Edge-Based Video Analytics vs. Server-Based Analytics: Key Differences and Applications
Over the past decade, the credibility and performance of video analytics have advanced to the point where the question of “Should video analytics be used?” is no longer asked. Instead, the current question is, “How to implement video analytics within the technical structure?”
When deploying video analytics software, there are two options available that can be effective depending on the environment. The software, which includes video analytics algorithms, can either reside in one of two broad spaces: within an IP network camera (referred to as Edge-based because the software operates at the network’s edge) or on a central server (referred to as Server-based). One branch of the Edge approach can be used for analog cameras, but this branch lacks the computing power to run embedded analytics. This can be achieved by placing video analytics software on video encoders that also reside at the network’s edge.
Both the Server-based and Edge-based approaches have their strengths and weaknesses. Understanding these differences can help installers recommend the best approach to their clients. In some projects, a combination of both approaches can yield better results.
This method can be specifically used in companies like electrical organizations or other large-scale operations, as well as many smaller, remote locations with limited bandwidth. The table below describes the decision-making criteria for implementing and using video analytics within cameras (Edge-based) or on servers (Server-based).
In addition to the summary of the provided indices, there are factors that installers need to pay attention to, such as how the analytics will be used and where the recorded videos will be stored. When using advanced video analytics, the storage location of videos becomes crucial, as these analytics rely on the recorded footage.
More on Advanced Video Analytics
In cases where simple video analytics are used, the associated software analyzes video, detects anomalies, and triggers alarms. On the other hand, advanced video analytics go a step further by recording a summary of what the analytics are examining. A good example of this is using video analytics to review recorded footage from the last 24 hours to display all instances where someone wearing a red jacket passed through a certain area. This capability can also be used to detect size, objects, speed, direction (orientation), aspect ratio, or vehicle license plate numbers. These more advanced analytics create hidden information (metadata) for future use.
To analyze this metadata as part of a comprehensive study, it must be stored along with the video on an NVR, and the VMS system should have an interface for reviewing it. This is why video analytics vendors and VMS vendors must work closely together. In fact, when using very complex video analytics, it’s not unusual to find all components offered as part of a solution from a single source.
What to Choose?
Using Edge-based analytics in remote locations with low bandwidth offers many advantages. These analytics can provide a cost-effective and highly efficient solution. The high CPU power of Server-based analytics allows for more analytics to be executed in each camera. Since these analytics can work across all cameras, the flexibility of these systems means you can choose the camera you want based on the desired performance and location requirements.
Decision Criteria Comparison Table:
| Criteria | Server-Based | Edge-Based |
|---|---|---|
| Reliability | Servers are high-performance, and the maximum number of cameras can be defined per server to avoid overload. Server-based solutions are generally more reliable unless video streams have low quality. | IP network cameras usually have lower processing power unless they have specific onboard processors. However, they have access to uncompressed raw video, which increases the reliability of analytics. Reliability issues arise if the onboard processor isn’t strong enough for heavy analytics. |
| Video Analytics | No processing power limitations, meaning multiple analytics can run simultaneously. | Most video analytics can run at the network edge, but due to limited processing power, cameras are typically restricted to running one analytical function at a time. |
| Quality | Centralized (Server-based) analytics are more suitable for distant locations with one or two cameras. | With Edge-based analytics, cameras can perform everything, including storage in onboard memory. This independence ensures that even if the network is weak or down, the analytics still function. |
| Bandwidth | Server-based analytics are used only when video streams meet minimum quality levels. | Edge-based analytics are ideal for low bandwidth as they process analytics inside the camera, and only a simple alarm is sent to the central server. |
| Camera Selection | Analytics work easily on all cameras, making it easy to select cameras from different manufacturers. | Manufacturers often provide analytics on only a subset of their camera models, limiting choices for the user. |
| Replacement | If the existing analytics don’t meet the client’s needs, software on the server can be replaced, ensuring compatibility with the VMS. | Most manufacturers use a single analytical software for all their cameras, so if the analytics don’t work as expected, clients may need to replace all cameras with another brand. |
| Ease of Installation | The software can be easily implemented across all devices, regardless of brand or model. | Camera manufacturers usually use a single analytical software across their camera range, so if different camera brands are used in a client’s surveillance system, the installer must be trained in configuring all types of cameras. |
| Carbon Footprint | Using a single server for analytics and video recording reduces energy consumption for video analytics. However, additional servers for analytics can consume thousands of watts of energy. | Edge cameras use very little additional energy for running video analytics. |
| Total Cost of Ownership | The total cost is usually higher if analytics run on separate servers. | For Edge-based analytics, the total cost of ownership is zero as the IP camera with SD card takes up no rack space, uses little power, and doesn’t burden the HVAC system. |
| Price | Server-based analytics usually have higher per-camera costs, focusing on reliability, energy efficiency, and feature sets. | Many IP camera manufacturers bundle analytics into their systems at a very low price to create differentiation and a competitive edge. |
| Licensing | Server-based analytics are typically licensed per camera but can be re-assigned to another camera if needed. | Cameras are often sold with activated analytics as part of the package or software that can be unlocked by paying for a license, specific to that camera. |
Conclusion
Edge-based video analytics offers numerous benefits, especially in installations that require processing video directly in the camera, reducing bandwidth usage and improving efficiency. While more complex video analytics that require higher processing power might be better suited for Server-based solutions, a hybrid approach combining both may yield the best results for certain installations.
As technology continues to evolve, manufacturers are making their cameras more open, enabling software vendors to upload analytics and applications directly onto the cameras. This shift is transforming the surveillance industry, offering more flexibility and better integration between video analytics and camera hardware.
Edge-based video analytics is increasingly seen as a compelling alternative to traditional Server-based solutions, especially in environments with limited infrastructure. This approach is particularly advantageous in transportation systems, where limited infrastructure makes centralized video streaming a less viable option.
Source: SecurityInfoWatch, 3VR Magazine