Extracting extrinsic parameters of a laser scanner and a camera using EM

The robot's ability of perceiving their environments is primary functionality for a number of applications such as the rescue, navigation and manipulation. In these applications, the 3D colored map image is very useful information because it provides not only scene data but spatial data. For acquiring reliable 3D colored maps from the camera and the 3D laser scan system, one of the fundamental issues is the extrinsic calibration that calculates the rigid transformation (i.e., rotation and translation) between camera and laser coordinate system. In this paper we propose a novel method of getting extrinsic parameters which is insensitive to the noise of laser scanner measurements. Our approach introduces Expectation-Maximization(EM) algorithm to robust estimation of the relations between the camera and the laser scanner coordinate system in noisy environments. Simulations at various conditions with adding the laser measurement noise are performed to verify the effectiveness of our approach, and they show improved performances in estimating the extrinsic calibration parameters.