How to select the best behavioral recognition video analytics system

December 19, 2021 9:51 am

With AI-powered behavioral analytics smarter than ever, identifying the best security system requires thorough research and expertise.




Modern video systems produce vast amounts of content every second. Interpreting all that footage in real-time in order to react to potential threats is a daunting task for security and safety professionals. Advanced behavioral recognition video analytics can fill in the gaps, utilizing commodity hardware to analyze multiple real-time streams of footage at remarkable speeds.

Thanks to advancements in artificial intelligence, behavioral recognition video analytics are becoming even more insightful with the ability to learn, detect and classify abnormal behaviors automatically to stop potentially dangerous events from occurring or escalating.

The power of video understanding

AI is capable not only of detecting a crowd, but also whether a group of people are loitering or rioting, or that if a person has fallen to the ground or is tying their shoelace. The same holds true for vehicles that are stopped on a road or that have collided, or if two people are hugging to interact in an anti-social behavioral. The ability to interpret abnormal behaviors results in faster reaction times, such as initiating an automatic lockdown of a campus the moment a weapon is detected.

Wide range of scenarios

In general, advanced behavioral video analytics can be classified into one of five core functional categories: violence recognition, suspicious activity, crowd behavior, traffic monitoring, and environmental and personal safety, such as fire or hazards. The first step in determining whether behavioral analytics are the right enhancement for your system involves performing a thorough assessment of your organization’s needs and pain points. Are you protecting an office building, warehouse, airport, or a college campus? Monitoring a manufacturing site or a football stadium? In the world of video security and safety cameras, it’s all about context.




Working with a professional consultant
This is where an experienced system integrator (SI) or value added distributer (VAD) can be a tremendous resource. These professionals can help you identify the challenges facing your organization and whether visual analysis can provide a solution. What are some of your greatest concerns? Building security? Traffic flow? Business Intelligence? Crowd management? Weapons detection? Bullying? Process Management? Be as detailed as possible to determine the best suite of integrated solutions and intelligent analytics you need most.


Prioritizing your list
After an initial assessment has been completed, look to prioritize your security and safety challenges in a matrix. This will help determine the analytics capabilities and level of performance you need. Another point to determine is the range level of accuracy a behavioral recognition video analytics solution requires to delivers, also known as the recall rate. For example, if a video clip contains a scene with five vehicles, how many cars will the analytics detect? Most importantly is the estimate the false positive rate , meaning how noisy the system is,  your control command center will be able to handle, Again, this is where an experienced analytics provider, SI or VAD can help explain the nuances of the technology.


Preparing a document of requirements

Be prepared to create a document detailing your analytics functionally and technical requirements. The document should include a description of the current systems and technology you use and the characteristics you would like to see in a new or upgraded installation once an analytics solution is added.

When your document is complete, research the market and identify three behavioral recognition video analytics vendors from which you would like to receive proposals. Be rigorous in your analysis. Is a potential vendor a start-up or a well-established software developer? Are they financially stable? What’s their track record for customer service? Do they have credible references for similar projects, vertical market experience and/or installations? When you invite vendors to submit proposals, you should expect them to initiate a detailed analysis of your organization, including interviews with key stakeholders and in-house integration specialists, if they exist. If they don’t, consider it a red flag.


Evaluate the technology behind the analytics

Conventional  video analytics employ legacy pixel technology to recognize static objects and classify objects. More advanced behavioral analytics employ Artificial Intelligence (AI) to autonomously detect, analyze, and understand actions and events (dynamic evolving situation). This higher level of video understanding employs deep neural networks capable of analyzing and defining specific activities using video streams from conventional surveillance cameras.

More advanced analytics built this way can recognize the behavior of diverse objects and their relevant interaction and contexts. For example, an individual moving back and forth in a predefined area might be flagged as suspicious versa same individual that is engaged in some activity which considered valid. Advanced behavioral analytics will automatically distinguish between the two situations – without any manual setup or calibration.


This ability to distinguish a potentially dangerous situation from a normal occurrence provides security and safety professionals with the unique ability to initiate fast and appropriate responses to minimize or even prevent incidents from occurring.


A quick checklist of attributes to look for when selecting a video analytics solution includes the ability to…

  • Autonomously detect events of interest such as violent and suspicious activity, crowd behavior, perimeter (protection) violations, public health violations (e.g. COVID19), traffic congestion and accidents, environmental and personal safety threats, and occupancy and mobility counting.
  • Deliver implementation versatility in the Cloud or locally for one, some, or all networked video cameras.
  • Reduce or eliminate false alarms using a sequence of images versus a single image analysis
  • Eliminate the need for personnel to monitor large numbers of cameras simultaneously, effectively reducing operating costs associated with monitoring large surveillance systems and responding to false alarms.
  • Improve SOC/security personnel performance by allowing them to focus on real events of interest and provide additional services.



Privacy protection is critical criteria

The heightened level of situational awareness and risk mitigation provided by advanced behavioral recognition video analytics is further complemented by the need to ensure personal privacy protections. This should be accomplished by analyzing the general behavior patterns of individuals, groups, vehicles, and traffic-flows without identifying faces or license plates or any other biometric information. For high-security or law enforcement applications that require identification and authentication functionality, you should also look for a behavioral analytics solution that can be integrated into systems employing face and license plate recognition solutions.


Determining total cost of ownership

Once you identify your ideal solution, your analytics provider can help you to determine the overall cost of purchasing and operating your surveillance system, also called the total cost of ownership (TCO). The TCO should not only include the yearly software licensing fees or the initial perpetual licensing costs , but the cost of hardware required and  software maintenance and support.


Proof of concept is essential
When you finally select the preferred source for your new or enhanced solution, the selection process isn’t technically over. It is prudent to have the selected vendor provide a proof of concept to ensure the solution works as intended. This can involve a partial or light integration of perhaps of the total deployment. Essentially, you will put this scaled down version of the system to the test for a month or so. This is the only way to ensure the solution can meet your objectives.


The decision to employ intelligent behavioral analytics in a new or upgraded system requires a good deal of due diligence to identify existing and future needs, the ability to leverage existing and/or future investments in video systems, and the efficacy of the solution. Look for suppliers with the best combination of proven technology, experience in your specific business/vertical segment, and solutions that provide both versatility and scalability for future growth and changes. This may require some additional homework and discussion early on, but it will pay dividends down the road when you have the best possible solution up and running.

Blog Posts