Abstract: A modified Markov localization algorithm is given based on the correcting eigenvalue of the weighting matrix with the ridge estimation method. Through the modified algorithm, the problem about global existing of inverse of error covariance matrix has been solved, the problems caused by singular point in the application of Markov estimation algorithm have been improved. The simulation focusing on the accuracy geometric distribution of bistatic sonar system has been done. The quantitative analysis and comparison of the location accuracy of three data grouping types can also be obtained. The simulation results show that compared with optimum linear data fusion localization algorithm, the modified Markov localization algorithm has improved the accuracy of location; the positioning performance of the region near the baseline and the two base stations is also greatly improved, making the application of redundant data combinational estimation in sonar system widely.