通常情况下，融合kmeans聚类与Hausdorff距离的点云精简算法在目标曲面的曲率值过小时需要设定Hausdorff距离阈值，在模型表面复杂情况下曲率估算精度不高，针对以上问题对该算法进行改进。首先在kmeans聚类中k值的确定采用手肘法确定聚类数保证聚类精度，然后采用维数特征Hausdorff距离代替主曲率Hausdorff距离提取特征点，避免了曲率的估算和在曲率值过小时设定Hausdorff距离阈值，最后融合kmeans聚类簇心与采用维数特征Hausdorff距离提取的特征点实现数据精简。采用实际扫描的点云数据进行验证，实验表明改进后的算法在相近精简率下提取的特征点更多，精度更高。

In the past, the point cloud simplification algorithm integrating K-means clustering and Hausdorff distance needs to set the Hausdorff distance threshold when the curvature value is too small in the flat region, and the accuracy of curvature estimation is not high in the case of complex model surface. In view of the above problems, we improved the algorithm. Firstly, the determination of K value in K-means clustering was made by using the elbow method to determine the clustering number to ensure the clustering accuracy. Then, the Hausdorff distance was used to extract feature points by replacing the main curvature Hausdorff distance with the dimension feature, avoiding the estimation of curvature and setting the Hausdorff distance threshold when the curvature value is too small. Finally, the K-means cluster core was fused with the feature points extracted by the Hausdorff distance with the dimension feature to achieve data simplification. The experimental results show that the improved algorithm can extract more feature points at the same reduction rate, and the reduction accuracy is higher.