METHODS AND ALGORITHMS OF DATA STITCHING AND FUSION FOR MULTI-SCALE MEASUREMENT OF ULTRA-PRECISION FREEFORM SURFACES

摘要

Ultra-precision freeform surfaces with submicrometer form accuracy and surface finish in nanometric range are now widely adopted in opto-mechatronics applications due to their capabilities of design freedom and special functions, as well as compact configuration for the products. However, it also brings a lot of challenges for the measurement of freeform surfaces due to the geometrical complexity and variety, especially for the large-sized and multi-scale freeform surfaces, which requires a well balance between high precision and efficiency. Currently, there is still lack of a complete solution for measurement of ultra-precision freeform surfaces by one type of measuring instrument, regarding to the multi-scale measurement, different sample materials, etc. Many existing precision measurement instruments only possess limited measuring range at one single measurement. This is particularly true for some freeform surfaces with high slope. Besides, a holistic measurement of freeform surfaces also demands the data measuring from different sensors, or multi-sensor measurement. Freeform data stitching and fusion technology not only provides an enabling solution for multi-scale and multi-sensor measurement of freeform surfaces, but also enhances the measuring ability of some high precision measurement instruments. Data stitching and fusion is emerging technology since 1990s, and it is previously used for target tracking, automated identification of targets, and limited automated reasoning applications [1]. The technology is then adopted in reverse engineering and precision metrology [2, 3]. Generally, a procedure for data fusion includes pre-processing, data registration, data fusion and post processing, among which data registration is critical for data stitching and fusion. Due to the geometry complexity and variety, as well as a lack of common features, of freeform surfaces, it is still very difficult to register one freeform surface to another one. Open literature review shows that, data registration can be realized by feature based or surface descriptions based approaches, and the registration process generally includes coarse and fine registration [3]. Some research work has been found for data stitching of aspherical surfaces [5, 6]. For example, subaperture stitching method is adopted to measure aspherical surfaces based on least square-fitting [7]. Iterative Closet Point (ICP) method is often used for coordinates correspondence searching. However, they are susceptible to data noise and outliers involved in the measured data. There is still little research found for data stitching and fusion of ultra-precision freeform surfaces with sub-micrometer form.