旋转和平移双推测的ICP配准加速算法

Accelerated ICP Registration Algorithm Based on Expectation of Both Rotation and Translation

为了提高迭代最近点(Iterative Closest Point,ICP)的运算速度,在原始ICP及其加速算法(Accelerated ICP,AccICP)的基础上,提出旋转和平移配准参数双推测的加速算法。该算法通过对配准参数进行统一推测的AccICP进行分析,针对配准参数迭代求取过程中旋转和平移参数存在变化不同步的情况,提出了对该俩参数进行独立推测,只要任一参数迭代变化符合要求即可进行推测。实验结果表明,该算法比原始ICP具有较好的加速效果;与AccICP算法相比,做到了精准推测,减少了不必要的统一推测。

To increase the calculation speed of Iterative Closest Point（ICP） Algorithm, a new accelerated ICP is proposed based on the original ICP and its accelerated ICP（Acc ICP）. Both rotation and translation registration parameters are expected in the algorithm. The whole registration parameters are expected in the Acc ICP, and the changes of both rotation and translation registration parameters may not be synchronized during the process of iterative calculation. Both parameters are expected respectively in the new proposed method, and it can do expectation independently if one parameter’s change meets the requirement. The experimental results show that the proposed algorithm can accelerate calculation, compared with the original ICP. Compared with Acc ICP, it does accurate expectation, and avoids unnecessary unified expectation.