Measurement model development for the correlation of imaging sonar acoustic shadows and bathymetry for ROV terrain-relative localization

This paper details the development of a probabilistic measurement model for the use of sonar imagery as an automated pilot aid for localization of a remotely-operated vehicle (ROV) with respect to an a priori bathymetric terrain map. Specifically, acoustic shadows in sonar imagery are correlated with expected visibility images generated from a bathymetry map, where acoustic shadows are significant drops in the sonar image intensity. An expected visibility image is generated for a given ROV position estimate with respect to a stored terrain map by extracting a bathymetry profile along the sonar scan plane and then evaluating a metric termed “differential height” to quantify visibility probability in the pixel space of the sonar image. This paper provides the theoretical foundation behind the approach presented in [1], and further details the tuning of measurement model parameters using ROV sonar image field data collected in collaboration with the Monterey Bay Aquarium Research Institute (MBARI). The position estimation results for one dataset from [1] is repeated in this paper in order to demonstrate localization performance using the proposed approach.