Elevation recognition in architecture drawings for 3D converting to BIM modelsTools Yin, Mengtian (2019) Elevation recognition in architecture drawings for 3D converting to BIM models. MRes thesis, University of Nottingham.
AbstractBuilding Information Modelling (BIM) is recognized as the next generation of techniques that will bring about digitalization revolution to global AEC industry. Developing countries, such as China, still use CAD in building project. It is valuable to transform architecture drawings to BIM models in a cost-efficient way. Current researches on 3D reconstruction of architecture drawings work on floor plan detection. The problem is elevation and height information in floor plan is not always accessible. This research studies how to use layer property in CAD to detect elevation drawing and combine with floor plan recognition to deliver a comprehensive 3D reconstruction. The layer property and conventions are studied, and an automatic layer classification method (ALCM) is proposed to identify the semantic meaning of layer. Based on ALCM, a method to detect elevation drawing is proposed. Bounding box is created from the floor plan detection and visibility analysis. It can achieve elevation member measuring by searching the inside primitives. ALCM is tested with 70 plan drawings. The average accuracy is 95%, which implies reliable performance. Taking advantage of ALCM, the layer of demanding structural members or symbols are known. The results are inputted in the implementation of elevation detection testing. 50 pieces of elevation drawings are used to evaluate the orientation recognition method; two sets of complete architecture drawings, containing 8 elevation drawings and 36 floor plans, are adopted to test the rest of algorithms. It is found that the precision and recall rate of visibility analysis is 70.6% and 98.4%, which suggests an optimistic capability to catch visible objects, but some invisible objects are also mistaken. It leads to creation of invalid bounding box. Through the elevation detection algorithm, 87% of visible members in elevation drawing successfully acquire their height and offset value. The algorithm is applied in a practical use in University of Nottingham, Ningbo, China. A teaching building is automatically modeled in Revit with correct family dimension and position.
Actions (Archive Staff Only)
|
Tools
Tools
![[img]](/style/images/fileicons/application_pdf.png)
