跳到正文
arXiv cs.AI· Emanuele Iacobelli, Simone Teglia, Francesco Pro, Claudia Melis Tonti, Irene Amerini·· 2 小时前AI 评分35

OCTR2A:城市交叉路口实时 RGB 转 BEV 轨迹重建与预测

OCTR2A: Real-Time RGB-to-BEV Trajectory Projection and Forecasting at Urban Intersections

AI 导读

研究者提出 OCTR2A 架构,仅用城市基础设施上对角安装的 RGB 相机即可实时重建并预测路口 agent 的 BEV 轨迹,无需 LiDAR 或高精地图。该方法以 10Hz 下 12 帧观测和 12 帧预测,在单张消费级 GPU 上实现实时运行,投影—预测 ADE/FDE 为 0.144/0.140–0.154/0.241 米。

正文 · 原文

View PDF HTML (experimental)

Abstract:Trajectory forecasting at urban intersections is often studied assuming that reliable bird's-eye-view (BEV) motion histories are available, which in practice requires costly resources such as LiDAR, aerial views, HD maps, or calibrated infrastructure. We address this upstream perception-to-forecasting problem by reconstructing and predicting agents' BEV trajectories only from paired opposite-corner RGB cameras mounted on existing urban infrastructure at variable heights at inference time. We introduce the Opposite-Corner Trajectory Reconstruction and Anticipation Architecture (OCTR2A), a modular pipeline integrating per-view detection with class refinement, geometric cross-camera matching, transformer-based BEV projection, and egocentric GRU forecasting, trained in cascade and mainly supervised with a soft multi-scale spatial-overlap loss. Since no public benchmark provides this sensing configuration, we introduce R3ST++, a Blender-based synthetic benchmark built from SIND trajectories that provides synchronized opposite-corner views, cross-view correspondences, and BEV ground truth, while also serving as a deployment-support simulator for site-specific fine-tuning. Experiments on these datasets show that, with 12 observed and 12 predicted frames at 10\,Hz, OCTR2A achieves projection~-~forecasting ADE/FDE of 0.144/0.140~-~0.154/0.241~meters in real time on a single consumer GPU. Evaluated on ground-truth BEV histories, the forecasting module improves over the strongest state-of-the-art methods on SIND by 66.7\%/41.7\% ADE/FDE. Transfer experiments on TUMTraf-V2X show that R3ST++ pretraining supports adaptation to real-world roadside domains with different sensing configurations.
Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
Cite as: arXiv:2611.24352 [cs.CV]
  (or arXiv:2611.24352v1 [cs.CV] for this version)
  https://doi.org/10.48550/arXiv.2611.24352

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Emanuele Iacobelli [view email]
[v1] Wed, 18 Nov 2026 12:47:15 UTC (17,141 KB)

来源:arXiv cs.AI · arxiv.org