For those with legacy, analog video surveillance systems, switching to network-based AI cameras without changing the existing infrastructure is feasible, albeit with important caveats.
Increasingly, video surveillance systems are enabled with intelligence and AI. In an AI-enabled video surveillance system, AI cameras play a key role. For those with legacy, analog video surveillance systems, switching to network-based AI cameras without changing the existing infrastructure is feasible, albeit with important caveats.
Benefits of AI-enabled video surveillance
More and more, video surveillance is accelerating towards AI. The benefits are manifold. With AI, video is no longer just a passive surveillance tool, used for investigative purposes after something happens. Rather, it can now detect and classify objects and proactively tell users something is about to happen.
“More organizations are seeking to implement AI video surveillance systems. The driving force behind AI adoption is the advanced features, primarily algorithms facilitating proactive operational readiness compared with legacy surveillance systems that are often used passively to investigate reported incidents. AI benefits include actionable intelligence, real-time situational awareness, false alarm reduction, weapons detections, people counting, and tailgating detection, person’s down,” said Cathal Walsh, VP of Physical Security and Chief Security Officer at Guidepost Solutions.
“AI adoption is being driven by several factors, including the need to detect threats in real time rather than reviewing incidents after they occur; ongoing labor shortages and rising security staffing costs; the growing demand for proactive rather than reactive security; and increased concerns about workplace, school, healthcare, and public safety,” said Bob Mesnik, President of Kintronics. “Healthcare, education, retail, transportation, manufacturing, and critical infrastructure are among the industries experiencing the fastest adoption because they benefit significantly from automated threat detection and operational efficiency.”
AI cameras and video analytics
There are two approaches to adding intelligence to video surveillance systems. One is the use of AI cameras where most of the intelligence, for example object detection and classification, occurs.
“AI cameras contain powerful processors capable of analyzing video directly within the camera. Manufacturers such as Hanwha offer cameras that can distinguish between people, vehicles, and animals while generating valuable metadata, including object attributes, movement, and direction. This metadata is then passed to the video management system (VMS), allowing operators to perform more intelligent searches, trigger automated responses, and improve overall situational awareness,” Mesnik said.
Another approach is the use of powerful AI software/analytics at the backend.
“AI analytics software … enhances standard IP cameras without requiring intelligent cameras at every location. Solutions such as Scy-AI can detect firearms, recognize suspicious behavior, identify faces within crowds, detect slip-and-fall incidents, and even recognize aggressive or violent behavior. Because the AI processing occurs on the server, organizations can often upgrade existing IP surveillance systems without replacing every camera,” Mesnik said.
It should be noted that theoretically, AI analytics can work with legacy analog cameras. But ideally, these analytics should work with IP cameras to get the best results.
“Many AI companies offer agnostic solutions even for legacy analog cameras, meaning they can leverage existing video surveillance system cameras, which reduces complexity and lowers cost. But AI solutions work best with IP cameras,” Walsh said.
“Most modern AI analytics platforms are designed to work with digital IP cameras rather than analog video systems. Although analog cameras can be connected to an IP network through video encoders, the image quality from older analog cameras is usually insufficient for advanced AI analytics such as facial recognition, firearm detection, or behavioral analysis,” Mesnik said. “For this reason, upgrading analog cameras to AI capability is often less practical than replacing them with modern IP cameras.”
Migration paths
For legacy users seeking to migrate to a smart, AI-enabled video surveillance system, there are several migration options depending on the user’s existing infrastructure.
For analog users seeking to replace their old analog cameras with AI cameras while keeping the analog infrastructure and coaxial cable in place, this is theoretically feasible. What they need to do is connect the AI network cameras with Ethernet-over-coaxial (EoC) adapters which output signals that can be transmitted over coaxial cables. Metadata generated by AI cameras, for example object types, location of the video and other key information in the frame, can also be carried over EoC. An EoC adapter should also be deployed at the other end of the system to convert the video back to NVR-readable format. That said, the legacy DVR should be removed as it does not recognize IP signals; they should be replaced with NVRs.
But then again, for analog users seeking to deploy IP AI cameras, doing a rip-and-replace upgrade, where the entire analog infrastructure is replaced with an IP one, would result in a much cleaner system without converters or adapters. An IP architecture would also allow better integration with other networked security devices such as IP speakers or IP alarms. Ultimately, the key lies in whether rip-and-replace benefits outweigh the investment. If the legacy coaxial cable can be readily replaced with Ethernet without too much money or energy, then going all-IP would be ideal. If the coaxial cables are buried deep inside the walls and require tearing down or reconstructing the building to replace, then IP-over-coaxial is the better option.
For those whose infrastructure is already IP, then retrofitting the system with AI-enabled video surveillance is less complicated.
“If the existing surveillance system already uses IP cameras, organizations can often add substantial AI capability without replacing the entire system. Many AI analytics platforms work with existing IP cameras as long as those cameras provide sufficient image quality and resolution,” Mesnik said. “Additional AI software or an AI server can add capabilities such as facial recognition, person and vehicle detection, license plate recognition, intrusion detection, firearm detection, suspicious behavior analysis, slip-and-fall detection and crowd analytics. This allows organizations to modernize their surveillance systems incrementally rather than replacing every camera at once.”
According to Mesnik, a phased migration often provides the best balance between cost and performance.
“The first step is identifying locations where AI delivers the greatest value, including building entrances, school entrances, lobbies, parking lots, loading docks and high-security areas. For example, a school may only need AI analytics on entrance cameras to detect individuals carrying firearms or to identify suspicious behavior before someone enters the building. As budgets permit, additional cameras and locations can gradually be upgraded,” he said.
Product Adopted:Surveillance Cameras