Bearing-only localization is important in passive localization. Influenced by geometry and measurement errors, the precision is unfavorable. To track target rapidly and accurately, the implementation of best linear unbiased estimation (BLUE) filter in bearing-only system is explored. The converted measurements model in Cartesian coordinate is built, the statistics of converted measurements are derived, corresponding BLUE filter is deduced. Numerical comparisons with two state-of-art techniques show: proposed algorithm outperforms EKF in precision because it can estimate the high-order statistics of converted measurement errors; meanwhile its computation cost is lower than UKF with equal performance. Proposed filter can provide optimality under different state noise levels, which proves its applicability.