Improving sidescan sonar mosaic accuracy by accounting for systematic errors

Estimating the systematic errors associated with sensor geometry during data acquisition and applying the appropriate corrections can improve the accuracy of a sidescan sonar mosaic. We first review existing techniques for assembling mosaics and explain where systematic errors arise within typical sidescan sonar surveys. We then describe a new technique called systematic error reduction and compare its advantages to the classical "warping" or "tie-point alignment" technique. The main advantages of the systematic error reduction technique are that it works on an absolute, geographic reference frame, does not require fully overlapped images, and can be applied/modified in real time. We also describe how to find the optimal vector offset and show the results obtained on a real data set. As the success of our method depends on user interaction (e.g., identifying seafloor features or targets to be used for vector offset estimation, specifying acceptable offset ranges), considerable effort has been made to develop an intuitive GUI as well as a set of consistency tests to help guide the operator toward appropriate choices. Consistency tests include checks on whether a reference target has been allocated to the appropriate cluster, visual confirmation of whether targets are converging or diverging with vector offset application, and graphs for each offset component showing how its range of values affects the global dispersion of clusters. Finally, we provide some thoughts on future development directions, in particular to the estimation of offsets that vary spatially and temporally