Quantifying Surface Texture with Deep Learning on Laser Treated Surfaces

In industry, laser treatment is a common approach to improve the surface characteristics of manufactured parts. The quality of laser treated surfaces depends on many process parameters, for example laser type, laser power or material, and it is essential to monitor the process to ensure the quality of the produced surfaces in an automatic way. Therefore, surface texture classification is indispensable in the production cycle. Traditional tactile-based methods for measurement of surface topography are limited in their application, since they either only measure a surface profile instead of the areal laser structure, or they typically have a long measuring time for areal measurements. Here, an optical areal measurement and classification solution is presented to automatically classify laser treated surfaces. The proposed solution offers advantages in terms of speed and automation by using a measuring instrument based on focus variation in combination with artificial intelligence for surface texture classification. This chapter presents a comprehensive case study on the application of deep learning and convolutional neural networks in the field of optical metrology for classification of laser treated surfaces. In addition, the chapter discusses various machine learning methods and their application for classification at an abstract level to highlight the differences between traditional methods and the deep learning method applied here. The presented applications and results demonstrate the effectiveness and potential of the proposed measurement solution for optical measurement classification tasks.