Loading [MathJax]/extensions/tex2jax.js

Noise-robust registration of microscopic height data using convolutional neural networks

Research output: Chapter in book/report/conference proceeding › Conference contribution › Research

Authors

Research Organisations

Details

Original languageEnglish
Title of host publicationSPIE Future Sensing Technologies 2023
EditorsOsamu Matoba, Joseph A. Shaw, Christopher R. Valenta
PublisherSPIE
ISBN (electronic)9781510657229
Publication statusPublished - 22 May 2023
EventSPIE future sensing Technologies - Yokohama, Japan
Duration: 18 Apr 2023 → 21 Apr 2023

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume12327
ISSN (Print)0277-786X
ISSN (electronic)1996-756X

Abstract

In this work, a deep convolutional neural network is proposed to improve the registration of microtopographic data. For this purpose, different mechanical surfaces were optically measured using a confocal laser scanning microscope. A wide range of surfaces with different materials, processing methods, and topographic properties, such as isotropy and anisotropy or stochastic and deterministic features, are included. Training and testing datasets with known homographies are generated from these measurements by cropping a fixed and moving image patch from each topography and then randomly perturbing the latter. A pseudo-siamese network architecture based on the VGG Net is then used to predict these homographies. The network is trained with a supervised learning approach where the Euclidean distance between the predicted and the ground truth gives the loss function. The 4-point homography parameterization is used to improve the loss convergence. Furthermore, different amounts of image noise are added to enhance the prediction’s robustness and prevent overfitting. The effectiveness of the proposed method is evaluated through different experiments. First, the network performance is compared to intensity-based and feature-based conventional registration algorithms regarding the resulting error, the noise-robustness, and the processing speed. In addition, images from the Microsoft Common Objects in Context (COCO) dataset are used to verify the network’s generalization capability to new image types and contents. The results show that the learning-based approach offers much higher robustness regarding image noise and a much lower processing time. In contrast, conventional algorithms have a smaller registration error without image noise.

Keywords

    Confocal Laser Scanning Microscopy, Convolutional Neural Network, Homography estimation, Image registration, Machine learning, Surface metrology

ASJC Scopus subject areas

Cite this

Noise-robust registration of microscopic height data using convolutional neural networks. / Siemens, Stefan; Kästner, Markus; Reithmeier, Eduard.
SPIE Future Sensing Technologies 2023. ed. / Osamu Matoba; Joseph A. Shaw; Christopher R. Valenta. SPIE, 2023. 123270D (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 12327).

Research output: Chapter in book/report/conference proceeding › Conference contribution › Research

Siemens, S, Kästner, M & Reithmeier, E 2023, Noise-robust registration of microscopic height data using convolutional neural networks. in O Matoba, JA Shaw & CR Valenta (eds), SPIE Future Sensing Technologies 2023., 123270D, Proceedings of SPIE - The International Society for Optical Engineering, vol. 12327, SPIE, SPIE future sensing Technologies, Yokohama, Japan, 18 Apr 2023. https://doi.org/10.1117/12.2644620
Siemens, S., Kästner, M., & Reithmeier, E. (2023). Noise-robust registration of microscopic height data using convolutional neural networks. In O. Matoba, J. A. Shaw, & C. R. Valenta (Eds.), SPIE Future Sensing Technologies 2023 Article 123270D (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 12327). SPIE. https://doi.org/10.1117/12.2644620
Siemens S, Kästner M, Reithmeier E. Noise-robust registration of microscopic height data using convolutional neural networks. In Matoba O, Shaw JA, Valenta CR, editors, SPIE Future Sensing Technologies 2023. SPIE. 2023. 123270D. (Proceedings of SPIE - The International Society for Optical Engineering). doi: 10.1117/12.2644620
Siemens, Stefan ; Kästner, Markus ; Reithmeier, Eduard. / Noise-robust registration of microscopic height data using convolutional neural networks. SPIE Future Sensing Technologies 2023. editor / Osamu Matoba ; Joseph A. Shaw ; Christopher R. Valenta. SPIE, 2023. (Proceedings of SPIE - The International Society for Optical Engineering).
Download
@inproceedings{0024074bd0954514929f12dedf48dbc9,
title = "Noise-robust registration of microscopic height data using convolutional neural networks",
abstract = "In this work, a deep convolutional neural network is proposed to improve the registration of microtopographic data. For this purpose, different mechanical surfaces were optically measured using a confocal laser scanning microscope. A wide range of surfaces with different materials, processing methods, and topographic properties, such as isotropy and anisotropy or stochastic and deterministic features, are included. Training and testing datasets with known homographies are generated from these measurements by cropping a fixed and moving image patch from each topography and then randomly perturbing the latter. A pseudo-siamese network architecture based on the VGG Net is then used to predict these homographies. The network is trained with a supervised learning approach where the Euclidean distance between the predicted and the ground truth gives the loss function. The 4-point homography parameterization is used to improve the loss convergence. Furthermore, different amounts of image noise are added to enhance the prediction{\textquoteright}s robustness and prevent overfitting. The effectiveness of the proposed method is evaluated through different experiments. First, the network performance is compared to intensity-based and feature-based conventional registration algorithms regarding the resulting error, the noise-robustness, and the processing speed. In addition, images from the Microsoft Common Objects in Context (COCO) dataset are used to verify the network{\textquoteright}s generalization capability to new image types and contents. The results show that the learning-based approach offers much higher robustness regarding image noise and a much lower processing time. In contrast, conventional algorithms have a smaller registration error without image noise.",
keywords = "Confocal Laser Scanning Microscopy, Convolutional Neural Network, Homography estimation, Image registration, Machine learning, Surface metrology",
author = "Stefan Siemens and Markus K{\"a}stner and Eduard Reithmeier",
year = "2023",
month = may,
day = "22",
doi = "10.1117/12.2644620",
language = "English",
series = "Proceedings of SPIE - The International Society for Optical Engineering",
publisher = "SPIE",
editor = "Osamu Matoba and Shaw, {Joseph A.} and Valenta, {Christopher R.}",
booktitle = "SPIE Future Sensing Technologies 2023",
address = "United States",
note = "SPIE future sensing Technologies ; Conference date: 18-04-2023 Through 21-04-2023",

}

Download

TY - GEN

T1 - Noise-robust registration of microscopic height data using convolutional neural networks

AU - Siemens, Stefan

AU - Kästner, Markus

AU - Reithmeier, Eduard

PY - 2023/5/22

Y1 - 2023/5/22

N2 - In this work, a deep convolutional neural network is proposed to improve the registration of microtopographic data. For this purpose, different mechanical surfaces were optically measured using a confocal laser scanning microscope. A wide range of surfaces with different materials, processing methods, and topographic properties, such as isotropy and anisotropy or stochastic and deterministic features, are included. Training and testing datasets with known homographies are generated from these measurements by cropping a fixed and moving image patch from each topography and then randomly perturbing the latter. A pseudo-siamese network architecture based on the VGG Net is then used to predict these homographies. The network is trained with a supervised learning approach where the Euclidean distance between the predicted and the ground truth gives the loss function. The 4-point homography parameterization is used to improve the loss convergence. Furthermore, different amounts of image noise are added to enhance the prediction’s robustness and prevent overfitting. The effectiveness of the proposed method is evaluated through different experiments. First, the network performance is compared to intensity-based and feature-based conventional registration algorithms regarding the resulting error, the noise-robustness, and the processing speed. In addition, images from the Microsoft Common Objects in Context (COCO) dataset are used to verify the network’s generalization capability to new image types and contents. The results show that the learning-based approach offers much higher robustness regarding image noise and a much lower processing time. In contrast, conventional algorithms have a smaller registration error without image noise.

AB - In this work, a deep convolutional neural network is proposed to improve the registration of microtopographic data. For this purpose, different mechanical surfaces were optically measured using a confocal laser scanning microscope. A wide range of surfaces with different materials, processing methods, and topographic properties, such as isotropy and anisotropy or stochastic and deterministic features, are included. Training and testing datasets with known homographies are generated from these measurements by cropping a fixed and moving image patch from each topography and then randomly perturbing the latter. A pseudo-siamese network architecture based on the VGG Net is then used to predict these homographies. The network is trained with a supervised learning approach where the Euclidean distance between the predicted and the ground truth gives the loss function. The 4-point homography parameterization is used to improve the loss convergence. Furthermore, different amounts of image noise are added to enhance the prediction’s robustness and prevent overfitting. The effectiveness of the proposed method is evaluated through different experiments. First, the network performance is compared to intensity-based and feature-based conventional registration algorithms regarding the resulting error, the noise-robustness, and the processing speed. In addition, images from the Microsoft Common Objects in Context (COCO) dataset are used to verify the network’s generalization capability to new image types and contents. The results show that the learning-based approach offers much higher robustness regarding image noise and a much lower processing time. In contrast, conventional algorithms have a smaller registration error without image noise.

KW - Confocal Laser Scanning Microscopy

KW - Convolutional Neural Network

KW - Homography estimation

KW - Image registration

KW - Machine learning

KW - Surface metrology

UR - http://www.scopus.com/inward/record.url?scp=85173430098&partnerID=8YFLogxK

U2 - 10.1117/12.2644620

DO - 10.1117/12.2644620

M3 - Conference contribution

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - SPIE Future Sensing Technologies 2023

A2 - Matoba, Osamu

A2 - Shaw, Joseph A.

A2 - Valenta, Christopher R.

PB - SPIE

T2 - SPIE future sensing Technologies

Y2 - 18 April 2023 through 21 April 2023

ER -

By the same author(s)