Nonlinear model outliers removed algorithm and its application in underwater positioning

This paper presented an algorithm of outliers removed for nonlinear model and the threshold selection method, which was used in ultra-short baseline (USBL) positioning system for the installation of underwater calibration. At first, the assumption that most of the observation points were correct was given. Then a parameter was set up on the data based on it. Larger the value of the parameter of one point was, greater the probability was that the point was a good one. Smaller the value, less the probability. It was combined with the least squares estimation algorithm. Before a USBL system was used, installation calibration was needed. And the validity of the data had a significant impact on the calibration. Considering the nonlinear of the calibration model, the outliers removed of nonlinear model was chosen as a data preprocessing method, and the threshold was given. With the data preprocessing method, it could get more accurate estimation of installation errors. So it improved the position accuracy of USBL underwater positioning system.