Generalization is a procedure used extensively in GIS data reduction, where polygon shape generalization is considered the most complicated. Since most of polygon generalization methods need to convert the polygon to the poly-lines, the generalized result is significantly affected by the decision of the starting point, which is required in initiating the poly-line generalization algorithm. The purpose of this study is to develop a curvature-based generalization technique for polygon GIS data. Curvature is an obvious characteristic of a curve and it can be used to represent the curve quantitatively. In this study, the shape's feature points which with relatively large curvature are detected initially. Then the original polygon can be separated into several curves by the feature points and each curve will be simplified respectively. In order to reach different stage of the simplification, a shape distortion index is used as a criterion to decide those data points should be preserved. An experiment is performed to compare the proposed method with the Douglas-Peucker method. The comparisons show that the proposed method has better simplified result and less distorted shape both visually and quantitatively than Douglas and Peucker method.