High Precision Georeferencing using GPSINS and Image Matching BIOGRAPHIES

BIOGRAPHIES
Michael Cramer is a research associate in the Institute of Photogrammetry. He received his diploma in surveying from the University of Stuttgart in 1993. From 1993 till 1995 he was employed in the special research group "SFB 228 { High precision navigation", mainly related to the topic attitude determination using multiantenna GPS and low-cost INS. Now he is working in the eld of GPS/INS integration for georeferencing of digital airborne and satellite three line CCD scanner. Dirk Stallmann took his Dipl. Ing. in Geodesy and Surveying Engineering from the University of Hannover, Germany, in 1989. From 1989 till 1995 he was working at the Swiss Federal Institute of Technology, Zurich, on image matching for image/object tracking, precise measurement and automatic digital terrain model generation from airborne and satellite images (SPOT, MOMS02). Since 1995 he is working at the University of Stuttgart on photogrammetric processing of airborne and satellite based three line CCD scanner and sensor integration. Norbert Haala received his diploma in surveying from the University of Stuttgart in 1990. He started as a research associate at the Institute for Photogrammetry as a member of the special research group "SFB 228 { High precision navigation". He has been working extensively in the eld of image reconstruction and nished his Ph. D. thesis "Building reconstruction from aerial images and DSM data" in 1996. Since 1995 he is associate professor and head of the research group "Sensor-Integration". (INS). Additionally, due to the along-track stereo capability of the camera error terms for position and attitude can be estimated by photogrammetric constraints utilizing ground control points and corresponding image points. In order to get a high accuracy position and attitude are determined by integrating all available information (GPS, INS and stereo image data). Within the article the geometric processing of the high precision and high resolution scanner imagery will be described and results of the performed test ight in October 1996 will be presented.
ABSTRACT
The registration and geometric recti cation of airborne scanner imagery is an essential prerequisite for the processing and analysis of this type of images. The Digital Photogrammetric Assembly (DPA) is an airborne camera consisting of three pan-chromatic line arrays for stereo imaging and four line arrays for multi-spectral imaging. For georeferencing the sensor system is completed by a module consisting of a di erential GPS receiver con guration and an Inertial Navigation System
determined in a least squares adjustment.
2 THE DPA SENSOR SYSTEM
The devtem started in the late eighties and was done by the Daimler{ Benz Aerospace (DASA), formerly the Messerschmitt{ Bolkow-Blohm (MBB) company. The philosophy of the system is to get single path pan-chromatic stereo coverage and multispectral coverage of the terrain. This is similar to other imaging airborne sensors like the MEIS imager 7] or the MOMS-02 sensor 1] for space applications. Originally, the camera system consists of three panchromatic CCD line arrays for the in- ight stereo imaging connected with an INS providing linear accelerations and angular rates measurements. First tests to evaluate the potential of the stereo module were done in 1992 and are published in 5]. Later, the camera was completed by the multispectral module for acquisition of multispectral images in the red, green, blue and nearinfrared band. Additionally, a di erential GPS receiver con guration was synchronized with the system. The optical part together with the sensor electronics and the INS are mounted in a Zeiss Jena SM2000 platform for stabilization of the imaging system during the ight. The components for operating the system and storing the recorded data are xed in two racks. The complete camera system is shown in gure 1.
1 INTRODUCTION
The DPA is an integrated airborne push-broom scanning system with real time data collection capability. The fundamental idea of this system is to obtain high quality 3D and multispectral information of a scene simultaneously. In order to acquire geometric information three high resolution CCD lines are used providing panchromatic images in forward, nadir and backward direction of the ight path. The ground pixel size for these images at the ying height of approximately 2000 m above ground is in the order of 0.25 m. For multispectral purposes four CCD lines are used to acquire images of the spectral bands red, green, blue and near infrared. For georeferencing of the digital image data, a six degree of freedom strap-down INS is utilized. Additionally, a di erential GPS receiver con guration is synchronized to overcome the well known systematic INS error e ects. Hence, for every scan line the fully exterior orientation parameters of position and attitude are available. More details about the georeferencing process in general are given in e. g. 8], 9]. Within this paper the algorithm for georeferencing the image data is described. In contrary to the common approach of determining the exterior orientation for the complete trajectory by using a Kalman lter starting with an initial alignment procedure, a strip-wise approach is utilized. The in- ight alignment of the INS is realized using ground control points in the three channels. By using the collinearity condition of the the o sets and drift (linear rates) for the attitude angles are