路侧轨迹位置数据异常检测方法、设备、介质及产品
By performing orthogonal decomposition of longitudinal and lateral distances and multi-dimensional cross-validation on trajectory data, the problem of misjudgment in trajectory anomaly detection under extreme conditions in existing technologies is solved, achieving higher detection accuracy and robustness.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TUS CLOUD CONTROL (BEIJING) TECH LTD
- Filing Date
- 2026-03-13
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, trajectory anomaly detection methods based on Euclidean distance have a high false positive rate under extreme conditions such as vehicle occlusion, and cannot accurately distinguish whether the position change is a sudden change along or perpendicular to the driving direction, resulting in poor detection accuracy.
By acquiring the target's historical state information and current frame state information, the displacement between adjacent frames is orthogonally decomposed into longitudinal and lateral distances using the heading angle. Anomaly detection is performed by combining historical averages and statistical thresholds. The detection box size change rate is introduced for secondary verification. Multi-dimensional cross-validation is performed by combining the road reference yaw angle and lane line topology data from the high-precision map.
It improves the accuracy of trajectory anomaly detection, reduces the false positive and false negative rates, and enhances the robustness and reliability of roadside sensing trajectory data quality evaluation.
Smart Images
Figure CN122416769A_ABST