Probabilistic Absolute Position Sensor Based on Objective Laser Speckles

Abstract

This paper presents a vision-based probabilistic absolute position sensor, which is able to operate on spacious technical surfaces without the modification of the surface. Instead of using artificial surface markings, the surface structure itself is used to generate objective laser speckle patterns (OLSPs), which are observable on most technical surfaces. This allows the extraction of features on surfaces, which might be too smooth for image processing under white light observation. In addition, the uniqueness of the individually observed laser speckle patterns enables a determination of the absolute position even after a sudden power loss of the sensor system. Experiments confirm the applicability of such OLSPs within a feature-based probabilistic framework on a smooth stainless steel rod and compare the results to images taken under white light illumination. The proposed sensor system successfully recovers from unknown initial states and measures the position with a peak-to-peak error of 22.1 $\mu \text{m}$ and an rms error of 5.6 $\mu \text{m}$ over a measurement range of 100 mm.