发表机构
University of Wyoming（怀俄明大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对民用标准图纸的OCR自动化方法会丢失关键几何信息，本文提出视觉优先多模态RAG框架PlanSightRAG，整合多向量检索与智能体架构，构建基准数据集并验证其在检索、合规检查任务上的性能，还实现了自主视觉规则 grounding。
AI 中文摘要
民用基础设施合规检查长期依赖工程师手动读取传统二维图纸，但基于光学字符识别（OCR）的自动化方法会剥离解读这些图纸所需的几何形状与布局信息。本文提出一种名为PlanSightRAG的视觉优先多模态检索增强生成（RAG）框架，该框架直接对图纸图像进行索引与推理，整合了ColNomic-3B多向量检索模块、智能体式Planner-Retriever-Auditor-Synthesizer架构以及MaxSim热图作为证据追踪。我们基于五个州交通部（DOT）的标准图纸（共1898页）构建了包含4056对样本的基准数据集。PlanSightRAG在零样本检索任务上达到91.47%的Recall@5，在保留的密歇根州交通部语料库上达到91.40%。在合成的参数化生成合规图纸上，我们的Qwen2.5-VL-72B管道仅在提供预先确定的规则阈值时达到100%的判定准确率，而未使用视觉语言模型（VLM）的OCR基准已达到76.4%的准确率，这一结果构成了可控的上限。最后，我们展示了自主视觉规则 grounding能力，即无需任何人工提供的规则，直接从规范语料库中提取数值限值。
英文摘要
Civil infrastructure compliance checking has long relied on engineers manually reading legacy 2D plans; however, OCR-based automation strips away the geometry and layout essential for interpreting these plans. We present a Visual-First Multimodal Retrieval-Augmented Generation (RAG) framework called PlanSightRAG. It indexes and reasons directly over plan imagery, integrates a ColNomic-3B multi-vector retrieval, an agentic Planner-Retriever-Auditor-Synthesizer, and MaxSim heatmaps as an evidence trail. We introduce a 4,056-pair benchmark from five state Departments of Transportation (DOT) standard plans (1,898 pages). PlanSightRAG achieves 91.47% Recall@5 on zero-shot retrieval, while on a held-out Michigan DOT corpus, it achieves 91.40%. On synthetic, parametrically-generated compliance drawings, our Qwen2.5-VL-72B pipeline reaches 100% verdict accuracy only when supplied a pre-resolved rule threshold, a controlled ceiling that a non-VLM OCR baseline already reaches at 76.4%. Finally, we demonstrate autonomous visual rule-grounding by extracting numeric limits directly from a specification corpus without any human-supplied rules.
Comments32 pages, 9 figures, 25 tables. Preprint submitted to Automation in Construction
