Smart city, smart policing and AI: Going hand-in-hand
Source: William Pao
Any smart city initiative begins with making the city safe and secure. Luckily, video analytics and AI can help in this regard. This note examines how.
Smart city has become a hot topic. Needless to say, to make cities smarter and safer, administrators must deal with crimes, which are serious municipal issues. According to the Council on Criminal Justice, the number of murders in US cities in 2021 was 5 percent higher than in 2020 and 44 percent greater than in 2019. Fighting crimes therefore has become a critical component in any smart city initiative.
Challenges and pain points
Yet law enforcement officers are often met with certain challenges and difficulties, chief among them a lack of data to work with or act upon. “Based on our experiences throughout Southeast Asia, Mexico, South Africa and back home in the Unites States, it is the lack of accurate information in the most crucial time to best take action. Even today agencies seem to be acting on after-the-fact data rather than able to make real-time decisions. The challenge now is to find ways to improve existing support technology and infrastructure,” said David Ly, CEO of Iveda.
“Working out how to prioritize and manage all crimes/events in a city, including investigations and predicting crimes, in an efficient way can be difficult. In many situations it is already too late when the officers arrive at the scene,” said Andreas Göransson, Global Marketing Manager at Axis Communications.
Smartening up policing with AI
To be more responsive and situationally aware, authorities often turn to video surveillance, which has long played a critical role in law enforcement. “Video surveillance can play a significant role in crime prevention when combined with other crime deterring methods, and it can be pivotal for a criminal investigation or in response to an active crime scene. It provides crucial real-time situational intelligence to first responders, like the number of shooters and firearms in the case of an active shooting scene, and unbiased evidence to detectives after the fact,” said Augusto Chiaravalloti, Industry Marketing Manager at Genetec.
Yet video alone has its limitations. This is where policing can be aided by advanced analytics, powered by deep learning and AI, which can generate alerts based on abnormalities detected – loitering, loud noises, gunshots and double parking, among others. Officers on patrol nearby can be instantly alerted and go to the scene.
“Video alone can only tell you what had already occurred if you know where to look in the first place. Knowing what had gone wrong will always allow an organization to better mitigate their risks in the future but may not help in prevention. The good news is for those who already use video surveillance system, there is a way to immediately enhance its effectiveness,” Ly said. “AI in video surveillance can now truly allow for all the value and functionality expected from a traditional system. This benefits law enforcement and security personnel by providing more accurate alerts and meaningful data where better decisions could be made quicker. AI can also recognize past problems, then help us avoid them all together. That’s prevention at its best.”
“Early pixel-based analytics offerings were used in fighting and solving crime with mixed results. Many solutions in the market today still use this older technology. With the advent of AI, solutions have matured to the point where police experience good positive detections without the many false detections that made earlier analytics unusable,” said AJ Frazer, Chief Revenue Officer at Irisity.
“AI – and more specifically deep learning – can automatically generate alerts based on specific events in any scene. In combination with more accurate analytics and data, this helps operators make informed decisions based on location, time, and environment. Tens and even hundreds of cameras can be added without the need for more operators,” Göransson said. “With video analytics, law enforcement can even follow a specific vehicle on footage or monitor a person’s dwell time at a particular location. This is vital to fighting and solving crime today.”
“As an example,” Göransson added, “knowing who the drug dealers are in a specific neighborhood isn’t always sufficient to make an arrest. Proof is required. By combining heat mapping and dwell time analysis, it’s possible to see where a dealer hangs out most often, and when they are the most active. With technology pointing the way, officers can catch them in the act and use the recorded video as evidence.”
Edge or server
AI and analytics software require large computing power to run; therefore they usually reside on servers. Yet more and more, cameras have the capacity to run these programs, and this has certain advantages for law enforcement authorities.
“Analytics taking place within the camera means that only the valuable data needs to be transferred to the operator. Conversely, analytics taking place on the server requires that all of the data from the camera is transferred to the data center for analysis, and with that comes a much greater need for costly bandwidth,” Göransson said. “Secondly, analyzing video within the camera and as close as possible to its capture means that the images being reviewed are of the highest possible quality: there is no degradation that can come through compressing images prior to transfer.”
But ultimately, the architecture that the law enforcement end user chooses depends on their own requirements and use cases.
“The use of analytics on the edge and on-prem or in the cloud, all have their benefits and uses. Using edge analytics allows you to reduce the workload on a central computer as well as provide more real-time options but provide a higher per-unit camera cost. On-prem or in-cloud analytics could increase the overall requirements for server/cloud compute power but may be more beneficial when using analytics forensically (after-the-fact) on select or as needed cameras,” Chiaravalloti said.
“The decision for edge vs. server vs. cloud is influenced by many factors. Some customers use a combination of the three. The main factors are computing power, maintainability, and IT policies. The best camera OEM’s are just beginning to include enough processing power in their cameras to run the latest AI analytics. But most customers already have many older cameras deployed. Server or cloud is the only option in this case. Some IT department prefer server based AI since it’s easier to cyber-secure and maintain. Some IT departments prefers cloud, while other reject it on principle,” Frazer said.