Consistency based calibration approach for visual-laser scanner

Recently, the robotics community has widely used color-depth (RGB-D) sensors for various applications, such as collision avoidance, navigation and robotic path planning. To effectively utilize the benefit of depth and color modalities, it is important to know the relative pose between these sensors. In this paper we proposed a theoretical and practical approach to estimate the rigid body transformation between these complimentary sensors (extrinsic calibration). The proposed work is based upon the exploitation of some common characteristics, between theses different modalities, by integrating a consistency criteria in the nonlinear optimization framework. The calibration procedure does not require any calibration object and directly operate on the visual correspondences. We have presented calibration results on low-cost built visual-laser scanner for indoor and outdoor environments. We have also compared our approach with the existing work and found the results to be more accurate.