Roadside camera system improves traffic occupancy prediction accuracy
RoadOcc Learns When to Persist, Transport, or Refresh Memory for Roadside Occupancy Prediction
Computer Vision and Pattern Recognition
Summary
Roadside cameras watch the same street scene over and over, but it can be hard to know exactly where moving cars and bikes are, especially when the evidence is weak or old. The authors created RoadOcc, a system that learns when to keep past information, when to move it based on vehicle speed, and when to refresh with new camera observations. This way, it makes better guesses about where vehicles currently are on the road. Their tests showed improved accuracy in predicting occupied road areas compared to earlier methods.
What this means in practice
- •For traffic monitoring teams: Improve real-time roadside occupancy maps for better traffic management using learned memory routing that accounts for vehicle motion and sensor data.
- •For autonomous vehicle developers: Enhance roadside environment understanding by integrating temporally consistent occupancy data refined through motion-aware memory techniques.
Authors
Xiaokai Bai, Lei Yang, Songkai Wang, Lianqing Zheng, Si-Yuan Cao, Hui-liang Shen
Abstract
Fixed roadside cameras repeatedly observe a stable scene overlaid by sparse moving traffic. Temporal memory can recover weak observations, but reusing moving evidence at stale locations can corrupt occupancy predictions. Motion compensation addresses displacement, while reliance on the resulting history remains a separate learning problem. We introduce RoadOcc, which learns soft routing among fixed-coordinate history (\emph{Persist}), velocity-addressed history (\emph{Transport}), and current evidence (\emph{Refresh}). Motion state and class-consistent historical support supervise these source choices. Dynamic-aware cross-attention (DCA) updates candidate locations, multi-scale voxel velocity estimation (VVE) constructs transport addresses from multi-scale current--history correspondence, and velocity-guided dynamic sparse fusion (VDSF) combines routed evidence under fixed sparse-token budgets. On InfraOcc, RoadOcc reaches 65.29 mIoU and 32.37 dynamic mIoU, gains of 4.44 and 4.71 over STCOcc. Controlled address experiments show that VVE raises dynamic mIoU by 0.87 over fixed-coordinate reading. Across three seeds, supervised P/T/R adds 1.40 dynamic points over motion-corrected retrieval, while removing Refresh costs 0.32 points. Results from two transfer models, Occ3D-nuScenes, and longer intervals provide additional support. Code will be released.