Stochastic Geometry Models for Texture Synthesis of Machined Metallic Surfaces: Sandblasting and Milling

TL;DR

Stochastic texture models for sandblasted and milled surfaces based on topography measurements of such surfaces are developed, which use separate modeling approaches for the two cases as the surface patterns differ significantly.

Abstract

Training defect detection algorithms for visual surface inspection systems requires a large and representative set of training data. Often there is not enough real data available which additionally cannot cover the variety of possible defects. Synthetic data generated by a synthetic visual surface inspection environment can overcome this problem. Therefore, a digital twin of the object is needed, whose micro-scale surface topography is modeled by texture synthesis models. We develop stochastic texture models for sandblasted and milled surfaces based on topography measurements of such surfaces. As the surface patterns differ significantly, we use separate modeling approaches for the two cases. Sandblasted surfaces are modeled by a combination of data-based texture synthesis methods that rely entirely on the measurements. In contrast, the model for milled surfaces is procedural and includes all process-related parameters known from the machine settings.

Figures (25)

• Figure 1: Illustration of the synthetic visual surface inspection system and comparison between acquisition and rendered images of a milled surface. Images are provided by Fraunhofer ITWM.
• Figure 2: Test objects made of aluminum having a base about $5\text{ cm}$ in size and differently machined surfaces: sandblasted (left), parallel milled (middle) and spiral milled (right). The objects were produced by Fraunhofer IOF.
• Figure 3: Measurement of a milled surface provided by Fraunhofer IOF. 2d height image (left) and its representation as a surface in $\mathbb{R}^3$ (right). Imaged region is $3\text{ mm}\times 3\text{ mm}$.
• Figure 4: Measurements of sandblasted surfaces using an air-jet with pressure $2.5$ bar (left) and $6$ bar (right) provided by Fraunhofer IOF. The imaged region is $5.6\text{ mm}\times 7.7\text{ mm}$.
• Figure 5: Output of texture generation methods using the same initial Gaussian white noise image. Image size is $512\times 512$ and pixel spacing $1.75\,\mu$m which corresponds to an imaged region of $0.9\text{ mm}\times0.9\text{ mm}$.