Ship track estimation from single hydrophone data
MetadataVis full innførsel
Passive acoustic range estimation and source tracking using hydrophones have been a problem because of the complexity of the underwater channel. This thesis is concerned with a fast method for the estimation of ship velocity and horizontal range to a surface going vessel using acoustic data from a single hydrophone. The problem is of interest in applications for monitoring of noise from commercial shipping and ship traffic. There has been many who have developed different method for similar problems and this thesis is based on a matched field inversion method. This method used the analytic Lloyd mirror propagation model which assumes constant sound speed, homogeneous horizontal layers and a single bottom layer, which has reduces accuracy at larger range (300 meter), but is simple and fast. The cost function for matching the recorded data and the modeled data was a modified Bartlett function which calculates the deviation between the ratio of the two data sets and was sensitive to variation in the signal. To reduce the variance in the recorded signal the welch spectral estimation was used. The ASHS and ASDE was implemented for a more efficient search through the parameter space, but a 2D search through the velocity and range parameters was also made to study the uncertainty of the estimations and the effect the different parameters had on the cost function loss. The results of the estimation was compared with the recorded AIS data of the unknown parameters. The results in this thesis indicate that a close estimation of the ship velocity and horizontal range was possible with the 2D search when presented with a signal with good SNR, reduced variance, low range, good estimation of the source depth, low search interval and a sufficient amount of data which can be used in the estimation. Which also makes it possible to use these parameters for ship tracking. An observation which where made was that the source depth changed the outcome of the estimation drastically. Close estimation with the global search algorithms required higher SNR and knowledge of the unknown parameters that could reduce the search interval.