Bathymetric Sidescan Sonar Bottom Estimation Accuracy: Tilt Angles and Waveforms

This paper presents a detailed analysis of bottom estimation performance of bathymetric sidescan sonar. Both general and closed-form expressions for the cross correlations of the received backscatter across a receive array are determined for a square pulse, match-filtered square pulse, and a chirp Gaussian pulse. The closed-form expressions clearly show the contributions that the various error mechanisms make to the correlation. Through geometry, simulation, and the Cramer-Rao lower bound (CRLB), it is demonstrated that when the array is tilted, a double-angle region results, which requires two angles to be estimated rather than just one, if the location of the bottom is to be determined correctly. Estimating two angles requires an array with at least three elements as opposed to two elements typically employed in simple relative-phase bathymetric sidescan sonar. It is shown through simulation that multiple angle estimates made with a simple linear prediction algorithm attain estimates near the CRLB and that the CRLB can be used to establish confidence limits. The bottom estimation performance related to a match-filtered Gaussian pulse and Gaussian chirp pulse are compared with that obtained with a match-filtered square pulse. Little difference in performance is found except when the angle estimation accuracy is thermal noise limited and then the chirp pulse performs better because of the higher resultant snr. A simple approximation procedure for predicting estimation performance is developed along with conditions for its application. Examples of bottom estimation performance are given for different array sizes, tilt angles, frequencies, and water types.