3D正态分布变换在多波束自动校准中的应用

Exploration and application of 3D normal distribution transformation in multi-beam automatic calibration

多波束测深数据存在横摇误差、纵摇误差、艏摇误差等,需进行校准。当前主要采用商业软件Caris进行人工手动校准,自动校准主要采用迭代最邻近算法等。针对Caris需要人工干预以及ICP容易陷入局部循环的缺陷,采用3D正态分布变换的配准方式,通过建立联合概率密度函数,采用似然函数建立匹配点云与目标点云之间转换关系。利用Hessian矩阵和梯度向量求最优化转换参数,完成多波束条带的自动校准。本文结合Caris中的人工配准结果,通过对3D-NDT算法匹配效果进行对比分析,验证了文中算法在平坦地区、斜坡区域等多种地形中都取得了更优的匹配效果,为实现多波束自动校准提供了重要的算法依据。

Multibeam Bathymetry System measured depth data has Roll errors,Pitch errors,Yaw errors,and it must be calibrated.Currently,commercial software named Caris is mainly used for manual labor calibration.Automatic calibration is the Iterative Closest Point algorithms(ICP).For the defects that the Caris requires manual intervention and ICP is prone to falling into a local cycle,this article attempts to use the 3D normal distribution transformation(NDT)configuration method to establish a joint probability density function and establish a matching point cloud Conversion between clouds and targets by establishing a combined probability density function.Using the Hessian matrix and gradient to optimize the conversion parameters to complete the automatic calibration of multiple beam bands.Combined with the manual registration results in Caris,this paper verifies that the algorithm achieves better matching effects in flat areas,slope areas and other terrains by comparing and analyzing the matching effect of 3D-NDT algorithm,and provides an important algorithm basis for automatic multi-beam calibration.