Recognition of orthogonally oriented micromarkings using curvature detectors

The problem of recognizing micro-dimensional markings used to identify products and certain areas of the surface is considered. The marking used for these purposes was a pattern of separate orthogonally oriented segments applied using contact nanolithography. Crosses and rectangles are considered as such patterns in the article. A scanning probe microscope was a tool for obtaining images of micromarkings and their application. The peculiarities of microlabeling images obtained by a probe microscope are analyzed. It was found that the recognition of markings is hindered by the relief of the sample surface and the markings skeleton distortions (ruptures) caused by it, as well as image noises inherent in probe microscopy (drop-out lines, brightness jumps, reduced contrast). It was proposed to use two different surface curvature detectors to recognize images with such distortions. The curvature in these detectors was estimated by inscribing the geometric shapes of a circle and a sphere into the relief. The inscribing and evaluation of the radius took place at each point of the surface image raster. This ensures sensitivity to the local centers of curvature associated with the markings skeleton. It is shown that the use of two different curvature detectors, characterized by different sensitivity to the central homogeneous part of the segments and the places of the contact or intersection of the segments, makes it possible to increase the accuracy of the skeletons location and the reliability of recognition. The method of detectors application and the criteria for the presence of markings on the image are described. From the set of local curvature maxima localized by both detectors, adjacent maxima are selected, from which isolated skeletons are formed. The criterion for marking recognition is the presence of intersecting or orthogonally oriented skeletons of a certain length. The results of testing the technique on real images are presented. The results showed that the use of two curvature detectors made it possible to eliminate breaks in the marking skeletons and increase the reliability of recognition.