Abstract: Aiming at the massive discrete tool position data generated in the computer numerically controlled programming stage,a non-uniform B-spline curve interpolation method based on curvature adaptive select-ing feature point is proposed under the condition of satisfying the preset interpolation accuracy.Firstly,the curvature of each discrete data point is calculated by the method of forming an approximate arc from three adjacent points,and the feature data points such as the segment point of the curvature segment and the maximum point of curvature are used as the initial interpolation data points to construct and generate the initial non-uniform B-spline interpolation curve.Secondly,the interpolation error calculation model is es-tablished and used to calculate the interpolation errors between all the data points not participating in the interpolation and the non-uniform B-spline interpolation curve,and the new feature points that are beyond the preset interpolation error in the curve segments are added to generate a new non-uniform B-spline inter-polation curve.Repeat the above processes until all data points that are not on the non-uniform B-spline int-erpolation curve can meet the interpolation accuracy condition. The simulation calculation results of the ac- tual tool position data show that the method can better retain the characteristics of the original data curve in terms of shape and accuracy even when a large amount of the original tool position data is removed, and it has advantages of fewer iteration calculations and larger removing the data points. The method will have high application value to the spline computer numerically controlled programming of massive discrete tool position data.