Extrinsic Calibration for Vehicle-based Mobile Mapping System

Abstract

Having the advantage of 360°imaging and rotation invariance,panoramic camera has gradually been used in mobile mapping systems(MMS).Calibration is an essential requirement to make sure that MMS can get high quality geo-information.This paper presents a way to address the extrinsic calibration for vehicle-based MMS composed of panoramic camera and Position and Orientation System(POS).Firstly,control points in the natural scene are set up,whose spatial coordinates are measured with high precision.Secondly,apanoramic spherical model is constructed and panoramic image can be projected to this model by means of spherical reverse transformation projection.Then,localize and select the control points in 3Dspherical panoramic view but not in panoramic distorted image directly,the spherical coordinates of control points in panoramic image are gotten.After points correspondence is established,make use of direct geo-reference positioning equation and coordinate transformation,the translation and rotation parameters of panoramic camera relative to POS are computed.Experiments are conducted separately in space city calibration site located in Beijing and the Binhai New Area in Tianjin using our approach.Test results are listed as follows.When the GPS signal are of good quality,absolute positioning mean square error of a point is 10.3cm in two-dimension plane and 16.5cm in height direction;Otherwise,it is 35.4cm in two-dimension plane and 54.8cm in height direction.The max relative error of distance measurement is about 5cm over a short distance(distance3km),which is not obviously affected by the GPS signal quality.