“AI Urban Management” is here: an automated unauthorised street trading detection system built on OpenMV

Illegal street vending recurs repeatedly; the challenge of governance lies not in “detecting it once” but in its “long-term persistence.” OpenMV enables continuous visual monitoring, shifting management from periodic rectification to daily perception.

Unauthorised street trading is a common issue in urban governance; trading facilities such as stalls, shelves, and tables encroach upon footpaths or carriageways, not only disrupting traffic order but also posing safety hazards—particularly during peak morning and evening hours or in areas with high pedestrian density.

Management relying on manual inspections incurs high labour costs and struggles to provide long-term, continuous oversight, easily resulting in management blind spots where issues recur immediately upon personnel departure.

Based on the SingTown OpenMV smart camera, automated and routine visual monitoring can be conducted in areas prone to recurring incidents, such as key road sections, areas surrounding schools, and commercial districts.

After the camera is fixed in place, OpenMV runs object detection algorithms on-device to automatically identify objects related to encroachment on public space, such as stalls, shelves, sunshades, and trolleys. It also determines in real time whether encroachment has occurred by combining road boundaries, pedestrian walkway extents, and configured virtual rule lines.

When the system detects that business facilities are encroaching upon roads, obstructing pedestrian pathways, or operating beyond their licensed scope, it can instantly output identification results at the edge device for on-site alerts, evidence collection for violations, or automatic reporting to the urban management platform, thereby providing objective evidence for subsequent law enforcement.

Leveraging the on-device AI capabilities and edge deployment advantages of OpenMV, the management of unauthorised street occupation has shifted from periodic rectification to continuous perception and real-time detection, significantly reducing the burden on manual inspections and enhancing both the responsiveness and precision of urban governance.