Indoor Single Space Matching Based On Design Kernel

Traditional floor plan matching techniques often fail to comprehensively account for design factors. To address this limitation, a fine-ranking algorithm is proposed based on the concept of spatial design cores. The algorithm extracts key geometric features, including core areas, critical placement lines, design-compliant customized wall lines as well as door and window positions, and evaluates feature similarities with an improved intersection over union algorithm to assess floor plan resemblance. Additionally, feature computation is simplified to obtain a coarse-ranking algorithm. A highly efficient and precise framework is established to identify the most similar floor plans with both ranking algorithms, and experimental results on 300 manually annotated housing layouts demonstrate the proposed system improves matching accuracy by over 40% and reduces computation time by 84.9% of that required by existing methods.