基于弱监督学习的视频监控异常事件检测模型训练方法
By constructing reliability weights through the calculation of differential mutation values and manifold coherence values, the problem of noise false alarms in video surveillance is solved, achieving highly accurate and robust anomaly detection, which is suitable for video surveillance in modern industrial parks and smart factories.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ANYU HEZHONG TECH CO LTD
- Filing Date
- 2026-04-20
- Publication Date
- 2026-07-17
AI Technical Summary
Existing video surveillance anomaly detection technologies suffer from high-frequency burst noise false alarms in complex environments due to the distance-semantic gap. Existing methods cannot effectively distinguish between real anomalies and noise, resulting in a high false alarm rate.
By calculating differential mutation values and manifold coherence values, a reliability weighting mechanism is constructed. By integrating spatiotemporal features and utilizing the temporal continuity of real anomalies and the transient isolation of noise, a weighted ranking loss function is constructed for model training.
It significantly reduces the false alarm rate in complex monitoring scenarios, improves the accuracy and robustness of detecting real abnormal events, and can automatically identify and suppress high-frequency noise without precise time positioning labels.
Smart Images
Figure CN122049790B_ABST