Demelsa González
MARKETING MANAGER
22 November 2022.
Comparte
The application of deep learning in video analytics enhances both security and mobility within cities. Below, we present seven examples.
A city is only as smart as the technology it employs to improve efficiency and utilise the data it generates effectively. In the realm of video security, CCTV cameras placed along public thoroughfares capture images that deep learning algorithms can convert into valuable information by categorising them based on type and behaviour.
This real-time, automated classification can be customised based on various criteria. As a result, it produces datasets that facilitate the generation of alerts and the strategic allocation of resources in a city, thereby enhancing the safety and mobility of its residents.
In practice, image analysis through artificial intelligence (AI), specifically using deep learning, can be applied in cities in the following scenarios:
Índice de contenido
1. Protecting areas of interest
Any monument or building can become a target for criminals, and it’s far more practical to prevent theft or vandalism than to deal with the consequences.
Traditional intruder detection systems distinguish movement, size and location but do not identify the type of object or other crucial details.
Deep learning technology can accurately detect individuals exhibiting suspicious behaviour, such as lingering around a building for extended periods. This behaviour, known as ‘loitering’, can be recognised by artificial intelligence, triggering an alarm to activate an inspection protocol.
2. Preventing traffic congestion
Poor management of congestion in cities can lead to traffic accidents and increased pollution.
By counting and categorising vehicles, road traffic analytics can trigger alerts when an established threshold is exceeded. Analysing statistics is also crucial. Understanding the average duration vehicles remain stationary, as well as identifying peak days and hours of congestion, enables informed decisions for city mobility planning.
In addition, it enhances traffic-light control and allows for the adjustment or alteration of lane directions as necessary.
“DEEP LEARNING” Step by Step
Deep learning to prevent crowd congestion- The City Council installs Lanaccess’s image processor in the video recorder connected to a camera monitoring a public park. The processor contains the computer algorithm called LAVA-DENSITY, a piece of code programmed to detect occupancy levels in a specific area.
- Thanks to the algorithm’s learning capabilities, it can analyse patterns, identify moving individuals and count them without human supervision. This technique, known as ‘deep learning’, enables the recognition of behavioural patterns from a large number of images autonomously and incrementally.
- Programmed to send alerts, the system will notify the control centre if it detects an alarmingly high number of people in a virtually defined area.
- The control centre will decide whether to dispatch personnel to evacuate the park.
3. Improving traffic flow at roundabouts
Vehicles consistently interact at these frequently congested junctions.
Advanced analytics detect pedestrian activity within the roundabout, providing statistical data on vehicle types entering, their crossing times and main exit routes.
The deep learning algorithm will also help devise the best strategy to ensure a seamless and safe traffic flow pattern.
4. Enforcing fines for serious offences
As traffic volumes rise, so do offences.
The algorithm can identify serious vehicle traffic violations that jeopardise road safety, such as driving against the flow of traffic or making improper turns.
The system alerts Traffic Headquarters, providing an image of the vehicle committing the offence.
5. Monitoring scooter traffic
The improper use of scooters or riders not wearing helmets poses a major safety risk and is a growing concern in cities.
Using deep learning, analytics can classify the types of vehicles in circulation, including scooters.
In busy areas, traffic officers can be deployed to enforce regulations on the use of these modes of transportation.
6. Alerting authorities to stationary persons or vehicles at level crossings
A level crossing is where a railway line crosses a road without a tunnel or bridge separating them. Despite barriers in place, these crossings pose a risk in cities, compounded by unpredictable driver and pedestrian behaviour.
An intelligent detection model integrates deep learning technology with CCTV surveillance systems to identify stationary objects at level crossings and sends real-time alerts to the control centre.
7. Mitigating crowd congestion
Counting moving objects in a video sequence is one of the straightforward tasks that an intelligent algorithm can handle.
One potential application is monitoring and managing unauthorised large gatherings, such as street parties. If the defined maximum occupancy threshold is surpassed, the system sends a real-time alert to the control centre, enabling personnel to be dispatched to the event location.
Cities are becoming more technologically advanced to improve the quality of life for their residents. Our analytics are programmed to contribute to this goal, with algorithms designed for all the applications mentioned in the article.
