A new algorithm for on-line estimation of moisture distribution in non-hygroscopic granular materials is presented. The dynamics of moisture movement in granular materials under drying are frequently modeled by non-liner diffusion equations. By application of Kirchhoff transform, the equations are linearized and then approximated to a lumped parameter system (LPS) model by means of the finite integral transform technique. Kalman filter is applied to the LPS model to estimate moisture distribution. The results show that the moisture distribution can be estimated satisfactorily from measurements of surface and average moisture content.