Since the traditional statistical and physical algorithms in the inversion of refractivity from radar clutter (RFC) cannot track omni directions in real time, a new filter algorithm (extended Kalman filter and unscented Kalman filter) is proposed. The parameter equation of atmospheric duct, observation operator, state equation of filter arithmetic are derived separately. Finally, the implementation of the iterative inversion filter algorithm is derived. On the theoretical basis, two algorithms above are tested separately with or without considering the variation of refractivity with time. The experimental result indicates that unscented Kalman filter is suited to solve the nonlinear inversion problem, which has significance in theoretical foundation and technological support for practical applications for the future.