LTA looks for ways to detect traffic offences using bus cameras
Cameras on all 6,000 public buses here could in the future be part of a network used to detect traffic offences, after the Land Transport Authority (LTA) issued a call for proposals for such a video analytics system.
The authority is looking to detect traffic violations that include vehicles that occupy bus lanes during restricted hours, stop in yellow boxes or fail to give way to buses at bus priority boxes, or are illegally parked.
The system would have to record the offending vehicle’s model, colour, type and licence plate number, according to tender documents.
Besides detecting traffic offences, LTA also wants the new system to use bus cameras to spot infrastructural defects on the roads and surrounding areas, and to improve the authority’s operational response time.
LTA, which issued a call for proposals on Oct 7 via government procurement portal GeBIZ, also hopes to detect pedestrians who are crossing at undesignated crossings, and identify areas with a large number of pedestrians.
At present, video footage from bus cameras is recorded during bus service hours and must be retrieved manually from on-board recorders. They are typically used in incident investigations and to ensure the safety of passengers.
Current enforcement efforts and checks on infrastructure are resource- and labour-intensive, limited by the deployment of enforcement officers and the use of mobile cameras stationed at bus stops and illegal parking hot spots.
Submitted proposals should include the use of artificial intelligence (AI) to scan and automatically detect various offences and infrastructural defects.
They have to identify the type and location of the offence or defect being detected, capture an image for reference, monitor the repair status for defects, and alert LTA officers within 24 to 48 hours.
The proposed systems should not require much involvement from bus captains while they are driving, and have minimal impact on bus operations.
Shortlisted companies will have to conduct live demonstrations of their proposals and, if selected, will need to do a further proof-of-concept test that will last four months.
Asked about the rationale and expected timeline for the project, LTA said it regularly evaluates how to use technology to improve its processes.
“As video analytics/AI detection solutions become more developed, LTA is seeking solutions that tap data from bus cameras facing the roads to enhance traffic and road operations,” said an LTA spokesman, who did not say if this would be rolled out on all 6,000 buses.
He added that the authority will assess the feasibility and reliability of the proposed solutions before determining subsequent steps.
AI experts welcomed the move to tap bus cameras, but highlighted several challenges.
Professor Pradeep Varakantham, director of the Singapore Management University’s (SMU) collaborative, robust and explainable AI-based decision-making lab, said the proposed solution may face difficulty in differentiating visual clutter – generated from the continuous movement of people and vehicles on the roads – from an offence or defect.
Prof Pradeep, who also teaches computer science at SMU, added that bus cameras may not provide a detailed three-dimensional view to recognise offences or defects properly, and additional laser sensors might be needed to provide clearer images.
Mr Laurence Liew, director of AI innovation at research programme AI Singapore, said the success of this system would hinge on robust camera specifications, which can provide high-quality footage, as cameras must also be able to withstand Singapore’s tropical weather conditions such as heavy rain and variable lighting conditions.
Motorists like Mr Sam Yew are looking forward to the new system.
The 38-year-old regional manager called it a “great idea” to leverage AI to detect traffic offences and structural defects, as it would be a “very efficient way” of improving road safety and reducing traffic violations in the long run.
Mr Yew, who has been driving for 14 years, added that he is anticipating better driving conditions and improved road safety, as more drivers may comply with traffic rules since they know that their actions will be easily detected by bus cameras.
But he noted that AI could be susceptible to technical issues and detect “false positives” – misidentify traffic offences and road defects – which could lead to motorists sending in appeals and contribute to more work for LTA.
Mr Yew also said LTA should put stringent data protection measures in place to prevent unauthorised usage of the data.
Get The New Paper on your phone with the free TNP app. Download from the Apple App Store or Google Play Store now