A framework of an image-based integrated approach to create as-is building information models for existing buildings

Abstract

Building Information Modelling (BIM) provides an intelligent and parametric digital platform to support activities throughout the life cycle of a building and has been widely used for new building construction projects. However, most existing buildings today often do not have completed as-is information documents, nor existed meaningful BIM models. Despite the growing use of BIM models for relatively new buildings and the improvement in as-is records, incomplete or even incorrect information is still one of the main reasons for the low-level efficiency in facilities management. Furthermore, as-is BIM modelling for existing buildings is considered to be a time-consuming and expensive process, which requires great effort, time and skilled workers. Hence, developing an efficient way to create an as-is BIM model would be essentially the foremost step for effective operations and maintenance of existing buildings in their life cycle using cutting-edge BIM tools. We aim at developing a framework to establish a high efficient as-is BIM modelling system that integrates diverse building data in different formats (e.g., photo images, drawings, text data) with low cost, in order to improve efficiency and effectiveness of operations and maintenance, and furthermore possibly support refurbishment of a building. With this ultimate goal, this paper focuses on analysing state-of-the-art object recognition and reconstruction methods applicable for creating as-is BIM models, and summarises merits and limitations of these existing techniques and methods. Then, a framework is established with the most suitable methods on the basis of applying image sources and existing documents of the target building. The framework mainly includes a mechanism that supports automated creation of as-is objects with their meanings and information, and an assisting information library for the target building. The mechanism will use intelligent reasoning algorithms as appropriate such as the hybrid neuro-fuzzy algorithm in order to take into account possibilities of information shortage. The information library will mainly consist of four functional modules and four information modules (i.e., surface information, geometric representation, extracted information, and features information). Furthermore, key challenges and current progress are also addressed in this paper.