LTA to roll out AI bus-camera system to detect encroachments and road defects

Singapore's public transport system is about to lean much harder on computer vision. The Land Transport Authority (LTA) will progressively roll out an AI-powered video analytics system from the fourth quarter of 2026 that taps the cameras already sitting on public buses to spot errant motorists, road defects and non-compliant road works — turning the daily commute itself into a rolling surveillance network for the road network.

The system, which completed trials with an average accuracy of 90 per cent across the scenarios tested, detects vehicles that encroach into bus lanes or park illegally along roads, and flags infrastructure defects and poor road works. It will be expanded to cover all bus routes by 2030, LTA said on Friday (Aug 21). Senior Minister of State for Transport and National Development Sun Xueling framed the move as relieving bus captains of manual monitoring duties and unlocking "the next bound of how we assess risks to safety."

The rollout sits alongside other safety measures announced at the Public Transport Safety and Security Awards 2026 at One Punggol. LTA has shortened driving time on 10 routes with runtimes exceeding two hours, converted continuous loop services to bi-directional ones, and will deploy a fully automated safety audio announcement feature across the fleet from end-2027 after piloting it on 11 services since August 2025.

The shift is emblematic of how Singapore is quietly embedding AI into everyday civic infrastructure rather than only chasing flashy deployments. Here, computer vision is doing the dull, high-volume work — monitoring lanes, checking roadworks, counting encroachments — that previously depended on human review. That is where AI tends to deliver its most reliable returns in a small, densely organised city-state.

Why it matters for Singapore: LTA's bus-camera system is a textbook example of Singapore's pragmatic AI playbook — applying machine learning to solve a concrete, measurable civic problem instead of pursuing AI for its own sake. With deployment beginning in months and full network coverage by 2030, it also signals renewed public-sector confidence in computer-vision technology, which has quietly matured into a dependable enforcement and inspection tool across government agencies.