Reducing uncertainty for laser triangulation measurements on moving samples

Abstract

Permanent quality control directly in the production line is of high interest in today’s manufacturing facilities to increase throughput and reduce production rejects. As optical sensor systems enable contactless, fast, and precise measurements of surface topologies, they are frequently used for in-line measurements on moving samples. Edges are often defining elements of spatial features on 3D structures, e.g. two edges may define the critical width of a trench structure. Thus, to detect deviations from the feature’s dimensional specifications, the lateral position of its edges must be determined accurately. As optical sensors have finite exposure times, the motion of the sample during the exposure induces lateral uncertainty. This thesis aims to reduce the lateral position uncertainty of edge features in laser triangulation measurements on moving samples. An experimental laser triangulation sensor is developed, which allows access and adjustment of relevant parameters such as exposure time and laser intensity and the readout of its imaging sensor’s raw pixel data. Combining 1D Gaussian most-likelihood estimation, matched filtering, and Gaussian-mixture-model fitting, the intensity distribution on the imaging sensor is described and analyzed. Using a model of the laser intensity distribution on the moving sample, the lateral position of edge features is estimated by comparing the peak power ratio on the imaging sensor and the power distribution on the sample. Experimental performance evaluation shows a significant reduction of the lateral edge position uncertainty. The performance of the proposed method is determined by comparing statically and dynamically measured feature widths of a 3D printed sample.The results show a significant reduction of the mean absolute error in the corrected measurements of more than 60% compared to uncorrected measurements. Moreover, features that are missed due to motion blur in the uncorrected measurements are detected with high lateral accuracy by the proposed method.