The Hague tests AI cameras to detect laughing gas abuse

The Hague in the Netherlands recently tested how well cameras with video analytics can detect the abuse of nitrous oxide or ‘laughing gas’.

During a test day on the Scheveningen boulevard, the solution from smart camera provider Avutec and data platform provider Monotch was assessed with the help of the municipality.

Since January 1 of this year, it has been illegal in the Netherlands to transport, sell, manufacture or possess laughing gas for recreational use due to concerns over health risks and road safety.

Avutec’s AI cameras have been trained on how to detect various nitrous oxide balloons. Detection data – not footage, due to privacy – was sent straight to the municipality’s enforcement team via Monotch’s platform, to enable staff to assess the situation directly.

“The municipality of the Hague wants to reduce any drugs-related nuisance and therefore uses the most advanced technologies available,” said Anouk Zwaan, Safety Coordinator, The Hague.

Results

The system was trained to detect a balloon that is held at mouth level. To see how well this works, The Hague ran tests with actors and balloons as well as other objects such as a football, beach bucket and a balloon on a string.

Frank Hitzert, a spokesperson for The Hague, told Cities Today: “The test shows that in most cases the system is able to give a signal at the right times.

“The test has been completed and the camera has been removed. We are now investigating whether we should implement it, and if so, whether it will be in Scheveningen again or in other relevant places in the city.”

The trial was part of The Hague’s SHIELD initiative through which the municipality and the police collaborate on technology innovations related to safety and security.

Fast alert

“Harnessing the power of artificial intelligence, we have trained our cameras with specially curated datasets to identify both balloons used for laughing gas and people,” said Walter Verbruggen, Sales Director at Avutec.

“These datasets incorporate a diverse range of people, groups, and balloons of various shapes and colours. The system’s algorithm integrates these identifications, thereby minimising the risk of false alarms. It enables the surveillance system to accurately target and follow up on the actual perpetrators.”

“What makes this total solution unique is immediately sending a signal to the enforcement department so that it can respond quickly and adequately,” added Gary Lin, International Business Developer at Monotch. “This all happens in fractions of seconds via our platform.”