AI: Smart cameras to save people from drowning in swimming pools

Artificial intelligence (AI) – your friend and swimming helper: Swimming pool companies in Munich and Wiesbaden are currently testing an Israeli surveillance system with cameras and motion detection, which should be able to save swimmers in distress from drowning. In suspicious situations, the integrated technology, which is based on machine learning according to the manufacturer, alerts the lifeguard via a smartwatch. This should enable lifeguards to provide targeted help more quickly.

“Smart” swimming pool with motion detection

Stadtwerke München (SWM) started the “Smart Swimming Pool” pilot project in the Südbad at the end of July. During the revision period, “new technology was installed for all pools” as part of the usual maintenance and repair measures, the municipal supply and service company announced at the beginning of June: “Artificial intelligence is intended to help identify movement patterns in the water, including data-controlled .”

After a test phase lasting several weeks, the cameras are now focused. In a two-year pilot project, SWM want to gain insights into whether the technology is “useful” for all the pools they operate.

Detect patterns in vector data – no faces recorded

The system used comes from the company Lynxight from Tel Aviv. It consists of two components: In the first step, the cameras used are to record the movements in the water. “They do not capture real images of individual people – so no faces either,” emphasizes the SWM. The solution converts details of the recordings into vector data and derives movement patterns from them. The images would then be deleted immediately. The camera angles covered the entire water surface.

The electronic eyes are linked to smartwatches for the on-site supervisory staff. The intelligent watches should be able to warn in real time and with an exact position if the movement in the water after around 20 seconds indicates an unusual situation and possible danger. Swimmers can also be counted: The SWM write: “In addition, it is possible to use the vector data obtained over longer periods of time to analyze how the utilization in the pools is.”

According to the Munich utility, the system is compatible with the General Data Protection Regulation (GDPR). Visitors to the indoor pool in Sendling, where the gates on the facade are lowered when the weather is nice, would be informed in writing about the AI ​​control in the entrance area and when entering the indoor pool.

Number of false alarms should decrease

“The system is not yet error-free,” the “Süddeutsche Zeitung” (SZ) found out. The AI ​​is therefore in a learning phase that should last around 60 days. So far, the solution has not been able to distinguish between people who are sunbathing motionless in the water and swimmers who need help. False warnings are the result. In such cases, the lifeguards could give feedback via their watches and train the AI ​​in this way.

“The system learns with every action and should therefore make more and more concrete predictions and classifications in the course of the project,” says SWM optimistically. The AI ​​should soon be able to support the lifeguards in situations “in which reflections, bubbles or shadows in the water or the number of people make the situation confusing”.

Despite the initial weaknesses, the supervision in the southern pool was enthusiastic, the “SZ” quoted an SWM employee as saying. The system is already a great help, especially on busy days, even if there hasn’t been a real emergency yet. The visitors would hardly have noticed them until now. No one has yet expressed the data protection concerns that have already been raised by other parties.

The third eye”

The Wiesbaden outdoor and indoor swimming pool in Kleinfeldchen has been putting the WLAN-networked system to the test for several months. Plant managers there praise it as a “third eye” that always keeps an eye on all corners and provides valuable support at the edge of the pool.

Lynxight calculates, “The typical medical cost for a near-drowning victim ranges from $75,000 for initial treatment to $180,000 per year for long-term care. The total cost of a brain injury in such a case could be more than 4 $.5 million.” Monitoring a tank with the solution, on the other hand, only costs 800 US dollars a month. According to the German Life Saving Society (DLRG), 85 percent of the fatal swimming accidents recorded in 2021 did not occur in public swimming pools, but in inland waters such as lakes and rivers.

(tw)