Abnormal loitering behaviour, on-site recognition by OpenMV

In key areas, those who truly pose a risk are often not the “people entering” but rather the “loitering individuals”. Manual monitoring struggles to continuously assess whether behaviour is abnormal, and reviewing footage after the fact means missing the optimal intervention window. OpenMV enables the system to “observe behaviour”, not merely “view images”.

In locations such as residential community entrances and exits, school gates, factory premises entrances and exits, and park entrances and exits, prolonged loitering, repeated back-and-forth movement, or lingering in sensitive areas by individuals often indicates potential security risks—for example, reconnaissance activities, unauthorised entry, or suspicious observation. If such activities are not detected and addressed promptly, subsequent security incidents may occur.

Although traditional security systems can record video footage, they primarily rely on manual real-time monitoring or post-event playback of recordings. Manual monitoring is not only labour-intensive but also makes it difficult to promptly and accurately determine whether a behaviour is abnormal; by the time an issue is identified, the optimal intervention window has often already passed.

Based on the SingTown OpenMV intelligent camera, the system can continuously perceive and analyse personnel behaviour at the camera end, enabling automatic identification and immediate response to abnormal loitering behaviour.

After the camera is fixedly deployed, OpenMV continuously captures footage of the monitored area and performs human detection, target tracking, and trajectory recording on-device, enabling continuous observation of individuals entering the area.

By comprehensively analysing behavioural characteristics such as personnel dwell time, activity range, movement trajectory, and repeated paths, OpenMV can determine whether personnel exhibit loitering behaviour that significantly deviates from normal passage patterns and distinguish such behaviour from daily activity patterns, thereby reducing false alarm rates.

When the relevant behavioural metrics reach the preset threshold, the system can immediately output identification results at the edge, triggering local audio-visual alerts or integrating with security systems to enable rapid response and on-site intervention, without requiring cloud-based computation at any stage.

By embedding behavioural analysis capabilities directly into the camera, OpenMV enables security systems to achieve on-site intelligent perception with low latency, low bandwidth usage, and high stability—upgrading safety management from “passive video recording” to “proactive alerting”.