AUTOMATED ROOM GRAPH EXTRACTION FROM ARCHITECTURAL FLOOR PLAN BASED ON DISTANCE AND VISIBILITY

In this paper, we introduce a unified framework for automatic room segmentation and graph extraction from floor plans. We sample random points within the activity space and cluster them using obstacle-aware travel distances and visibility to form initial room regions, then apply a linear mapping to infer inter-room relationships. Tested on the RPLAN dataset and representative spatial layouts, our method achieves high accuracy and produces visually intuitive room segmentations. Our method can facilitate large-scale plan comparison and retrieval, support spatial analytics, and assist computer-aided architectural design.