基于观测概率有效下界估计的二维激光雷达和摄像机标定方法

A Calibration Method of 2D Lidar and a Camera Based on Effective Lower Bound Estimation of Observation Probability

针对2D激光雷达和摄像机最小解标定方法的多解问题,该文提出一种基于观测概率有效下界估计的标定方法。首先,提出一种最小解集合的分级聚类方法,将每类最优解替换原来的解集合,从而减少解集合样本个数。然后,提出一种基于激光误差的联合观测概率度量,对解集合元素的优劣进行度量。最后,利用聚类结果和观测概率度量结果,该文提出基于观测概率有效下界估计的有效解选取策略,将优化初始值从最优解转化为有效解候选集合,提高了标定结果的准确性。仿真实验结果表明,在真解命中率性能上相比于Francisco方法,该文方法在不同棋盘格个数情况下提升真解命中率16%~20%,在不同噪声水平下提升真解命中率6%~20%,有效提高真解比例。

Considering the multi solution problem of minimum solution method for calibration of 2D lidar and camera,a calibration method based on the estimation of the effective lower bound of observation probability is proposed.Firstly,a hierarchical clustering method with minimum solution set is proposed which should be used to replace the original solution set with each kind of optimal solution,so as to reduce the number of samples in the solution set.Then,a joint observation probability measure based on laser error is proposed to measure the quality of solutions.Finally,using the clustering results and the measurement results of observation probability,an effective solution selection strategy based on the estimation of the effective lower bound of observation probability is proposed,which transforms the optimized initial value from the optimal solution to the candidate set of effective solutions,and improves the accuracy of calibration results.Comparing with the existing methods,results of both simulation and real data experiment show that the proposed algorithm improves significantly the true solution hit rate by 16%~20%under different number of checkerboards and 6%~20%under different noise levels.