well known that the stereo compilation of a pair of images with a y-parallax of larger than 20 microns is very inconvenient for the operators. Therefore, the potential of integrated sensor orientation approach in reducing the y-parallax is particularly worthwhile in photogrammetric mapping applications. From an economic point of view, integrated sensor orientation can be seen as a trade-off between direct orientation and conventional aerial triangulation since it eliminates the need for ground control points but still requires the measurement of tie points in overlapping images. An important issue in integrated sensor orientation is the number and distribution of the tie points. Despite the availability of automated point extraction and match­ ing algorithms, still many mapping organizations rely on the manual measurement of the tie points. To minimize the amount of this manual procedure, it is essential to know whether an acceptable range of y-parallax in stereo images can be obtained with a minimum number of tie points. Although previous studies have shown the effect of including a certain number of tie points, it is not known how the accuracy is influenced by variations in the number and distribution of the points. The objective of this research is to investigate the influence of the number and distribution of tie points on the accuracy of integrated Photogrammetric Engineering & Remote Sensing Vol. 75, No. 3, March 2009, pp. 305-311. 0099-1112/09/7503-0305/$3.00/0 © 2009 American Society for Photogrammetry and Remote Sensing PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING March 2009 305 sensor orientation. The focus will be on airborne frame cameras, as these are more commonly used in aerial map­ ping applications. The paper is structured in five sections. The next section describes the calibration of integrated GPS/INS system followed by tlie transformation of points from the image to the object space through direct and integrated sensor orientation. Experiments with various numbers of tie points in integrated sensor orientation are then presented followed by conclusions. Calibration of GPS/INS for Airborne Frame Cameras The calibration of GPS/INS IS basically a comparison of exterior orientation parameters measured directly by GPS/INS with those obtained by using a reference method (Forlani and Pinto, 2002; Honkavaara, 2004; Yastikli and Jacobsen, 2005b). The discrepancies are modeled by computing calibration parame­ ters mat relate GPS/INS position and attitude to the reference exterior orientation parameters. Bundle adjustment aerial triangulation is most often used as the reference method for the computation of exterior orientation parameters. Therefore, the determination of calibration parameters requires one or more test flights over a test field with preferably signalized control points. There are two main approaches to the compu­ tation of calibration parameters: one-step approach and twostep approach (Heipke et al., 2002). In die one-step calibration approach, a bundle adjustment of all available information in the image and object space is performed. The image space is related to the object space using the well-known collinearity condition equations augmented with a set of calibration parameters mat establish the relation between the INS, the GPS receiver, and the camera (Skaloud, 1999). The main calibration parameters include the unknown lever arm distance between the GPS/INS and the camera perspective center and the three misalignment angles that model the relative orientation of the INS with respect to the camera. Ideally, the camera exposure must be precisely synchronized with the GPS/INS; nevertheless, a possible time delay can be modeled by adding a synchro­ nization offset to the time of the GPS/INS measurements. Para­ meters of the interior orientation of UIB camera can also be estimated in this calibration procedure, provided that the calibration flights are designed in such a way that the effects of different parameters can be separated (Nilsen, 2002). The two-step calibration approach provides a more straightforward solution to the estimation of the calibration parameters. Since the two-step approach was adopted in this paper, it will be described in more detail in the following section. Two-step Calibration Approach A more straightforward way to compute the calibration parameters is to perform the aerial triangulation first, and then compare the estimated exterior orientation parameters with GPS/INS measurements, The discrepancies between GPS/INS measurements and the camera position and attitude parameters estimated in aerial triangulation are modeled by a polynomial function of time. The general form of the polynomial for position measurements is expressed as (Cramer and Stallmann, 2001):