With computer technology and three-dimensional imag ing software and hardware technology rapidly develo pment, three-dimensional digital technology is more and mo re concerned and applied by people. But limit graph ics card handling ability and network bandwidth seriously af fect such media transmission, simply rely on increa sing hardware equipment input is not enough to solve pro blems, it should adopt corresponding algorithms to compress three-dimensional surface model data so that can fu ndamentally solve the drawback. Though some common used compression algorithms and tools can be used to red uce three-dimensional geometric data sizes, efficie ncy is not ideal or it is improper for some specific applicati ons, three-dimensional model and previous voice frequency( one-dimensional), image( two-dimensional ), video( two-dimensional+time) the three kinds of medias types are different, three-dimensional network mode l also reflects as random crook, complex and lackin g of continuous natural parameterization features, and e ach kind of attributes that are defined in model su rface is non-regular sampled, which let classic orthogonal a nalysis tools cannot be directly used to handle 3D geometric signal. Therefore, look for new algorithms to effec tive express three-dimensional model is imperative. The paper based on Garland proposed surface model compression algorithm, it puts forward a kind of edge collapse -based surface data compression algorithm, the algorithm c alculates edge collapse cost and defines each edge collapse order according to quadric error measure criterion, judges shrinkage legality according to half-space test method, by the algorithm, it implements surface model bound ary zone and interior zone synchronization compress ion, it not only can ensure original model geometric features o n the condition of high compression ratio, but also can effective ensure compression model quality.