High-resolution defect detection and noise reduction using wavelet methods for surface measurement

Nowadays the demands on quality control are constantly increasing, hence an important step is a completely automated control with well defined risks. Very promising solutions are optical 3D, in addition to surface measurement. Here, an algorithm is presented to separate local defects from the surface and the noise (measurement error and surface roughness) with a given manufacturer's risk for piecewise smooth surfaces. The algorithm consists of a feature enhancement by means of special wavelets and thresholding and interpolation schemes to recover a defect- and noise-free surface and subsequently the extension and shape of the defects in all directions with reduced random errors. The limits of the algorithm such as accuracy, sensitivity, maximum cover rate of the surface with defects and rotation and translation invariance are shown theoretically and by numerical simulations. Experimentally, nanoindents are measured by means of confocal microscopy, and a reduction of the random errors by one order of magnitude is observed. Furthermore, a ceramic plate is measured by means of fringe projection and features are detected which are much smaller than the noise. Finally, a white light measurement is evaluated to demonstrate the scale and instrument independence of the method.