Statistical Characterization of Multibeam Bathymetry Data.

摘要

Multibeam sonar systems produce high-density depth data of the ocean floor. Point pattern analysis of the individual depth points is conducted to determine an appropriate parameterized point-to-event distance distribution, F(y). This distribution is needed to quantify the lower bound on the estimated position error of a terrain-based navigation system being developed. The null hypothesis assumes that a Poisson point process produced the two-dimensional pattern. Under the null, F(y) has been shown to follow an exponential distribution where y is the distance from any arbitrary point in the study area to the nearest point in the pattern. Several tests of the null hypothesis are conducted. These tests lead to a rejection of the null in favor of a regular alternative. Probability plots and likelihood ratio tests are used to suggest an appropriate distribution, F(y), for the regular point pattern. These tests suggest that a two parameter Weibull distribution with increasing hazard rate may be an appropriate model. Monte Carlo simulations are conducted to determine goodness-of-fit of the Weibull model.