K-Means++的声速剖面精简方法

Streamlined method for sound velocity profile based on K-Means++

在深海定位中,声线传播距离长、声速剖面层数多,采用常梯度声线跟踪虽然提高了定位精度,但明显降低了定位的计算效率。针对这一问题,本文引入优化聚类算法,提出了一种基于K-Means++的声速剖面精简方法。该方法将原始声速剖面的梯度分为正负2个部分,采用K-Means++对每部分进行初始聚类,再对聚类后的每一簇进行不同类别数的K-Means++聚类,将相邻相同类的层合并后得到精简声速剖面。采用精简声速剖面进行水下定位,并与原始声速剖面定位效果对比。实验结果表明:本文方法在确保原始声速剖面空间结构和水下定位精度的前提下,显著提高了定位计算效率,且在不同深度的定位精度与原始声速剖面定位精度保持一致。

In deep-sea positioning,the distances over which sound rays propagate are long and the number of layers in the sound velocity profile is large.The use of intra-layer constant-gradient ray tracing improves the positioning accuracy,but significantly reduces the efficiency of the positioning calculation.To address this problem,in this paper,we introduce an optimized clustering algorithm and propose a streamlined method for obtaining the sound velocity profile based on K-Means++.This method divides the gradients of the original sound velocity profile into positive and negative components,and uses the K-Means++algorithm to perform an initial clustering of each component,after which K-Means++clustering of different categories is performed for each cluster.The layers of adjacent classes of the same category are merged to obtain a streamlined sound velocity profile,which is used for underwater positioning,and which we then compared with the positioning result of the original sound velocity profile.The experimental results reveal that the proposed method can significantly improve the efficiency of the positioning calculation by ensuring the spatial structure and underwater positioning accuracy of the original sound velocity profile,with the positioning accuracy at different depths consistent with that of the original sound velocity profile.