Calibrating multiple cameras with non-overlapping views using coded checkerboard targets

This paper presents an approach for combined intrinsic and extrinsic calibration of multi-camera rigs using coded targets. It is suited for conventional mono or stereo camera setups as well as for arrangements of multiple cameras with non-overlapping fields of view. We use a static scene consisting of multiple checkerboard targets and a sequence of images. This gives us a large amount of different observations and leads to precise calibration results. To solve the association problem of checkerboard corners over time and between different cameras, we use binary patterns that surround each checkerboard. A sparse nonlinear least squares solver is finally used to estimate the optimal parameter set. The parameter set contains intrinsic and extrinsic parameters, time series of the multi-camera rig pose, board poses and a description of board deformation.