Extrinsic Camera Calibration for an On-board Two-Camera System without overlapping Field of View

Abstract

Most recent developments in car technology promise that future cars will be equipped with many cameras facing different directions (e.g.: headlights, wing mirrors, break lights etc.). This work investigates the possibility of letting the cameras calibrate and localize themselves relative to each other by tracking one arbitrary and fixed calibration object (e.g.: a traffic sign). Since the fields of view for each camera may not be overlapping, the calibration object serves as logical connection between different views. Under the assumption that the intrinsic camera parameters and the vehicle's speed are known, we suggest a method for computing the extrinsic camera parameters (rotation, translation) for a two-camera system, where one camera is defined as the origin.

Global calibration of non-overlapping cameras: State of the art

“…Optical mirrors [34,35], [43], [47], [49], [51] [39] [ 36], [40], [48] [ 37], [45,46], [50], [52] Structure from motion [55], [57], [59][60][61] [ 53,54], [63], [66,67], [69][70][71] [56], [58], [62], [64,65], [68], [72,73] Laser projection [77,78], [80] [74-76] Close range photogrammetry [87] [ [88][89][90] range of applications in visual tracking, pose estimation and robot navigation. The method based on laser projection can theoretically work in a large range, but it is difficult to obtain high calibration accuracy because the image quality of the spot or the light plane decays as the work distance increases.…”

Motion-based online calibration for non-overlapping camera views

“…However, such approaches are hardly transferable to mobile camera rigs. Lamprecht et al [6] use well-known patterns from the scene, e. g. traffic signs, to determine the extrinsic parameters of two non-overlapping cameras in a vehicle. The drawback of this approach is that the localization of the pattern must be very precise to perform a precise calibration.…”

On-line non-overlapping camera calibration net

“…In this case, standard stereo camera calibration methods do not work. For this task, many methods have been proposed such as SLAM-based [3,5], mirror-based [1,14,18,19], tracking-based [15,2], trajectory-based [16,6], and AR marker-based [21,22].…”