To develop a multi-modal in-traffic bridge deck scanning device, we need to calibrate extrinsic parameters of a ground penetrating radar (GPR). GPR output is in a non-Euclidean coordinate system because it only detects underground objects relative to road surface. When road surface is non-planar, its output cannot be trivially mapped to a 3D Cartesian system which is necessary for sensor fusion. We design an artificial planar bridge as the calibration device to ensure that the GPR output maintains Euclidean coordinate property during the calibration process. The bridge also allows us to adjust metal balls as calibration objects. We model the GPR imaging process and extract readings from hyperbolas generated from metal balls. We employ the maximum likelihood estimator to estimate the rigid body transformation and provide the closed form error analysis. We have conducted physical experiments to validate our calibration process. Results show that the calibrated model has an average error of 9.77 mm for testing samples. This is satisfying because the GPR signal wave length is 18.8 cm.