Abstract: In order to accurately obtain the underwater target position and velocity information, it is necessary to eliminate and correct the outliers in the long base line positioning. An improved residual detection method for the elimination and correction outliers is proposed. In this method, the absolute value of Kalman filtering residual error is taken as the judgment standard to identifying and eliminating outliers, and then the outliers are modified with adjusted Kalman filter estimation. Aiming at the problem that the filtering model does not match the actual motion to cause large deviation of data before and after filtering, the normal data will not be processed. The experimental results on the lake show that for the positioning trajectory where outliers exist, the root-mean-square error without removing outliers is 55.68 m, the root-mean-square error after processing by the residual detection method is 8.11 m, and the root-mean-square error after processing by the improved residual detection method is 2.04 m. Thus, the improved residual detection method can determine, eliminate and correct outliers in positioning trajectory, reduce positioning error, and improve the positioning accuracy of the long base line system.