Search

The automated visual surface inspection planning is an important part of the quality assurance in automated custom product manufacturing. Visual surface inspection planning tackles image acquisition design and defect detection. Both tasks greatly benefit from the utilization of realistic and automated image synthesis of the inspected object. The realism of synthesized images greatly depends on object material, whose properties are largely influenced by texture. In this work, we focus on parametric texture synthesis and its application for visual surface inspection planning. We start by analyzing texture present on physical samples and introduce the requirements for texture synthesis models in visual surface inspection. Based on observation and surface characterization standards we present a model capable of reproducing texture on physical samples. This approach is generalized and further models are presented with respect to requirements. Finally, we highlight the importance of surface texture from the visual inspection planning perspective.

Mithilfe von Computer Vision, Computergrafik, maschinellem Lernen und Robotik wird am Fraunhofer-Institut für Techno- und Wirtschaftsmathematik ITWM ein virtuelles Framework konzipiert, welches das iterative Design eines Inspektionssystems unterstützt und somit ein fixes Bildaufnahme-Setup als Ausgangspunkt umgeht.

Currently, thin-film coating adhesion classification is done manually guided by two standards, DIN 4856 and ISO 26443. Both standards provide classification guidance based on set of schematic illustrations. As such the interpretation of these standards is both prone to human error and lacks clear class parametrization, which would be necessary for enabling automatic adhesion classification at a finer resolution. In this paper we introduce a set of hand-crafted features used for the parametrization of the two thin-film hard coating adhesion classification standards, as well as a pipeline for automatic fine resolution adhesion classification. The developed features resemble key characteristics used by experts in the classification process, describing the properties of either delamination or cracking within the sample. Additionally, we explain the necessity for a revised approach to thin-film hard coating adhesion classification, specific requirements that the standard class parametrization process has to fulfill and the steps we have taken in tackling these challenges. The set of proposed features was used to train two different regression models whose results were further used for extensive evaluation of both feature applicability and model performance. The analyses show that the features are suitable for parametrizing standard's classes and highlight that ISO standard requires both delamination and cracking features, whereas DIN requires only delamination. This work introduces a novel method for fine scale thin-film hard coating adhesion classification which has not been available so far. Furthermore, the proposed features introduce a solid base for the definition of a future adhesion classification standard providing class parametrization necessary for automatic fine scale adhesion evaluation.

Viewpoint estimation is an important aspect of surface inspection and planning. Typically viewpoint estimation has been done only with the 3D model and not with the actual object. This, therefore, limits the flexibility of using the actual object in inspection planning. In this work, we present a novel pipeline that can efficiently estimate viewpoints from live camera images. The pipeline can be used for different sized industrial objects to efficiently estimate the viewpoint. The achieved results are real-time and the method is easily generalizable to different objects. The presented method is based on a 3D model of the object, which is self-supervised and requires no manual data annotation or real images to be used as an input. The complete solution, together with documentation and examples is available in the public domain for testing. https://gitlab.itwm.fraunhofer.de/dutta/real-time-pose-est/.

Die Haftung von Hartstoffschichten ist ein wesentliches Qualitätskriterium für die Funktions­sicherheit von beschichteten Bauteilen. Bei der visuellen Auswertung des Rockwell­ Schichthaftungstests sind die Ergebnisse teils subjektiv und nicht genau genug. Ein neues, auf maschinellem Lernen basierendes Verfahren ermöglicht es, die Bewertung zu automatisieren.

Surface inspection development requires large amounts of image data representing the inspected product surface. The image data should contain both the ideal surface and the defective surface that can appear during production. Although image synthesis comes as a natural solution to this problem, its application is not straightforward in automated surface inspection environments. The reason for that is a lot of manual work that should be done for creating defects, simulating the inspection environment, and setting up the acquisition system for validation of simulation. To address these issues, we present a novel pipeline that automatizes surface defect creation, provides realistic rendering in a predefined inspection environment setup, and an acquisition system that enables comparison with the real images. The pipeline creates geometry-imprinted defects which combined with physically based rendering methods enable realistic light response for different light and camera positions during image synthesis. Finally, synthesized images can becompared with the real image taken in the same setup enabling verification. Also, synthesized images enable the visualization of visible surfaces and defects for a given inspection plan.

The goal of visual surface inspection is to analyze an object’s surface and detect defects by looking at it from different angles. Developments over the past years have made it possible to partially automate this process. Inspection systems use robots to move cameras and obtain pictures that are evaluated by image processing algorithms. Setting up these systems or adapting them to new models is primarily done manually. A key challenge is to define camera viewpoints from which the images are taken. The number of viewpoints should be as low as possible while still guaranteeing an inspection of the desired quality. System engineers define and evaluate configurations that are improved based on a time-consuming trial-and-error process leading to a sufficient, but not necessarily optimal, configuration. With the availability of 3D surface models defined by triangular meshes, this step can be done virtually. This paper presents a new scalable approach to determine a small number of well-placed camera viewpoints for optical surface inspection planning. The initial model is approximated by B-spline surfaces. A set of geometric feature functionals is defined and used for an adaptive, non-uniform surface sampling that is sparse in geometrically low-complexity areas and dense in regions of higher complexity. The presented approach is applicable to solid objects with a given 3D surface model. It makes camera viewpoint generation independent of the resolution of the triangle mesh, and it improves previous results considering number of viewpoints and their relevance.

Efficiency of a visual inspection system is greatly influenced by the hardware configuration in terms of both arrangement and sensing parameters. Currently, the configuration is still determined manually. The available automated solutions are not yet ready to be integrated into everyday industrial use due to restrictions in applicability. Therefore, this paper offers a holistic approach to development and integration of an inspection planning pipeline which is both modular and application flexible. First, a clear distinction between the automated and semiautomated planning is proposed. Further, a set of requirements is established, ensuring system flexibility. The requirements are independent of the sensing technology and application. Finally, following the requirements, we implemented a semi-automated planning pipeline intended for visual inspection of rigid objects. For the first time, the pipeline introduces a flexibility in terms of object geometry, as well as interactive adaptation of planning results.

State of the art surface inspection planning requires an expert approach with a lot of trial and error. Because of the lack of available aiding tools, an engineer building inspection systems must rely heavily on his or her experience. In this work we proposed an interactive 3D visualization tool to help an engineer determine good viewpoints. It can be used both as a standalone tool for manual viewpoint placement or as an interface to the inspection planning algorithms giving the engineer a possibility to evaluate and modify automatically generated plans.

The failure of propulsion components of ships and ferries can lead to serious environmental and economic damage or even the loss of lives. For ultrasonic inspection of such large components we employ mechanized scanning and defect reconstruction using the Synthetic Aperture Focusing Technique (SAFT). We report on results obtained in view of the detection of defects with different inspection techniques. Also, we address the issue of Probability of Detection by reporting results obtained in POD and MAPOD-studies (Model-Assisted POD) using experimental and simulated data. Finally, we show recent results of surface and sub-surface inspection using optical and eddy current techniques.

Beruhrungslos arbeitende Verfahren der industriellen Bildverarbeitung gewinnen in allen Industriebranchen an Bedeutung. Die automatisierte Oberflachenprufung ist dem menschlichen Auge weit uberlegen. Sie hilft den Herstellern, die hohe Qualitat ihrer Produkte sicherzustellen und gleichzeitig Kosten zu sparen.