A nonlinear filtering algorithm for Markov processes based on a piecewise-polynomial representation of the components of an estimate of the state vector is developed on the basis of the spline-approximation method by using a third-order kinematic model in the primary measuring instruments.