15 December, 2021
Modern cities are continuously evolving and remain key to humanity’s present and future. Generating 80% of the world’s GDP and emitting more than 70% of the annual global carbon emissions, cities remain multi-hubs of commerce, technology, and culture but also major sources of pollution and inequality. Promoting and maintaining economic growth, prosperity and equity is both a fundamental target and a significant challenge, but as Azeem Azhar and others have argued “cities appear to be better judges of what their populations need, in comparison to national governments”. However, for cities to lead, respective authorities will need to emphasize on smart city digitalization, in order to attract and scale investments in urban infrastructure.
How methods of yesterday have exceeded their capacity
Over recent years, network-based technologies have been providing the tools for gathering data and making them available for officials to analyse and take well informed decisions. Either concerning the management of resources, traffic or public infrastructure, inputs from scattered sensors and cameras help to collect valuable information. However, data streams and volume have proliferated almost exponentially which means that this semi-automatic method of operation is quickly losing its past effectiveness.
Why Edge Computer Vision (ECV) systems powered by AI and ML are the solution
Thankfully, new technology, if applied responsibly, can bring about the solution. As Artificial Intelligence and Machine Learning (AI/ML) algorithms became popular and proved their prowess in complex applications, a whole new – and gradually more intelligent – range of capabilities became available. Indeed, “smart cities” are no longer a futuristic term, but a reality within grasp. The latest AI/ML algorithms can make much faster, more efficient and more secure use of visual information, collected either by already installed or new cameras, and offer real-time analysis to achieve faster response, more accurate predictions and automatic decision-making. These computer vision systems can train themselves to deeply analyse pictures and videos, and subsequently identify humans, animals, objects, movement, and behavioural patterns – all whilst protecting personal information and firmly adhering to the strictest data protection regulations.
A subcategory of the Computer Vision (CV) systems which has been gaining increasing popularity is the Embedded or Edge Computer Vision (ECV) systems. It may comprise a smart IP camera and a wired or wireless transceiver but the author’s favorite system is the one consisting of a standard IP camera and a mini-PC with wireless connectivity to the cloud via low-power wide-area networks (LPWANs), such as Long-Range WAN (LoRa) and Narrowband Internet of Things (NB-IoT). According to the monitoring objective, the centralized platform dictates which AI/ML algorithm should run on the mini-PC of these localised system. Morevoer, such algorithms can be remotely updated or upgraded without the need for time consuming and expensive site visits. Visual information is then processed locally – protecting privacy whilst greatly improving speed and reducing networks costs – and decision making can be optimised without the need to contact the main platform. ECV systems achieve:
- Reduced installation costs, using existing IP cameras (if available) and negating costly investments on network hardware devices or large data storage systems.
- Reduced operational costs, requiring much less bandwidth for the data exchange between the edge AI devices and the cloud-based platform.
- Protected citizen privacy, as the edge processing algorithms inherently safeguard anonymity and only communicate identification data to a central database when an alarm is raised based on pre-defined criteria (e.g. dangerous traffic violations).
In the following paragraphs, we will provide five examples of such ECV applications and the resulting benefits:
Application #1: Smart traffic and bicycle monitoring
Greater urban density typically results into a proliferation of vehicles and an increase in traffic jams, transport times, accidents, local air pollution and carbon emissions – not to mention a pervasive feeling of tiredness, stress and anxiety. With the use of new or sometimes existing street cameras, an edge-enabled computer vision system may capture a real-time view of traffic conditions and by correlating this information with specially trained machine learning algorithms, can enable the city authorities to:
- Predict probable congestion points and propose alternative routes to drivers through messaging on the highway illuminated boards or by automatically adjusting the traffic lights operation, preventing bottlenecks from developing.
- Monitor traffic conditions locally and send alarms to the local police departments when potentially dangerous driving behaviors are observed, i.e. blocked emergency or bicycle lanes, high speed driving or driving on bus lanes, red light or stop sign violations.
- Prioritize the movement of ambulances, fire engines and police vehicles in case of an emergency.
- Save time and reduce pollution from business-related transportation and promote productivity, as the delivery delays can be dramatically reduced through dynamic traffic re-routing.
- Minimize air pollution from the car exhausts and breaking by reducing journey times.
- Improve the quality of life for citizens and visitors by reducing congestion, optimizing traffic flows and reducing local air pollution.
Application #2: Public spaces monitoring
Cities are responsible for various types of public infrastructure, such as water treatment and distribution facilities, energy networks, telecommunication equipment, street lights, roads, tunnels and bridges. Consequently, local authorities are under immense pressure to maintain all critical assets in good working condition and to react in a timely fashion when faults or breakdowns occur. However, the inevitable multiplicity of infrastructure related incidents can stretch even the best resourced teams leading to a response which is often neither fast nor optimum.
On the other hand, if a cloud-based system could have access to the footage available from all the local CCTV networks, analyze it and automatically advise on prompt actions, city officials will be able to quickly take effective and efficient decisions. AI-enabled vision systems with edge processing can offer the above functionalities by:
- Quickly analysing the available visual data.
- Running AI algorithms to decode objects, people and movement.
- Using ML methods to predict dangerous situations, and
- Automatically inform the responsible personnel.
Application #3: Public health and safety in a pandemic era
As the Covid pandemic showed, there are times when city authorities need to respond to completely novel and unforeseen situations. Computer vision systems can aid public services (e.g., police stations, hospitals, water treatment facilities, traffic management control rooms) to adapt their operation to any new regulations, inform citizens appropriately, identify clusters of non-compliance and apply corrective actions where necessary.
For example, a health protocol breach in a public space can be identified, assessed and managed in a faster and more accurate manner, reducing the likelihood and impact of runaway risks to the local community.
Application #4: Waste dumping monitoring
Littering is the bane of modern cities. From the anarchic disposal of gardening waste, building materials and old furniture to the more dangerous dumping of electrical devices, broken machinery, used tyres, old batteries and chemical products, it’s impossible for the city authorities to maintain a near real-time monitoring and deterrence system.
By using street IP cameras with ECV capabilities, potential dumping spaces can be accurately monitored. Should the edge AI algorithms detect a suspicious activity, the following could help officials to be dully notified and decide on the actions they may wish to take:
- Recording of incident location, date and time.
- Photograph and video capturing of the potentially illegal activity.
- Vehicle plate recognition (if involved).
- Localised information processing.
- Automatic decision-making on whether an alarm should be raised.
- Prioritise alarms according to severity
Moreover, such data may help the authorities to take better decisions when planning the development of new waste facilities or identify which neighborhoods may benefit from a targeted waste and recycling information campaign. Ultimately, this results into cleaner streets, less air pollution, reduced health hazards and an improved management of municipal assets.
Application #5: Smart parking monitoring
Another application where the implementation of computer vision systems can help officials to offer better services is the management of parking spaces. In a busy city, a large chunk of the everyday traffic jams and the associated pollution is created by drivers who are trying to locate available parking spaces.
With the use of parking lot AI-cameras it is possible to detect the cars arriving or leaving a parking area, automatically identify plate numbers and record the parking duration. All the above information can be fed to a central, cloud-based database, enabling the city authorities to reduce parking surveillance costs and provide real-time information to drivers over mobile phones, pointing them to the closest empty parking space. This can save time, fuel costs and frustration, while the city roads will suffer less from congestion, noise and air pollution.
Furthermore, by applying machine learning algorithms to the collected real-time parking monitoring data, the authorities can make better use of parking spaces and plan more effectively when deciding upon future investment on public parking.
From the above, it is evident that the use of computer vision systems with AI/ML functionalities can significantly improve public services and the life of citizens and visitors in our cities. Either by running visual data processing entirely on a cloud-based IoT platform, or with complementary edge AI capabilities (ECVs), these methods reduce monitoring costs, whilst allowing for reliable predictive analysis and much faster decision making.
If you would like to find out how such a computer vision system could be implemented and configured to provide robust solutions to your city’s challenges, you may book a free consultation call with one of Yodiwo’s Smart City experts.