基于离群算法组合曲面特征点提取的研究

Research on extraction of characteristic point of combined surface based on outlier detection algorithm

为了克服目前组合曲面提取特征点算法中阈值选取困难导致边界特征点误判的缺点 ,在对组合曲面特性进行分析的基础上 ,提出了一种基于离群算法的组合曲面特征点提取算法。该算法根据曲面特性定义了曲面域和曲面域深度 ,在空间统计学基础上引入正态分布的标准单位数和置信系数 ,采用空间数据挖掘中的离群算法提取组合曲面特征点。通过在某型摩托车零件中的应用 ,表明了该方法可以有效地避免阈值选取问题 。

In the existing methods of extracting the characteristic points of combined surfaces, to choose the proper threshold value, which affects the validity of the characteristic points, is difficult. A novel approach of extracting the characteristic points of the combined surfaces based on outlier detection algorithm was proposed. In this approach, the surface domain and its depth were defined according to the combined surfaces' characteristic. The characteristic points were extracted according to the normal unit introduced based on the outlier detection algorithm, which used the spatial statistics and was an important part of spatial data mining. This approach was applied to the reverse design for the parts of a motorcycle successfully. The experiment shows that this approach can accurately extract the characteristic points and meet the requirement of real-time.