INDEPENDENT AND RIGOROUS ASSESSMENT OF STANDARD AND REFINED METHODOLOGIES FOR GPS/IMU SYSTEM CALIBRATION IN AERIAL PHOTOGRAMMETRY Marica Franzini DIET – University of Pavia - Italy – marica.franzini@unipv.it KEY WORDS: Photogrammetry, GPS/INS, Direct Georeferencing, Calibration, Accuracy. ABSTRACT: The paper concerns selected results from the PhD thesis regarding calibration of GPS/IMU systems and quality of direct georeferencing in photogrammetry. Thanks to Pavia's Test Site (PTS) and to a complex structure of flights which were acquired above it, it is possible to perform rigorous and independent validation of results. This means: the possibility of calibrating on one flight and validating on another, totally independent; the usage of disjoint sets of points for calibration and for assessing the results of direct georeferencing. Issues are also investigated such as short term time stability of calibration and reestimation of camera focal length. 1. INTRODUCTION The thesis concerns direct photogrammetry and in particular some aspects of calibration strategies and procedures as well as result assessments. A preliminary careful reading of reference bibliography, both at national and international level, allowed a definition of the state of the art of the technique. In this first phase of the work problems connected with direct georeferencing (DG) were faced through the study of several project reports and papers. The main goal of this approach was to analyze and understand the new aspects introduced with inertial systems. This analysis also allowed a comprehension of which problems had already been faced and totally or partially solved and, last but not least, to recognize which issues are still open. The investigation described above was also useful as a point of reference for the planning and the creation of an ad hoc test-site and of an ad hoc data-set of flights. The realization of Pavia’s Test Site (PTS) started at the end of the 90’s in order to become a support for different purposes such as GPS survey, traditional photogrammetry and laser scanning acquisitions. During the last few years, several phogrammetric data-sets were also acquired with the aid of inertial systems. These data, even if useful for some analyses, weren’t sufficient for a deep and rigorous assessment of direct photogrammetry. To fill this lack, the test- site was enlarged with new ground points (both artificial and natural) and with new flights. As mentioned above, one of the main goals of the thesis was the analysis of calibration strategies and procedures. To achieve this purpose, a detailed study of the reference systems involved and of their mathematical relations was realized. This investigation was conducted for a double aim: to better understand the results of the calibrations performed with the commercial software and to realize a self-written program for the re-estimation of the calibration parameters. The second goal of the thesis is the assessment of calibrations in terms of accuracy and of presence of residual parallaxes. In order to better evaluate the results coming from DG, several traditional aerial triangulations (AT) were performed, one for each of the six blocks involved. This approach allowed the determination of the potentiality of the data and to assess the DG by the comparison between the AT accuracy results and those coming from direct sensor orientation (DSO). Moreover, two step procedure was chosen for the calibrations and the ATs, calculated in this phase, were also used for this purpose. Calibrations were performed both with commercial software and with a self-written one. Nevertheless, within the thesis, only results obtained with the latter were shown. This choice guaranteed the risk of an incorrect interpretation of results. The next step was the validation of the performed calibrations. Different strategies were adopted to evaluate the final accuracy of calibration parameters. To take advantage of the data-set structure, some independent validation approaches can be realized. For instance, it is possible to calibrate with one block and to apply the obtained parameters to another block flown at a different altitude; moreover the same parameters can be applied to a similar block (same altitude) but flown on another day. This approach allows the assessment of the quality of calibration parameters and the detection of the presence of possible systematic errors. All the validations were performed with specific programs written in Matlab language. Finally, taking into account the results of the validations, a further analysis on calibrations was performed focusing the attention on the re-estimation of focal length. Taking advantage of the presence of three flight heights, focal length was re- estimated and new calibrations were performed. Finally, some conclusions on direct photogrammetry and sensor calibration were drawn. As mentioned above, most of the elaborations contained within the thesis, were performed with specific programs purposely written in Matlab language. The author of the thesis ad of the present paper had only contributed to drawing up a few modules. 2. PAVIA’S TEST SITE Between 2003 and 2005 a national project in inertial positioning in photogrammetry was realized and was concluded. The project leader was Riccardo Galetto from the University of Pavia and the title of the project was “Integrated Inertial Positioning Systems in Aerial Photogrammetry”. The aim of the project was to investigate almost all the features of inertial positioning: sensor calibration; precision and accuracy; direct sensor orientation (DSO) versus integrated sensor orientation (ISO); stability of calibration; influence of Ricerche di Geomatica - SIFET Edizioni - 2007