Forging measurement using passive stereo vision

Abstract

Dimensional inspection during manufacturing of high-temperature forging workpieces is still an unsolved research topic. In this work, an approach to reconstruct the hot forging surface using passive stereovision was proposed. The challenges are the irregular thermal radiation of the forging in the visible spectrum, which is affected by scales, heat haze, and an unstable environment. Unlike the approaches presented in the literature, we use the radiated light, without an external light source. High dynamic resolution images were used to image the forging surface with highly varying brightness. The effect of optical filtering on the contrast of the fine surface structure was investigated and evaluated by spectral analysis of the images, with the best results obtained in the green spectrum. A comparison of feature detectors was performed for the specific surface structure of the hot forging. The best results in terms of density of reconstructed points were obtained using the KAZE method. The final verification of the achievable accuracy was performed in a measurement volume of approx. 500×400×350 mm. An 877 °C hot steel pipe sample was measured with a single point standard deviation error of 0.28 mm, which is an order of magnitude higher than under ideal conditions, but acceptable in comparison with manufacturing tolerances. Further developments could be related to the application of dense image matching, further investigation of interfering effects, and application of the proposed methods in the industrial environment.