Abstract: The continuous multi-scale representation is one of attractive problems in mapping technology. Conventional mapping which uses the subdivision of multilevel tiles or multi-version storage to accomplish spatial multi-scale representations has the problems including data redundancy, inconsistency between diffident levels, the dramatic change in scale range and others. This study comprehensively applies various multi-scale representation strategies, including the cartography generalization and the spatial index technology, and put scale dimension and operations information in data model to build The Vector Spatial Data Pyramid Model. This model builds up key scale representations data and controls the scales space of every target through the off-line processing. And the process information, including the representation status of object, the longitudinal association between object A and object B in diffident levels, the generalize operators and parameters, is encapsulated in the attributes of objects through object-oriented modeling. Meanwhile, this model establishes scale transformation structure through one end and both end of the control. Based on the pyramid model, the generalization action can be executed in real-time to obtain continuous scale representation.