Interferometric Synthetic aperture radar（InSAR） is based on the concept of observing the same scene with two slightly different radar trajectories.Each point of the scene is seen from two different positions along the two radar trajectories

and the differences between those points constitute the interferometric baseline.For quantitative SAR interferometry and differential interferometry applications

e.g.building digital elevation models（DEMs） or monitoring terrain displacement

the baseline must be estimated with a higher accuracy than generally achieved from satellite or airplane trajectories or attitude measurements.Traditional techniques for accurate baseline estimation need a priori information

like orbit data（to be known with a high precision）

fringe frequencies in flat areas

ground control points（GCP） and existing digital elevation models（DEM）.In many cases it is not possible to meet these requirements

i.e.neither orbit data or a digital elevation models（DEM） are available nor Ground Control Points（GCP） or flat terrain are within the scene.In this paper

a method of the InSAR baseline estimation is proposed based on Kalman filter and co-registration parameters of the interferometric SAR images and is not limited to the restrictions mentioned above.The proposed method are free from the following restrictions: 1.ground control points（GCPs）;2.existing digital elevation model（DEM）;3.flat terrain in the scenes;4.independence of orbit data.Moreover

the novel method allows the determination of time varying baseline parameters and is independent of the accurate knowledge of the height of the tie point.So the method can be applied when precise orbit data

ground control points（GCPs） or existing digital elevation models（DEM） are unavailable

or in mountainous regions where flat areas are difficult to be observed in the scene.The proposed method is made up of two steps: obtaining co-registration parameters of the interferometric SAR images and estimating the InSAR baseline based on Kalman filter.In the first step

the co-registration accuracy of the interferometric SAR images is required in 0.1pixels.In the second step

a geometrical relation between the pixel offset in range and the baseline components is derived

and the initial values of the baseline parameters are calculated using the forecasted satellite orbit data.Additional use of the knowledge about the stochastic characteristics of the involved parameters allows the determination of the baseline components by using Kalman filter.In the method a flat Earth approximation was considered

namely considering the terrain height as a constant equalling to the average value of the terrain height in the scene.The proposed method is examined using the ERS-1/2 Tandem data from the European Space Agency（ESA）.The data are imaged in Nanjing test site

which is located in the east of China.The results show that the proposed method can obtain the accuracy of baseline parameters at dm level and make up the limitations of baseline estimation when the accurate orbit data and ground control points can not be obtained.And the errors caused by orbit shift have been reduced.Thus

the method proposed can improve the accuracy of the DEM produced.