A Framework for Constructing Semantic As-is Building Energy Models (BEMs) for Existing Buildings Using Digital Images

Abstract

Retrofits of existing buildings have great potential to reduce global energy consumption and greenhouse gas emissions. Energy modeling of existing buildings, which is commonly conducted to prioritize retrofit strategies, relies on as-is building energy models (BEMs) that represent actual conditions of buildings. Recent efforts have focused on leveraging sensing technologies such as laser scanning and photogrammetry to capture as-is conditions of buildings and developing automatic methods for creating BEMs using the captured data. However, the majority of these efforts are limited to reconstructing 3D facade geometries with poor semantic information for rough BEM use. To this end, this paper presents a framework for an image-based approach to construct complete and semantic as-is BEM geometry models for existing buildings. The framework consists of four modules: 1) the data capture module that collects digital images of building facades and interior spaces and relevant “placement” information for geometry definition; 2) the building surface geometry reconstruction module that recognizes main building components required by BEM and reconstructs their 3D surface geometries from captured images; 3) the semantic enrichment module that adds the required geometry-related semantic relationships among the reconstructed building elements and interior spaces; and 4) the BEM creation module that stores the semantic geometry model in IDF data model. This framework is expected to extend existing research by creating complete (i.e. include not only building facades but also interior spaces) and semantic-rich as-is BEM geometry models.