Parametrization of Cross-Sections by CNN Classification and Moments of Area Regression for Frame Structures

Denk M, Rother K, Neuhäusler J, Petroll C, Paetzold K (2022)

Publication Type: Conference contribution

Publication year: 2022

Original Authors: Martin Denk, Klemens Rother, Josef Neuhäusler, Christoph Petroll, Kristin Paetzold

Pages Range: 93-103

Conference Proceedings Title: Proceedings of the Munich Symposium on Lightweight Design 2021

ISBN: 9783662652152

DOI: 10.1007/978-3-662-65216-9_9

Abstract

The shape reconstruction of volumetric images can require classifying cross-sections such as circles or rectangles. Depending on the type of cross-section, shape parameters such as radius, width, or height must be regressed. This article addresses cross-section classification by convolutional neural networks (CNN), with further regression of the shape parameter using the moments of area. This fully recognizes the cross-section type and the corresponding geometric shape parameter for 2D binary images.

Authors with CRIS profile

Martin Denk

Involved external institutions

Hochschule für angewandte Wissenschaften München DE Germany (DE) Universität der Bundeswehr München DE Germany (DE) Technische Universität Dresden DE Germany (DE)

How to cite

APA:

Denk, M., Rother, K., Neuhäusler, J., Petroll, C., & Paetzold, K. (2022). Parametrization of Cross-Sections by CNN Classification and Moments of Area Regression for Frame Structures. In Proceedings of the Munich Symposium on Lightweight Design 2021 (pp. 93-103).

MLA:

Denk, Martin, et al. "Parametrization of Cross-Sections by CNN Classification and Moments of Area Regression for Frame Structures." Proceedings of the Proceedings of the Munich Symposium on Lightweight Design 2021 2022. 93-103.

BibTeX: Download