A Scale Independent Algorithm for the Detection of Fault Indicating Structures in Range Images

Abstract

Reliable real-time surface inspection of extended objects is needed in several industrial applications. At the one hand this inspection should be performed with the highest possible resolution to achieve the recognition of fault indicating structures with lateral dimensions down to the range of a few microns. On the other hand the object should be measured during an industrial production process “on the fly” and - if possible - at once. In the case of a car body this would mean that the inspection system had to perform a robust measurement with a ratio of > 10^-6 between depth resolution and lateral extension. This ratio is at least one order beyond the ones which existing technologies offer.

The concept of scaled topometry consists of arranging different optical measurement techniques with overlapping ranges of resolution systematically in order to receive characteristic surface information with the required resolution. In such a surface inspection system, measurements on different scales of resolution have to be combined by a discrimination algorithm which should be sensitive on faults independent on the scale of resolution. Starting from a global measurement with a resolution not high enough for a sufficient description certain critical areas have to be detected in which a local measurement with higher resolution has to be performed.

We have refined our decision algorithm based on the concepts of fractal geometry. This algorithm extracts the order and homogeneity of fractality from the surface images and selects local deviations using different forms of filtering (Fourier and Wavelet). It is now able to use input parameters depending on which the detection is performed. The output determines the next inspection system and is used as input for the next step of measurement.