Calibrating a mobile camera's parameters

An analytical least-squares solution to the odometer-camera extrinsic calibration problem

“…The main drawback of this approach is that it requires measurements of known landmarks. Furthermore, the first step of this method is the same as [1], which ignores the measurement noise and cannot exploit more than two images.…”

“…However, since these techniques require the sensors to rotate around at least two different axes, they cannot be directly applied to a robot moving in 2D whose rotation is constrained around a single axis (i.e., the axis perpendicular to the robot's plane of motion). To address this issue, Chang and Aggarwal present an adaptation of the hand-eye calibration that relaxes this requirement [1]. In their method, the odometer-camera calibration problem is solved in two steps.…”

“…The main drawbacks of the EKF-based approaches are that they are iterative, they provide no guarantee of optimality, and their performance strongly depends on the accuracy of the initial estimate. One can initialize these iterative methods with the analytical solution of [1]. In this case, however, the EKF inherits the inaccuracy of the initial estimates, and may converge to a local minimum far from the optimal solution.…”

“…In [2], the second step of the approach proposed in [1] is extended to exploit additional measurements in a least-squares framework that is solved analytically. Moreover, the intrinsic odometer parameters, such as the wheel radius, are estimated simultaneously.…”

“…Several approaches exist in the literature for performing odometer-camera calibration. In [1], a closedform solution is proposed for the minimal problem of the odometer-camera extrinsic calibration. However, this approach is very sensitive to noise, and it cannot handle more measurements than the minimum required.…”