SceneGTMM：基于共形映射的场景感知可迁移GNN-Transformer双图交互地图匹配框架
原标题：SceneGTMM: A Conformal Mapping-based Scene-Aware Transferable GNN-Transformer Dual-Graph Interaction Framework for Map Matching
AI 摘要
阿里巴巴高德地图团队提出SceneGTMM框架，一种基于共形映射的场景感知可迁移GNN-Transformer双图交互地图匹配方法。该框架通过共形映射场景相对策略、GNN-Transformer双图交互架构和CRF增强结构化预测，在跨区域迁移和噪声鲁棒性上取得突破。实验表明，在定位误差16-50米的多源轨迹数据上，准确率超过80%，比传统HMM提升5.3%，并在跨城市迁移场景中优于现有深度学习方法。
正文节选
[Page 1] SceneGTMM: A Conformal Mapping-based Scene-Aware Transferable GNN-Transformer Dual-Graph Interaction Framework for Map Matching Yongliang Zhang†‡ Feng Song Ji Chen zhangyongliang.zha@alibaba- feng.song@alibaba-inc.com cj270839@alibaba-inc.com inc.com Amap, Alibaba Amap, Alibaba Amap, Alibaba Group,be