A calibration method with anistropic weighting for LiDAR and stereo camera system

Calibrating the extrinsic matrices between sensors is a significant pre-processing step of sensor fusion. Most of existing calibration methods use point-based rigid registration algorithm which considers the point coordinate error isotropic and uses the least square solution to estimate the extrinsic matrices. However, the error distribution of point coordinates is anisotropic due to the internal measurement properties of sensors, leading to decreased calibration accuracy. To solve this problem, we proposed an anisotropy weighting method: first we construct weighting matrices based on error distributions models of sensors; second we use surveying adjustment to further improve the calibration accuracy iteratively. We verified the effectiveness of our method through simulations. Compared with traditional methods, the accuracy is improved by about 45%. Moreover, our method can be applied in most of calibration methods to reduce the influence of anisotropic data and improve the accuracy.