复杂曲面测量规划及定位技术研究-仪器科学与技术专业论文.docx

PAGE PAGE VI process is carried out by a data registration algorithm after the data points are captured. In the algorithm, registration transformation matrix is decomposed and constraints are applied on some degrees of freedom and a more accurate result is achieved. Experiments on localizing an axial-flow impeller are conducted and show that the average error of the blade profile is reduced. 4.A new shape registration technique considering the regional difference in precision is proposed. The shape of a mechanical part is composed of a number of specific feature regions, each with their own precision level. If the shape registration between the scanned data and the CAD model is performed when ignoring the precision difference, an inaccurate alignment result will be produced due to the deviation of low fidelity regions.A new registration method taking into consideration of the region precision is proposed. The procedure is divided into several region-to-region matching steps. Firstly, the regions with high precision are matched. This matching transformation brings a set of standard orthogonal bases for the transformation space, and at the same time it gives registration constraints on some principal directions. Then the low precision regions are used to do a new matching. By projecting the new transformation on the previous standard orthogonal bases and setting some terms to zero, the model transformation is restricted and has no movement on the principal directions determined by the high precision regions. The method produces high matching accuracy on high precision regions and low matching accuracy on low precision regions, which is suitable for practical utility in engineering. The method is verified by experiments on a real impeller. KEY WORDS: complex curved surfaces, unified coordinate system measurement method, stepwise refinement localization, difference of regionalprecision,shaperegistration 目录 摘要 III ABSTRACT IV HYPERLINK \l _bookmark0 目录 VI HYPERLI