Dr. Lennart Hinz Profile

Dr. Lennart Hinz

at Leibniz Univ Hannover

SPIE Involvement:

Author

Publications (10)

Proceedings Article | 2 April 2024 Presentation + Paper

Medical instrument detection with synthetically generated data

L. Wiese, L. Hinz, E. Reithmeier

Proceedings Volume 12931, 1293109 (2024) https://doi.org/10.1117/12.3005798

KEYWORDS: Equipment, Education and training, 3D modeling, Neural networks, Data modeling, Image segmentation, Surgery, Mathematical optimization, Instrument modeling, Image processing

Read Abstract +

Proceedings Article | 15 August 2023 Presentation + Paper

Identification and correction of magnification factor deviations of a telecentric fringe projection system

Pascal Kern, Lennart Hinz, Markus Kästner, Eduard Reithmeier

Proceedings Volume 12618, 126180I (2023) https://doi.org/10.1117/12.2673617

KEYWORDS: Calibration, 3D metrology, Imaging systems, 3D acquisition

Read Abstract +

Proceedings Article | 11 August 2023 Presentation + Paper

Improved segmentation of damages on high-resolution coating images using CNN-based ensemble learning

Kolja Hedrich, Lennart Hinz, Eduard Reithmeier

Proceedings Volume 12623, 126230A (2023) https://doi.org/10.1117/12.2673821

KEYWORDS: Image segmentation, Coating, Inspection, Video, Image processing, RGB color model, Endoscopes, Binary data, Video processing

Read Abstract +

Proceedings Article | 3 October 2022 Presentation + Paper

GPU implementation of FSR simulations: performance improvements and limitations

Nils Melchert, Jérôme Degallaix, Lennart Hinz, Eduard Reithmeier

Proceedings Volume 12225, 122250E (2022) https://doi.org/10.1117/12.2633434

KEYWORDS: Electromagnetism, Radio propagation, Computer simulations, Mirrors, Wave propagation, MATLAB, Phase shifts, Visualization, Resonators, Fourier optics

Read Abstract +

Proceedings Article | 3 October 2022 Presentation + Paper

Pose estimation of optical resonators using convolutional neural networks in a simulation environment

Nils Melchert, Kolja Hedrich, Leon Wiese, Lennart Hinz, Eduard Reithmeier

Proceedings Volume 12222, 122220D (2022) https://doi.org/10.1117/12.2633416

KEYWORDS: Resonators, Optical resonators, Fabry–Perot interferometers, Monte Carlo methods, Convolutional neural networks, Network architectures, Optical design, Mirrors, Computer programming, Neural networks

Read Abstract +

Proceedings Article | 3 March 2022 Presentation + Paper

Damage segmentation using small convolutional neuronal networks and adversarial training methods on low-quality RGB video data

Kolja Hedrich, Lennart Hinz, Eduard Reithmeier

Proceedings Volume 12019, 1201902 (2022) https://doi.org/10.1117/12.2610123

KEYWORDS: Image segmentation, Coating, Inspection, Endoscopes, Video compression, Convolutional neural networks, Computer vision technology

Read Abstract +

Within the aviation industry, considerable interest exists in minimizing possible maintenance expenses. In particular, the examination of critical components such as aircraft engines is of significant relevance. Currently, many inspection processes are still performed manually using hand-held endoscopes to detect coating damages in confined spaces and therefore require a high level of individual expertise. Particularly due to the often poorly illuminated video data, these manual inspections are susceptible to uncertainties. This motivates an automated defect detection to provide defined and comparable results and also enable significant cost savings. For such a hand-held application with video data of poor quality, small and fast Convolutional Neural Networks (CNNs) for the segmentation of coating damages are suitable and further examined in this work. Due to high efforts required in image annotation and a significant lack of broadly divergent image data (domain gap), only few expressive annotated images are available. This necessitates extensive training methods to utilize unsupervised domains and further exploit the sparsely annotated data. We propose novel training methods, which implement Generative Adversarial Networks (GAN) to improve the training of segmentation networks by optimizing weights and generating synthetic annotated RGB image data for further training procedures. For this, small individual encoder and decoder structures are designed to resemble the implemented structures of the GANs. This enables an exchange of weights and optimizer states from the GANs to the segmentation networks, which improves both convergence certainty and accuracy in training. The usage of unsupervised domains in training with the GANs leads to a better generalization of the networks and tackles the challenges caused by the domain gap. Furthermore, a test series is presented that demonstrates the impact of these methods compared to standard supervised training and transfer learning methods based on common datasets. Finally, the developed CNNs are compared to larger state-of-the-art segmentation networks in terms of feed-forward computational time, accuracy and training duration.