Pinzer, Bernd
Refine
Document Type
Publication reviewed
- peer review (12)
- begutachtet (6)
- nicht begutachtet (2)
Keywords
- EffPro (2)
- Automatic labeling (1)
- Computer vision (1)
- Crystal structure (1)
- Crystallinity (1)
- Gießerei (1)
- Laser sintering (1)
- Online Messung (1)
- Permeation properties (1)
- Physically Based rendering (1)
This study investigates the influence of data distribution and image quality on the performance of fully supervised deep neural networks (DNNs) from the YOLO family for individual plant detection. Several sugar beet datasets (Sugar Beets 2016, Lincoln beet, CropAndWeed, and a newly collected dataset) were used for this analysis. A series of experiments was conducted to examine the effects of model size, random training seeds, dataset size and composition, image quality, and varying agricultural conditions on detection accuracy. The objective of this work is to identify the factors with the greatest impact on model performance in order to enable targeted dataset design and reduce annotation effort. The results show that missing or imprecise annotations have a significantly smaller impact on detection performance than noisy or blurred images. Furthermore, the findings indicate that using approximately 30–40% of the total 26,449 images is sufficient to achieve an optimal balance between detection accuracy and annotation effort.
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.
The utilization of uncrewed aerial vehicles (UAVs) in search and rescue (SAR) operations has become increasingly prevalent because the deployment of UAVs is expected to facilitate a higher degree of operational flexibility while simultaneously reducing costs. Currently, commercially available UAVs can be equipped with low-resolution thermal infrared (IR) cameras with typical resolutions of 640 × 512 pixels, which generally are evaluated manually by the SAR teams during an operation. Automatic person detection in IR images still remains a challenge. The objective of the proposed AIResQ dataset is to significantly enhance the performance of object detectors in the IR domain, employed in SAR operations for missing and potentially injured persons. AIResQ comprises 9,788 IR images with a resolution of up to 2048 × 1536 pixels captured from drone perspectives with a handheld camera under
varying weather conditions and in different terrains. Additionally, AIResQ displays persons in atypical poses. In order to test new object detectors in the context of SAR, we established a benchmark dataset stemming from exercises with real drone flights together with SAR organizations.
Die individuelle Markierung von Gussteilen ermöglicht die direkte Zuordnung von Prozesseinflussgrößen aus Teilprozessen wie Formen, Schmelzen und Gießen zu beliebigen Qualitätskenngrößen. Sie bietet damit den Schlüssel zu einer datenbasierten Analyse der Wirkzusammenhänge zwischen Prozess und technischen, ökologischen und ökonomischen Eigenschaften eines Gussteils sowie einer darauf basierenden KI-gestützten Prozesskontrolle.
In an industrial context, AI-based methods are becoming increasingly important in the optical systems used for identification, inspection and classification. The reasons for this are that AI-based image processing algorithms are easy to use on the operator side and often achieve superior results. E.g. in complex classification tasks. In the sand cast industry, the complexity in optical inspection of cast parts is connected with strong variations in the local surface topography and in the global object geometry change. Despite the great potential of AI-based methods, application is often hindered by the immense effort involved in acquiring a suitable training dataset. This refers not only to the acquisition of the required number of images but also to the tedious labelling. In this work, we investigate the capabilities and limits of synthetic training data on an AI-based optical scanner used to identify and track cast parts. The optical scanner is capable of detecting and classifying a codification specifically designed for the casting industry. By reading the code, the scanner can deduce the specific number of the cast part. For synthetic image generation, we use physically based rendering, which has advantage of full control over all rendering parameters. This allows for both a systematic investigation of the importance of the parameters and, an automatic labelling process of the training datasets. Our results show that, in particular, a detailed geometric modelling of the local surface topography and global object geometry of the pins have a positive influence on the recognition rate of the neural network. With that accuracy rates up to 56 % are achieved using synthetic training datasets, only.
Punching is a process that is sensitive to a multitude of parameters. The estimation of part and punch quality is often based on expert knowledge and trial-and-error methods, mostly carried out as a separate offline process analysis. In a previous study, we developed an optical inline monitoring system with subsequent image processing which showed promising results in terms of capturing every manufactured part, but was limited by slow image processing. Here, we present a more efficient image processing technique based on neural networks. For our approach, we manually identify the burnish parts in images based on criteria established via an expert survey in order to generate a training dataset. We then employ a combination of region-based and boundary-based losses to optimize the neural network towards a segmentation of the burnish surface which allows for an accurate measurement of the burnish height. The hyperparameter optimization is based on custom evaluation metrics that reflect the requirements of the burnish surface identification problem as well. After comparing different neural network architectures, we focus on optimizing the backbone of the UNet++ structure for our task. The promising results demonstrate that neural networks are indeed capable of an inline segmentation that can be used for measuring the burnish surface of punching parts.
Zunächst wird die aktuelle Situation der Schnittflächenkenngrößenermittlung innerhalb des Stanzprozesses in der Industrie beleuchtet und den steigenden Qualitätsanforderungen sowie anwendungsindividualisierbaren Produkten/Prozessen gegenübergestellt.
Auf dieser Basis wird ein optisches Inline-System entwickelt. Dieses System ermöglicht die Aufzeichnung der Schnittflächenqualität in einer Prozessgeschwindigkeit von 250 Hub/min mittels Bildaufnahme. Zusätzlich werden über einen integrierten Triangulationssensor dreidimensionale Schnitte der Schnittkante aufgenommen. Eine entwickelte Bildverarbeitung ermittelt aus den Daten der Bildaufnahme automatisiert und reproduzierbar die Glattschnitthöhe. Die Ermittlung weiterer Schnittflächenkenngrößen ist denkbar, wobei dreidimensionale Daten aus dem Triangulationssensor entstehen. Zusätzlich wird die erreichte Genauigkeit der Systems validiert. Abschließend wird ein Ausblick gegeben, welcher im Besonderen dieses System als Grundlage für die Öffnung der Steuerung des Stanzprozesses mittels intelligenter Methoden sieht. Dies wird an einem Praxisbeispiel unterstrichten.
A novel tracking system for the iron foundry field based on deep convolutional neural networks
(2022)
In modern manufacturing the ability of retracing produced components is crucial for quality management and process optimization. Tracking is essential, especially for analyzing the influence of the production parameters on the final quality of the castings. In the iron foundry industry, common marking methods, such as a datamatrix code, cannot be used due to harsh environmental conditions and the rough surface of the cast parts. This work presents a new coding and reading system that guarantees unique marking in the casting process.The coding is built up over several beveled pins and is read out using an optical 2D handheld scanner. With a deep convolutional neural network approach of object detection and classification, a stable image processing algorithm is presented. With a first prototype a reading accuracy of 99.86% for each pin was achieved with an average scanning time of 0.43 s. The presented code is compatible with existing foundry processes, while the handheld scanner is intuitive and reliable. This allows immediate benefits for process optimization.
