Abstract: In tracking maneuvering targets application with radar, target dynamics are usually modeled in the Cartesian coordinates. In the cases target motion model are always very accurate but nonlinear ,while target measurements are directly available in the original sensor coordinates. By means of combinatin of unscented transformation and best linear unbiased filter , a new filter named unscented transformation-best linear unbiased estimator(UT-BLUE) filter was proposed to solve the above nonlinear tracking problem. In this filter, by way of nonlinear target motion model, unscented transformation was first used to predict state of the true target and its covariance, and then they were directly estimated while keeping the measurement error in sensor (polar) coordinate system. Algorithm analysis and simulation were conducted to compare it with extended Kalman filter(EKF) and unscented Kalman filter(UKF) , and results indicate that the new filter is more effective and available.