Toward automated generation of parametric BIMs based on hybrid video and laser scanning data

Abstract

Only very few constructed facilities today have a complete record of as-built information. Despite the growing use of Building Information Modelling and the improvement in as-built records, several more years will be required before guidelines that require as-built data modelling will be implemented for the majority of constructed facilities, and this will still not address the stock of existing buildings. A technical solution for scanning buildings and compiling Building Information Models is needed. However, this is a multidisciplinary problem, requiring expertise in scanning, computer vision and videogrammetry, machine learning, and parametric object modelling. This paper outlines the technical approach proposed by a consortium of researchers that has gathered to tackle the ambitious goal of automating as-built modelling as far as possible. The top level framework of the proposed solution is presented, and each process, input and output is explained, along with the steps needed to validate them. Preliminary experiments on the earlier stages (i.e. processes) of the framework proposed are conducted and results are shown; the work toward implementation of the remainder is ongoing.

Introduction

The current state-of-the-art approach to collecting, organizing and integrating as-built data of a constructed facility into a single data structure is to model it using building information modelling (BIM) tools [1]. This approach generates parametric building models by producing logical building objects and the parametric relationships among them. The process starts by collecting spatial data on site through state-of-the-art surveying technologies, such as laser scanning (LIDAR) and photogrammetry. The resulting spatial data must then be manually stitched into a 3D surface with some algorithmic help for fine stitching. The points on the 3D surface are then manually replaced by objects, by having an operator observe the data, identify each object type, search for it in a database of standardized objects, and fit it on the surface with some help from fitting algorithms for optimal fitting. Following that, any as-built attributes can be assigned to each object manually.

Although this process is significantly assisted by recent technological advancements, most of it remains manual. Researchers along with professional modellers such as VECO [2] and Reality Measurements [3] have reported that more than two thirds of the efforts needed to model even simple facilities are spent on manual conversion of the surface data to a BIM. This problem results in significant cost and effort that is needed to convert the sensed surface of constructed facilities to the desired model, which undermines any benefits of automated spatial modelling for the majority of facilities. According to studies by Minhindu and Arayici [4] and Young et al. [5], BIM adoption is growing in some countries such as U.S., Denmark, Finland and Norway. However, as McCarthy [6] had predicted, for small construction projects, the net savings can barely justify adoption and utilization of this technology. As a result, the penetration of innovative spatial modelling technologies to smaller projects and companies in the Architecture, Engineering & Construction (AEC) industry is slow and they will wait unless significant savings can occur.

This paper presents a novel framework that holds promise to automate almost entirely the generation of as-built parametric BIMs of constructed facilities, ranging from residential housing to industrial structures. This framework uses spatial and visual data collected in the field to generate images and the 3D surface represented as a point cloud. The next step is to stitch images together in order to integrate them into a single 3D representation. Then, by analyzing geometric surface and surface texture information simultaneously against an established taxonomy of building material samples and shapes and their relationships to one another, it is possible to generate representations that can be used to classify common building objects into corresponding object categories. This is followed by novel fitting algorithms that identify the exact object from the identified category that fits the spatial and visual description. This allows the automated identification of most frequent items and presents them to the modeller for verification. The user/modeller is then responsible only for modelling specialty/infrequent items and checking/correcting the modelling results of the automated process. The major scientific breakthrough of this framework is the automation of several repetitive processes that are unwieldy and but time-consuming to be completed by human operations. These additions are expected to significantly reduce the cost, time and resources that are needed to build a parametric building model compared to current practice.