Scan registration for underwater mechanical scanning imaging sonar using symmetrical Kullback–Leibler divergence

Abstract

Due to its advantages in size and energy consumption, mechanical scanning imaging sonar (MSIS) has been widely used in portable and economic underwater robots to observe the turbid and noisy underwater environment. However, handicapped by the coarseness in spatial and temporal resolution, it is difficult to stitch the scan pieces together into a panoramic map for global understanding. A registration method named symmetrical Kullback-Leibler divergence (SKLD)-distribution-to-distribution (D2D), which models each scan as a Gaussian mixture model (GMM) and evaluates the similarity between two GMMs in a D2D way with the measure defined by SKLD, is proposed to register the scans collected by MSIS. SKLD not only weights the difference between distributions with the prior probability but also increases the numerical stability with the symmetrical constraint in distance measure. Moreover, an approximation strategy is designed to derive a tractable solution for the KLD between two GMMs. Experimental results on the scans that were collected from the realistic underwater environment demonstrate that SKLD-D2D dramatically reduces the computational cost without compromising the estimation precision.

Underwater Pose SLAM using GMM scan matching for a mechanical profiling sonar

“…To fix that, unstructured methods could be used. Jiang et al (2019) Dempster et al (1977). Tabib et al (2018) proposed to use the EM algorithm to fit a GMM to each scan.…”

“…To fix that, unstructured methods could be used. Jiang et al (2019) proposed to use $$K$$‐means algorithm to fit a point cloud into a GMM. They use a fixed number of components $$K$$—depending on the number of points in the scan—and, instead of minimizing the $${{\rm{ {\mathcal L} }}}_{2}$$ distance like in the D2D algorithm, they minimize the Kullback–Leibler (KL) divergence between the GMMs that represent the two registering scans.…”

A Survey of Underwater Acoustic SLAM System

“…This process is repeated until the entire scan sector is covered. The scanning period often lasts few to dozens of seconds inducing the motion distortion problem of the MSS image if the vehicle's pose and location alters during the scanning process [18].…”

“…Later, works of [14,26] successfully apply the pIC algorithm to the MSS image registration taking the motion induced distortion problem into consideration. Further, [18] proposed to model both the reference scan and floating scan as Gaussian mixture models and used the symmetrical Kullback-Leibler divergence as the distance measure between two Gaussian mixture models. It was validated that this method could dramatically reduce the computational cost without compromising the estimation precision allowing lower intensity threshold to be applied to segment the MSS image.…”

Underwater Loop-Closure Detection for Mechanical Scanning Imaging Sonar by Filtering the Similarity Matrix With Probability Hypothesis Density Filter

“…Explicitly, we divided the global map into equally sized voxels and counted the occupied voxels, with a lower value indicating higher consistency. We refer the reader to [36], [37] for more detail. The map generated by the proposed filtering method has sharper walls (its crispness measure is 11359), which is also consistent with the physical environment.…”