Least Square Innovation Method For Ocean Acoustic Tomography

Tomography refers to combining multiple one-dimensional projections into a single two-dimensional image or combing several poor images into a single improved image,which is an image formation problem.Ocean acoustic tomography(OAT),aims at image formation,allows to reconstruct the internal structure of large ocean areas based on the characteristics of acoustic signals transmitted through these areas.It’s a typical inverse problem.For image formation,the main three factors are data,model and algorithm,so this paper researched the OAT on these three aspects,used the least square innovation process to solve the inverse problem and made an experiment verification.Traditional acoustic tomography is based on ray theory,also perturbation method,academically.This method uses ray travel time between source and receiver to inverse the sound speed profile under vertical slice.The linear integral relation between ray path and slowness(reciprocal of sound speed)forms the basic function of acoustic inverse problem,which can be solved by method like least square(LS)of statistical theory.However,perturbation method may lead to low resolution and precision,due to the restriction of timing-space resolution of ray signal and the underdetermined problem caused by the larger amount of estimated parameters than observed ones.We proposed the method of empirical orthogonal function(EOF)to reduce the dimension of estimated parameter,modeled the sound speed profile time evolution to increment dimension and solved the inverse problem by innovation Kalman filter recursively.Compared with LS method,innovation Kalman filter uses the evolution knowledge of model and data,updates the old data with new data and solves the problem sequentially,which improves the accuracy and stability.Another key focus of this paper is OAT experiment data processing.This paper researched the technology about data processing and constructed the processing system of OAT.We use filter,Doppler shift calculation,matched filter to get the delays of signal,match them with eign-rays by beamforming and estimate the actual sound speed profile by innovation Kalman filter.Through the processing of experiment data,we validate the correctness of the theory and prove that using EOF,state-space model and Kalman filter can improve the accuracy and stability of OAT.