AI video surveillance is continuing to change the face of the security industry, reports Seagate.
In the UK, someone is reported missing every 90 seconds. Tracking down these individuals involves deploying millions of operational cameras and thousands of security personnel are often involved in monitoring the footage.
This requires a lot of storage space for the data, a lot of money and a lot of people-power.
Unsurprisingly, artificial intelligence (AI) is changing the face of video surveillance. By 2028, the total global edge computing infrastructure will be worth more than $800 billion and by next year, it’s estimated that the AI-based surveillance and security market will be worth over $4.46 billion.
‘AI’ is an umbrella term that encompasses different levels of technology. We speak of AI when a program can sense, reason, act and adapt. Machine learning is a subset of AI technology that can automatically learn and improve based on experience. Deep learning is another subset that mimics the neurons in the human brain and is capable of learning from unstructured data.
Finally, pattern recognition is the lowest tier of AI and the technology we see most often in surveillance applications. As a branch of machine learning, it aims to recognise patterns and regularities in data.
Strengthening security teams
Why the change? AI is cheaper, less labour intensive and less time consuming than traditional video surveillance systems (VSS). Ultimately, AI isn’t prone to human error.
In traditional systems, security guards are obliged to watch countless hours of footage on multiple screens. Understandably, this is tiring. A study by 3S Security Systems found that staff experience fatigue in as little as 12 minutes and overlook up to 45% of activity. After 22 minutes, that number increases to 95%.
This leads to higher insurance premiums. Worse still, fewer crimes are solved. In a cost-benefit analysis, traditional systems look dire, thus requiring companies to pay employees to perform easily automated tasks.
In comparison, AI VSS boasts a benchmark of 98% accuracy. It saves time and improves efficiency, liberating operators from the necessity of watching multiple screens at any given time. Also, operators needn’t search through hours of footage – by learning what is normal (and what isn’t), AI video surveillance can provide critical highlights for a team of security professionals.
AI offers endless possibilities for video surveillance
It doesn’t stop there. AI-enabled video surveillance has revolutionised how we use video surveillance; far from merely helping security operators catch criminals, video analytics technology allows companies to better understand patterns of human behaviour, criminal and otherwise.
Video analytics can help businesses understand their customers, for example, by learning how long they spend at a particular aisle. In turn, this allows them to redesign floor strategies and avoid long checkout lines.
Moreover, in warehouses and factories, metadata allows managers to view footage from their smartphones, enabling them to reduce the number of staff in control rooms and redeploy them to more value-added tasks.
Smart cities use video surveillance to monitor busy roads and intersections, helping to reduce congestion and traffic accidents as well as identify stolen vehicles through automatic number plate recognition (ANPR). Singapore is a case in point; their top of the range Polycom security solution has evolved over the years to incorporate video analytics technology.
This isn’t all though. Over the last two years, the healthcare sector has used video analytics to reduce the spread of COVID-19. AI-enabled video analytics allow staff to reduce overcrowding in busy cafeterias and lobbies and check whether individuals entering the hospital are wearing a mask. Thermal imaging and body temperature detection cameras in public spaces can also be deployed for similar reasons.
In China, Japan and New Zealand, casinos are now using AI-based facial recognition systems to combat fraud and cheating and even tackle gambling addictions.
Let’s not forget access control and monitoring in multiple occupancy buildings, education campuses, hospitals and prisons etc. – the possibilities for new applications are simply endless.
How does AI increase pressure on surveillance data storage?
There’s a huge demand, and as this demand grows, the demand for huge storage capabilities grows too. Research conducted by IDC (Worldwide Hard Disk Drive Forecast, 2021–2025) forecasts that compound growth for surveillance HDDs will grow at a rate of 8% through 2025, coupled with a 19% growth in petabyte shipments.
As government regulations across the world increasingly demand longer surveillance video retention times, the need for a fast, durable, stable and retentive solution is becoming more pressing than ever before. After all, where will all this data go, and how can we ensure that its original quality remains?
Meanwhile, more and more businesses are moving to VSaaS (video storage as-a-service), which means that these data pressures will continue to grow at an exponential rate. Organisations will require huge amounts of storage space coupled with high bandwidth, a capacity for long term data retention and super-fast access to imagery without dropping frames.
Traditional video streams are written in large sequential blocks. From a workload per block perspective, this is fairly straightforward for a hard drive. However, the workload increases due to the amount of ‘large blocks’ coming in from high-resolution and multiple cameras.
When AI-enabled cameras capture an image from live video, this image event is then compared to a library in the network video recorder (NVR). NVRs are configured to receive video with AI data (image and metadata) on a separate simultaneous channel.
Because of the AI data, the event will have two times the size of an event without AI. In other words, capturing more AI events per minute will have an exponential demand on storage workload and capacity.
To make this clearer, let’s look at a simplified example. Imagine that it takes you five minutes to write this article while someone is dictating it to you. You are then asked to write the text again (still dictated) within five minutes, but this time you need to tally up how many times you see the letter ‘T’. If you want to stay within the five-minute time frame, you will have to increase your
work rate.
To go another step further, you now need to tally up how many times you hear the letter ‘T’, the letter ‘A’, and the letter ‘S’. Needless to say, your work rate will increase exponentially if you have to respect the allocated five minutes.
Writing this dictated article is quite similar to recording a sequential video. And you can compare counting and making a tally of the letters to capturing AI events. Finally, the tally of the letters should correspond to the moment you heard them, which is the metadata stored on the drive.
A future-proof storage solution for video surveillance
AI applications in video inherently increase the work rate and required capacity of the storage used. Hard drives designed for desktop applications (eight hours per day x five days per week) are not suitable for basic video usage. Video drives may struggle over time in AI applications if the work rate increases due to the random writing of metadata and images.
In response to this increasing use of AI in video surveillance, storage manufacturers have designed firmware customised for AI with the sort of features not available in the more traditional, desktop drives.
The blend of AI firmware and specific hardware enhancements helps to create drives perfected for NVRs, AI edge security and machine learning applications.
Seagate’s SkyHawk AI 20TB is the latest, highest capacity, most up-to-date solution on the market — the only product that tackles these issues head-on. The drive has been perfected for NVRs and AI applications and its round-the-clock durability, five-year warranty and high mean time between failures rate (MTBF) make it a future-proof solution.
It is designed for 64 HD cameras and 32 additional AI streams, with a capacity of up to 20 TB. Importantly, it supports heavier workloads associated with AI applications and includes ImagePerfect AI firmware to enable zero dropped frames.
AI has rewired our global economy and society and organisations are increasingly using AI to solve crimes and understand human behaviour. The future for AI in video surveillance looks bright — it would be a setback if a data avalanche were to overwhelm an organisation that does not have a scalable, future-proof data solution.
Seagate’s SkyHawk AI can weather the storm. Based on more than 40 years of experience with storage solutions, they are built to run 24 hours a day and support multiple full high-definition cameras at the same time.
To find out more information, please visit: www.seagate.com
This article was originally published in the July 2022 edition of Security Journal UK. To read your FREE digital edition, click here.