Stream-Based Data Filtering for Accelerating Metrological Data Characterization

The main task of engineering surface metrology is to characterize a surface by assessing components such as form, waviness and roughness that correspond to different wavelength segments in the frequency domain, which are often extracted by deploying filtering techniques. The effectiveness of a specific kind of filtering algorithms is jointly determined by their filtering accuracy and computational efficiency. In this paper, a data stream-based programming paradigm is introduced which takes advantage of the programmability and parallel computation capacity of modern graphics process unit (GPU) to execute and accelerate the Gaussian filtering process that is extensively used in surface metrological data processing. In contrast to the results obtained by running MATLAB simulation kit for similar processes, the software framework speeds up the filtering process substantially while yielding satisfying accuracy as that of the corresponding MATLAB program, which proved the practicability and validity of the proposed computation model.