Refine
Document Type
Language
- English (7)
Publication reviewed
- peer review (5)
- begutachtet (2)
Keywords
- EffPro (2)
- AI-based quality inspection (1)
- Automatic labeling (1)
- Computer vision (1)
- Physically Based rendering (1)
- Synthetic data generation (1)
- Synthetic images (1)
- automated labeling (1)
- physically based rendering (1)
- synthetic data (1)
We present a highly miniaturized endoscopic point distance sensor based on a spatial confocal measurement principle. The sensor uses a new technique called spatial confocal point distance measurement. A special feature of the proposed sensor design is the high degree of miniaturization through femtosecond direct laser writing and the use of optical fiber bundles, which enable an endoscopic application. We show the complete sensor measurement principle, sensor head design, experimental setup, and experimental results.
Increasing miniaturization requires improved and highly miniaturized optical 3D metrology systems. In this paper a basic measurement principle and a proposed optical design of a highly miniaturized endoscopic spatial confocal point distance sensor are presented. The sensor uses a, to our knowledge new technique called spatial confocal point distance measurement. A special feature of the proposed sensor design is the high degree of miniaturization, through femtosecond direct laser writing and the use of optical fiber bundles, which enable an endoscopic application.
Chromatic confocal metrology is a widely established optical metrology technique, that allows for non-contact high-speed three-dimensional surface profiling without the need of mechanical depth scanning. However current methods are limited by the use of some sort of surface scanning method with mechanically moving parts. Furthermore the setups involve a spectrometer setup, either through prisms, gratings or multi-spectral cameras. This drastically limits the simultaneously measureable positions in lateral direction, as the spectrometer setup
will utilize one spatial dimension for the wavelength domain. We present a novel method for chromatic confocal metrology, that enables high-speed and high resolution one-shot aerial surface metrology. This method is scalable with respect to measurement range in axial as well as in lateral direction and in the number of measurement points that can be measured simultaneously. After deriving the theoretical basis of the approach a virtual optical design with a FOV of 10mm by 10mm and a depth range of 1.5mm with roughly 1000 measurement points, based mainly on off-the-shelf components will be presented. This virtual system design was used to perform various simulations and explain the design process and considerations as well as the expected system response of the proposed system.
Purposefully induced axial chromatic aberration is the core of the chromatic confocal metrology technique.Through the resulting generation of separated focal planes for each wavelength of a broadband light sourcea measurement volume is created and a three-dimensional reconstruction of the topography of technical and biomedical surfaces and layers can be performed. Based on the chromatic confocal metrology technique various metrology sensors and measurement systems have been developed, with high axial and lateral resolution, accuracy and precision. For a significant increase in measurement points, that can be measured simultaneously and the resulting reduction in measurement time, a chromatic confocal method utilizing a micro-lens array in combination with a improved spectral peak detection, has been developed. Through a single image acquisition, the object topography can be measured for multiple points simultaneously and therefore mitigating the need for axial aswell as lateral scanning of the object. For this reason in-situ applications have become a viable domain. First preliminary results of testing a laboratory setup of the proposed system design are presented.
Chromatic confocal metrology suffers from a limitation in the number of measurement points that can be measured simultaneously in a single frame acquisition. We propose chromatic confocal areal metrology (ChromaCAM), in which the surface height for each point in a 2D grid of measurement spots, generated by a rectangular micro-lens array, is parallely analyzed through the utilization of a pinhole multiplexer unit, an analog optical analysis unit, and postprocessing algorithms. An experiment shows the viability of the simultaneous acquisition of multiple measurement points and the advantages over exisiting areal chromatic confocal approaches. Compared with conventional chromatic confocal metrology, the increase in the acquisition rate is significant and enables one-shot measurements.
Synthetic Data Generation for AI-Based Quality Inspection of Laser Welds in Lithium-Ion Batteries
(2025)
Manufacturing companies are increasingly confronted with critical challenges such as a shortage of skilled labor, rising production costs, and ever-stricter quality requirements. These challenges become particularly acute when defect types exhibit high visual variance, making consistent and accurate inspection difficult. Traditionally, visual inspection of high variance errors is performed manually by human operators—a process that is both costly and prone to errors. Consequently, there is a growing interest in replacing human inspection with AI-based visual quality control systems. However, the adoption of such systems is often hindered by limited access to training data, labor-intensive labeling processes, or the absence of real production data during early development stages. To address these challenges, this paper presents a methodology for training AI models using synthetically generated image data. The synthetic images are created using Physically Based Rendering, which enables precise control over rendering parameters and facilitates automated labeling. This approach allows for a systematic analysis of parameter importance and bypasses the need for large real training datasets. As a case study, the focus is on the inspection of laser welds in battery connectors for fully electric vehicles—a particularly demanding application due to the criticality of each weld. The results demonstrates the effectiveness of synthetic data in training robust AI models, thereby providing a scalable and efficient alternative to traditional data acquisition and labeling methods. The trained binary classifier reaches a precision of 0.94 with a recall of 0.98 solely trained on synthetic data and tested on real image data.
In industrial applications, AI-based methods are gaining increasing significance in optical systems for tasks such as identification, inspection, and classification. The appeal of these methods lies in their operator-friendly implementation and often superior performance, particularly in complex classification tasks. However, despite the substantial potential of AI-based methods, their application is frequently constrained by the significant resources needed to acquire an appropriate training dataset. For instance, in the sand casting industry, the complexity of optical inspection for cast parts is compounded by considerable variation in local surface topography and global object geometry. This challenge pertains not only to the volume of images required but also to the extensive process of accurate labelling. In this work, we investigate the effectiveness and limitations of synthetic training data for an AI-based optical code reader, designed for the identification and tracking of cast parts. This scanner is capable of detecting and classifying a unique code, called Cast Code, developed specifically for the casting industry, enabling the identification of individual cast part numbers. For synthetic image generation, we utilize physically based rendering, which allows comprehensive control over all rendering parameters. This approach facilitates both a systematic investigation of parameter relevance and an automated labelling process for the training datasets. Our results indicate that detailed geometric modelling of both, the local surface topography and the global object geometry of the pins has a notable positive impact on the neural network’s recognition accuracy, achieving accuracy rates of up to 84 %, on real pin images, using synthetic training datasets, only.
