A hybrid top-down, bottom-up approach for 3D space parsing using dense RGB point clouds

M. Mehranfar; A. Braun; A. Borrmann

doi:10.1201/9781003354222-70

A hybrid top-down, bottom-up approach for 3D space parsing using dense RGB point clouds

M. Mehranfar, A. Braun, A. Borrmann

Chair of Computing in Civil and Building Engineering

Technical University of Munich

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

1 Scopus citations

Abstract

Nowadays, despite the advanced developments in engineering, automatic large-scale point cloud processing is still one of the challenging topics in many applications. In this regard, segmentation of the indoor point clouds into partitioned spaces is highly demanded in building information modeling (BIM) and robotic society. This paper proposes a novel automatic hybrid top-down, bottom-up approach for the 3D space parsing in the building environment and inferring relations between spaces. The proposed method is based on applying a deep convolutional neural network (CNN) for semantic segmentation of main elements and the use of existing knowledge in the construction of buildings. Unlike the existing methods, the proposed approach does not require pre-knowledge about the space layout. The results of evaluating the proposed method on two datasets with different designs highlight the capability of the proposed approach in 3D space parsing, extracting wall footprints, and particularly finding the topological relation between them.

Original language	English
Title of host publication	eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022
Editors	Eilif Hjelseth, Sujesh F. Sujan, Raimar J. Scherer
Publisher	CRC Press/Balkema
Pages	551-558
Number of pages	8
ISBN (Print)	9781032406732
DOIs	https://doi.org/10.1201/9781003354222-70
State	Published - 2023
Event	14th European Conference on Product and Process Modelling, ECPPM 2022 - Trondheim, Norway Duration: 14 Sep 2022 → 16 Sep 2022

Publication series

Name	eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022

Conference

Conference	14th European Conference on Product and Process Modelling, ECPPM 2022
Country/Territory	Norway
City	Trondheim
Period	14/09/22 → 16/09/22

Keywords

BIM
Point cloud
Segmentation
Top-down approach
convolutional neural network (CNN)

Access to Document

10.1201/9781003354222-70

Cite this

Mehranfar, M., Braun, A., & Borrmann, A. (2023). A hybrid top-down, bottom-up approach for 3D space parsing using dense RGB point clouds. In E. Hjelseth, S. F. Sujan, & R. J. Scherer (Eds.), eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022 (pp. 551-558). (eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022). CRC Press/Balkema. https://doi.org/10.1201/9781003354222-70

Mehranfar, M. ; Braun, A. ; Borrmann, A. / A hybrid top-down, bottom-up approach for 3D space parsing using dense RGB point clouds. eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022. editor / Eilif Hjelseth ; Sujesh F. Sujan ; Raimar J. Scherer. CRC Press/Balkema, 2023. pp. 551-558 (eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022).

@inproceedings{4a0c15344ba742cfaf39d010ef0586ac,

title = "A hybrid top-down, bottom-up approach for 3D space parsing using dense RGB point clouds",

abstract = "Nowadays, despite the advanced developments in engineering, automatic large-scale point cloud processing is still one of the challenging topics in many applications. In this regard, segmentation of the indoor point clouds into partitioned spaces is highly demanded in building information modeling (BIM) and robotic society. This paper proposes a novel automatic hybrid top-down, bottom-up approach for the 3D space parsing in the building environment and inferring relations between spaces. The proposed method is based on applying a deep convolutional neural network (CNN) for semantic segmentation of main elements and the use of existing knowledge in the construction of buildings. Unlike the existing methods, the proposed approach does not require pre-knowledge about the space layout. The results of evaluating the proposed method on two datasets with different designs highlight the capability of the proposed approach in 3D space parsing, extracting wall footprints, and particularly finding the topological relation between them.",

keywords = "BIM, Point cloud, Segmentation, Top-down approach, convolutional neural network (CNN)",

author = "M. Mehranfar and A. Braun and A. Borrmann",

note = "Publisher Copyright: {\textcopyright} 2023 the Author(s).; 14th European Conference on Product and Process Modelling, ECPPM 2022 ; Conference date: 14-09-2022 Through 16-09-2022",

year = "2023",

doi = "10.1201/9781003354222-70",

language = "English",

isbn = "9781032406732",

series = "eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022",

publisher = "CRC Press/Balkema",

pages = "551--558",

editor = "Eilif Hjelseth and Sujan, {Sujesh F.} and Scherer, {Raimar J.}",

booktitle = "eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022",

}

Mehranfar, M, Braun, A & Borrmann, A 2023, A hybrid top-down, bottom-up approach for 3D space parsing using dense RGB point clouds. in E Hjelseth, SF Sujan & RJ Scherer (eds), eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022. eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022, CRC Press/Balkema, pp. 551-558, 14th European Conference on Product and Process Modelling, ECPPM 2022, Trondheim, Norway, 14/09/22. https://doi.org/10.1201/9781003354222-70

A hybrid top-down, bottom-up approach for 3D space parsing using dense RGB point clouds. / Mehranfar, M.; Braun, A.; Borrmann, A.
eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022. ed. / Eilif Hjelseth; Sujesh F. Sujan; Raimar J. Scherer. CRC Press/Balkema, 2023. p. 551-558 (eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - A hybrid top-down, bottom-up approach for 3D space parsing using dense RGB point clouds

AU - Mehranfar, M.

AU - Braun, A.

AU - Borrmann, A.

PY - 2023

Y1 - 2023

N2 - Nowadays, despite the advanced developments in engineering, automatic large-scale point cloud processing is still one of the challenging topics in many applications. In this regard, segmentation of the indoor point clouds into partitioned spaces is highly demanded in building information modeling (BIM) and robotic society. This paper proposes a novel automatic hybrid top-down, bottom-up approach for the 3D space parsing in the building environment and inferring relations between spaces. The proposed method is based on applying a deep convolutional neural network (CNN) for semantic segmentation of main elements and the use of existing knowledge in the construction of buildings. Unlike the existing methods, the proposed approach does not require pre-knowledge about the space layout. The results of evaluating the proposed method on two datasets with different designs highlight the capability of the proposed approach in 3D space parsing, extracting wall footprints, and particularly finding the topological relation between them.

AB - Nowadays, despite the advanced developments in engineering, automatic large-scale point cloud processing is still one of the challenging topics in many applications. In this regard, segmentation of the indoor point clouds into partitioned spaces is highly demanded in building information modeling (BIM) and robotic society. This paper proposes a novel automatic hybrid top-down, bottom-up approach for the 3D space parsing in the building environment and inferring relations between spaces. The proposed method is based on applying a deep convolutional neural network (CNN) for semantic segmentation of main elements and the use of existing knowledge in the construction of buildings. Unlike the existing methods, the proposed approach does not require pre-knowledge about the space layout. The results of evaluating the proposed method on two datasets with different designs highlight the capability of the proposed approach in 3D space parsing, extracting wall footprints, and particularly finding the topological relation between them.

KW - BIM

KW - Point cloud

KW - Segmentation

KW - Top-down approach

KW - convolutional neural network (CNN)

UR - http://www.scopus.com/inward/record.url?scp=85160439757&partnerID=8YFLogxK

U2 - 10.1201/9781003354222-70

DO - 10.1201/9781003354222-70

M3 - Conference contribution

AN - SCOPUS:85160439757

SN - 9781032406732

T3 - eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022

SP - 551

EP - 558

BT - eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022

A2 - Hjelseth, Eilif

A2 - Sujan, Sujesh F.

A2 - Scherer, Raimar J.

PB - CRC Press/Balkema

T2 - 14th European Conference on Product and Process Modelling, ECPPM 2022

Y2 - 14 September 2022 through 16 September 2022

ER -

Mehranfar M, Braun A , Borrmann A. A hybrid top-down, bottom-up approach for 3D space parsing using dense RGB point clouds. In Hjelseth E, Sujan SF, Scherer RJ, editors, eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022. CRC Press/Balkema. 2023. p. 551-558. (eWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022). doi: 10.1201/9781003354222-70

A hybrid top-down, bottom-up approach for 3D space parsing using dense RGB point clouds

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this