Spatial structural patterns are very important reformation inflecting regional characteristics. However, they are seldom explicitly expressed in geo-data bases. In order to express them in generalized maps derived from geo-data set, spatial structural units need to be identified beforehand. This involves in the classification, identification, description and generalization of spatial structural patterns. This paper includes three sections. The first two parts discuss the classification and identification of spatial structure patterns. Spatial distribution can be subdivided into three categories, i.e., point, linear, and areal phenomena, which can be further divided into uniform, agglomerate, or random point/areal patterns, and isolated, grouped, nested, dendriform, or netlike linear patterns. Spatial structural units can be identified by means of comparison between digital model and theoretic model, which are derived from geographical entity. Theoretic models of structural patterns are usually described by a group of parameters, such as polarity, linearity, and symmetry. The third section studies the method to generalize spatial structure of point, linear, and areal phenomena. Among point and areal distribution, clusters are the most important structural units. Clusters can be generalized by means of boundary adjustment and inner structure simplification. As for linear distribution, dendriform and netlike patterns are the most important units. The generalization of both of them involves in trunk lines selection and typification of characteristic patterns, which need to be dealt with different methods according to concrete situation.