Research On The Multi-scale Algorithm Used In Measurement Of Roundness Error

Scale hides deeply in the attribute of time and space as well as the other closelyrelated attributes of all objective things. Scale is an easily-confused and easilymisunderstandingconcept, possessing different meanings to different subjects andapplication backgrounds. Scale means space and time unit used in studying someobject or phenomenon, and also means the range and occurring frequency involved inspace of the object or phenomenon. The former is to define scale from researcher view,and the latter is to determine scale based on the process or phenomenon studied.At present, the methods used in scale research mainly include graphic method,regression analysis, Variogram analysis, spatial autocorrelation analysis, spectrumanalysis, fractal analysis, scale variance analysis, wavelet analysis and so on. Waveletanalysis is developed on the basis of Fourier analysis,which resolves problems as thesize of window problem proposed by Gaber and transform window problem proposedby Fourier as well as shape fixed problem. It is a time-frequency localization analysisof which the window size is fixed but the shape is changed. With the ever-changing ofthe scale from crude to fine, Wavelet transform can gradually focus on any details ofthe analysis object, fully displaying all the changes of the object. Therefore, wavelettransform is denoted as"mathematical microscope"in scientific community.In this paper, on the basis of non-contact measurement to roundness error, a kindof data processing method naming multi-scale kalman filter is proposed, the mainresearch content and related conclusions is as follow:(1) Multi-scale kalman filter method is derived based on wavelet analysis,which is used to filter one-dimensional noisy simulation signal. It has moreeffectiveness than wavelet filter method.(2) On the basis of roundness error online inspection technology using lasertriangulation, a kind of data processing method naming multi-scale kalman filter isproposed. Data is collected by the data acquisition system of high-accuracy laser CCDsensor, the data treated by multi-scale kalman filter is used to calculated form and position error. The calculation shows that roundness error obtained by multi-scalekalman filter is closer to the value obtained by roundness instrument, and the signalto-noise ratio of the signal is between three and four.(3) Multi-scale kalman filter method for two-dimensional image signal isderived, and the simulation is conducted. The result shows that multi-scale kalmanfilter method is not only better then other methods in indicator, but also in visual effect,and the practicality is illustrated.(4) Based on the roundness testing technology using computer visiontechnology, multi-scale kalman filter method is used to filter and denoise the graphcollected, in addition, image enhancing, edge detection of pixel level and sub-pixelinterpolation are made, finally appraisal calculation of roundness error is conducted tothe diameter of cylindrical axis obtained by processing.