Abstract: This paper proposes a method,topologically constrained recognition method(TCRM) which can simultaneously recognize the blocks and their deformation parameters of a landslide by using only the surface displacements.In TCRM,the blocks are recognized according to whether the displacements on the blocks fit their deformation model properly.The judgment is a recursive process.The recognizing process is constrained by a topological matrix.As a statistical model recognition method,its idea may be used to recognize the blocks of any other objects only if each block has continuous linear characteristics.The recognizing process is accomplished automatically.In this paper,TCRM is first used to deal with landslide deformation model recognition.Some problems in practical use,such as the block concept,3D model and recognition approach,are discussed in detail.A famous landslide,called Baota Landslide,which lies in the Three Gorges Reservoir bank,is analyzed.The surface deformation of the Baota Landslide has been observed since 1997 using GPS receiver.We use the series observation data of six epochs to recognize the sliding block structure of every observation term.