Leon-Garza, Hugo and Hagras, Hani and Peña-Rios, Anasol and Conway, Anthony and Owusu, Gilbert (2022) A type-2 fuzzy system-based approach for image data fusion to create building information models. Information Fusion, 88. pp. 115-125. DOI https://doi.org/10.1016/j.inffus.2022.07.007
Leon-Garza, Hugo and Hagras, Hani and Peña-Rios, Anasol and Conway, Anthony and Owusu, Gilbert (2022) A type-2 fuzzy system-based approach for image data fusion to create building information models. Information Fusion, 88. pp. 115-125. DOI https://doi.org/10.1016/j.inffus.2022.07.007
Leon-Garza, Hugo and Hagras, Hani and Peña-Rios, Anasol and Conway, Anthony and Owusu, Gilbert (2022) A type-2 fuzzy system-based approach for image data fusion to create building information models. Information Fusion, 88. pp. 115-125. DOI https://doi.org/10.1016/j.inffus.2022.07.007
Abstract
Building Information Modelling (BIM) is a standard digital process that fuses buildings information from different sources into a 3D model during their lifecycle. For new construction sites using BIM, it is possible to monitor the cost, schedule, and changes throughout the lifecycle; however, existing buildings do not have a BIM model. Manually creating the BIM models for existing buildings is a high-cost task, both in time and money, hence there is a need for extracting information from available paper-based documentation and fuse it into a BIM model. The struggle of facility management and utility companies to fully adopt a BIM process (due to their high volumes of paper-based documentation of existing buildings) has led to the research on creating these 3D BIM models from 2D floor plan images. This paper presents a novel processing pipeline to extract 2D digital information from floorplans, fusing it into a 3D BIM model. The work focuses on fusing the available information to create the structure of the building in BIM format, which is considered the essential step before looking on working with other sources of data. In this process, we introduce a type-2 fuzzy logic based Explainable Artificial Intelligence (XAI) approach for the semantic segmentation step. The approach consists of using the outputs of type-2 fuzzy logic systems to classify a pixel as wall or background, by using information around and from the pixel of interest as the inputs to the system. After the semantic segmentation step, the output of the type-2 fuzzy logic goes through a noise removal process and finally a transformation from 2D to 3D by assigning the corresponding BIM tag to each identified element. The proposed type-2 fuzzy logic semantic segmentation approach produced comparable results (97.3% mean Intersection over Union (IoU) performance metric value) to the opaque box model approach based on Convolutional Neural Network (CNN) (99.3% mean IoU performance metric value). However, the type-2 fuzzy XAI system benefits from being an augmentable and interpretable model, which means that human users can understand the decision process and modify the model using their expert knowledge.
Item Type: | Article |
---|---|
Uncontrolled Keywords: | Building Information Modelling; BIM; Semantic segmentation; Type-2 fuzzy logic systems; Convolutional neural network (CNN) |
Divisions: | Faculty of Science and Health Faculty of Science and Health > Computer Science and Electronic Engineering, School of |
SWORD Depositor: | Unnamed user with email elements@essex.ac.uk |
Depositing User: | Unnamed user with email elements@essex.ac.uk |
Date Deposited: | 25 Nov 2022 18:37 |
Last Modified: | 16 May 2024 21:24 |
URI: | http://repository.essex.ac.uk/id/eprint/33262 |
Available files
Filename: BIM - Information Fusion_Final_Revised.pdf
Licence: Creative Commons: Attribution-Noncommercial-No Derivative Works 3.0