Automatic calibration of a multi-sensors system
This program propose an automatic multisensor system calibration
Let's suppose that you have a set of sensors (GPS / IMU, lidar, camera, ...) on a solid frame. The system sensors + solid frames (Solid) constitute a solid which means that for any couple of points X1, X2 belong to Solid dX1X2 / dt = 0 at any time.
hence, all sensors si have a constant position Tsi and a constant orientation Rsi in the solid frame coordinate system. We also suppose that you have the trajectory and the orientation over the time of all sensors relative to their own orientation / position coordinate system at time t0.
This tool will provide you an M-estimation of the calibration of all sensors using their trajectory / orientation over the time. To do that, we use the solid constraint of all sensors trajectory and orientation over the time. these constraints will lead to a non linear least square cost function solved using a Levenberg-Marquardt algorithm
In the end, the relative Orientation Rsi/sj and position Tsi/sj of the sensors will be estimated. The algorithm is proof to gaussian noise since it is a maximum likelihood estimator
For details about the maths, please read equations.pdf in the doc section