This paper presents a methodology and algorithms to estimate an off-line and an on-line calibration of a street-view like vehicle and to compare both calibrations for a bundle adjustement based georeferencing of the system. The objective of the calibration is to estimate the pose of all cameras in the reference frame of the vehicle.
In the first part of the paper we will present our mobile mapping system composed of a multi-camera panoramic head and of two stereo rigs, and an Inertial Navigation System to provide the georeferencing of the imagery.
In the second part we will develop the off-line calibration method based on a bundle adjustment of different poses of the vehicle in a submillimetre-measured 3D target network. The target residues of the bundle in image space are usually smaller than 0.5 pixels.
In the third part we will present the on-line calibration method where the calibration is estimated during the survey itself with the help of automatically measured homologous tie points (SIFT).
In both calibrations we consider that the position of the cameras is known and corresponds to the mechanical design of the system. Thus the calibration process only estimates the rotations of the cameras in the vehicle reference frame.
Finally we present a method to compute the space resection of the mobile mapping imagery on centimeter-accuracy ground control points measured by surveyors. This resection is unavoidable because the positioning is degraded in urban areas, even with a very good Inertial Navigation System, due to GPS masks and multiple paths. The idea is thus to measure on the field one or several ground points (usually in the middle of the GPS gaps) observable in one panorama or image. We will discuss the impact on the one hand of the number of points and their position in the GPS gaps, and on the other hand the calibration procedure (non, off-line, on-line) on the overall absolute positionning error which is better than 5 centimeters even for 1km long GPS gaps. We will provide results on different data sets.