Scholarlink

Automated extraction of information from floor plans greatly benefits architects by enabling the reuse of floor plans. This approach emphasizes on extracting and recognizing text and graphics to obtain detailed information from floor plan images. The paper proposes an algorithm for floor plan information retrieval that incorporates geometry extraction and room detection methods. The alpha shape technique is applied for precise shape detection, with actual room areas calculated from these shapes. Furthermore, a binary decision tree model is implemented to classify rooms into categories such as drawing rooms, bedrooms, and kitchens, as well as non-room types like parking porches, bathrooms, study rooms, and prayer rooms. When evaluated on the CVC-FP (de las Heras et al. in Int J Doc Anal Recogn (IJDAR) 18(1):15–30 [1]) dataset, the proposed model achieves a mean room detection accuracy of 87.81% and an area calculation accuracy of 95.67%.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Efficient Extraction and Classification of Room Information from 2D Floor Plans Using a Hybrid Binary Decision Tree Approach with DBSCAN Clustering and Alpha Shape

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Smriti Upmanyu,
  • Rajendra Gupta

摘要

Automated extraction of information from floor plans greatly benefits architects by enabling the reuse of floor plans. This approach emphasizes on extracting and recognizing text and graphics to obtain detailed information from floor plan images. The paper proposes an algorithm for floor plan information retrieval that incorporates geometry extraction and room detection methods. The alpha shape technique is applied for precise shape detection, with actual room areas calculated from these shapes. Furthermore, a binary decision tree model is implemented to classify rooms into categories such as drawing rooms, bedrooms, and kitchens, as well as non-room types like parking porches, bathrooms, study rooms, and prayer rooms. When evaluated on the CVC-FP (de las Heras et al. in Int J Doc Anal Recogn (IJDAR) 18(1):15–30 [1]) dataset, the proposed model achieves a mean room detection accuracy of 87.81% and an area calculation accuracy of 95.67%.