Due to the problem of redundant point cloud data in the automatic grinding operation of castings,the data processing time is too long and the processing efficiency is affected.In order to improve the computational efficiency in point cloud processing,a point cloud reduction method was proposed based on linear coupled curvature to combine efficiency and accuracy.A K-d tree data structure was constructed for the source scan point cloud,and a more robust linear coupled curvature feature model was proposed to extract feature domain.On this basis,the improved spatial dichotomy method was used to quickly simplify the non-feature domain.After fusing the feature domain and the simplified non-feature domain,the final simplified result was obtained.The results show that the method can obtain accurate reduction ratio,and there are only small data holes in the fitted surface model.Under the condition of 90% reduction ratio and 2.7% noise ratio,the conformal error is 0.867 mm,which is better than the traditional point cloud reduction method,and can realize the efficient reduction of casting point cloud.