Abstract: Aiming at the reverse problem of model by using Principal Component Analysis (PCA), a modification method was proposed. The transformed model was worked out by PCA. Through introducing the relative vector between model centroid and geometric center of the model, the reverse problem whether existed in transformational model was determined, and the correct direction was calculated by rotating axis. The experiment software was developed for matching the industrial Computed Tomography (CT) reconstruction model with the Computer Aided Design (CAD) model roughly, and the contrast tests were did for improvement. The result showed that the new method could effectively solve the reverse problem of model which generated in registration process of PCA.

Key words: error analysis, rough registration, principal component analysis, industrial computer tomography, computer aided design

摘要： 针对使用主成分分析法配准时产生的模型反向问题提出了修正方法。首先使用主成分分析法配准得到变换后的模型，再引入点云模型形心与质心之间的位置关系矢量，判断变换后的模型是否存在反向问题，并通过计算旋转轴线来修正模型方向。开发了工业CT三维重建模型与计算机辅助设计模型粗配准实验软件，对改进前后的方法进行了对照实验，表明所提方法可以有效解决使用主成分分析法配准时产生的模型反向问题。

关键词: 误差分析, 粗配准, 主成分分析, 工业CT, 计算机辅助设计