Abstract: In this paper, a terrain estimation algorithm based on kalman filter was proposed. Combining with gradient anomaly measured by gravity gradiometer, the algorithm achieved the purpose of terrain estimation by improving the accuracy and resolution of a low-resolution priori terrain. By introducing slope theory and rectangular prism method, the paper built system error model and measure equation of the kalman filter first. After that a simulation research was carried out to investigate the effect of algorithm. The simulation analyzed terrain mean noise, random noise and gravity gradiometer precision’s influence on estimation accuracy. At last, an area estimation simulation was also investigated. The results show that both the precision and resolution of estimated terrain have improved, which confirmed the correctness of the algorithm.