The application of AI across the CCTV sector is big business, but building networks of intelligence that enable rapid detection and response of illegal activity is no easy task.
SwissCognitive Guest Blogger: Eleanor Wright, COO at TelXAI – “AI And Video Surveillance”
So, let’s break down what it involves and more importantly the governance of data, integration into legacy systems, and the scalability of AI across CCTV networks. Identifying challenges and how they impact AI and video surveillance.
When you think of AI and CCTV you probably think of an Orwellian world of surveillance and you may not be wrong, London alone has approximately 785,100 CCTV cameras collecting and storing data.
But how well is this data utilised?
Firstly it’s important to note that many of these cameras are privately owned and those that aren’t, are owned by local councils, TFL, and the metropolitan police. This ownership is key, as those who own the cameras own the data and this data is crucial for building and implementing AI. This ownership, however, is governed by the SCC and ICO, which oversees and enforces UK data protection.
Thank you for reading this post, don't forget to subscribe to our AI NAVIGATOR!
Governance is crucial for accessing, cataloging, and implementing AI-driven CCTV systems. The storage, cataloging, and implementation of data is critical for training large AI systems capable of autonomously detecting illegal activity. Owned by government agencies and governed by the SCC, access to this data for commercial benefits is limited. Thus this limitation of access to data greatly restricts innovation and the training of AI-driven detection systems.
Secondly, integrating AI into legacy systems is key to the future of how smart these systems will be.
Legacy systems both enable and inhibit the deployment of AI systems. The implementation of AI-driven detection systems will only ever be as good as the sensors they are connected to. As the notion goes, garbage in garbage out; AI trained and deployed on garbage data will never result in accurate detection or reliable communications. Plugging in a well-trained and developed AI system into a network of historic or low-quality cameras will lead to nothing but a flood of false positives.
Finaly, applying AI at scale is another task within itself.
Scalability is critical; deploying technology developed in laboratories, into live environments is the first challenge, and scaling these systems across cities is an even larger challenge. When dealing with false positives, it is logical to expect the greater the number of sensors deployed the greater the amount of false positives will be generated. As such, the ability to filter out or reduce the number of false positives is critical to scaling large AI CCTV systems operating in real-time, across vast areas.
AI applied to physical security of the future however will not merely be limited to cameras; although they will most likely continue to be a key element. In addition to this, governance, integration into legacy systems, and scalability will continue to dictate the future of AI and CCTV. Whether it be implementing facial recognition, detecting suspicious activities, or enabling sensor fusion.
Where there are challenges, there will also be solutions. If data is not accessible, create it, if legacy technology is not up to the job, replace it, and if scalability is unmanageable, rethink the design underpinning the flow of information. There are many ways to better implement and utilise AI in physical security that goes beyond CCTV and video.
For example, audio sensors, CBRNE, access control systems, and LiDAR systems aid in detection. Combining online activity with physical sensors could enable the collection of critical data, at critical locations at the right time. Finally, these sensors have more to offer society than surveillance, in a world driving towards autonomy these sensors may be a key facilitator in the deployment of autonomous robots and vehicles.
About the Author:
Holding a BA in Marketing and an MSc in Business Management, Eleanor Wright has over eleven years of experience working in the surveillance sector across multiple business roles
The application of AI across the CCTV sector is big business, but building networks of intelligence that enable rapid detection and response of illegal activity is no easy task.
SwissCognitive Guest Blogger: Eleanor Wright, COO at TelXAI – “AI And Video Surveillance”
So, let’s break down what it involves and more importantly the governance of data, integration into legacy systems, and the scalability of AI across CCTV networks. Identifying challenges and how they impact AI and video surveillance.
When you think of AI and CCTV you probably think of an Orwellian world of surveillance and you may not be wrong, London alone has approximately 785,100 CCTV cameras collecting and storing data.
But how well is this data utilised?
Firstly it’s important to note that many of these cameras are privately owned and those that aren’t, are owned by local councils, TFL, and the metropolitan police. This ownership is key, as those who own the cameras own the data and this data is crucial for building and implementing AI. This ownership, however, is governed by the SCC and ICO, which oversees and enforces UK data protection.
Thank you for reading this post, don't forget to subscribe to our AI NAVIGATOR!
Governance is crucial for accessing, cataloging, and implementing AI-driven CCTV systems. The storage, cataloging, and implementation of data is critical for training large AI systems capable of autonomously detecting illegal activity. Owned by government agencies and governed by the SCC, access to this data for commercial benefits is limited. Thus this limitation of access to data greatly restricts innovation and the training of AI-driven detection systems.
Secondly, integrating AI into legacy systems is key to the future of how smart these systems will be.
Legacy systems both enable and inhibit the deployment of AI systems. The implementation of AI-driven detection systems will only ever be as good as the sensors they are connected to. As the notion goes, garbage in garbage out; AI trained and deployed on garbage data will never result in accurate detection or reliable communications. Plugging in a well-trained and developed AI system into a network of historic or low-quality cameras will lead to nothing but a flood of false positives.
Finaly, applying AI at scale is another task within itself.
Scalability is critical; deploying technology developed in laboratories, into live environments is the first challenge, and scaling these systems across cities is an even larger challenge. When dealing with false positives, it is logical to expect the greater the number of sensors deployed the greater the amount of false positives will be generated. As such, the ability to filter out or reduce the number of false positives is critical to scaling large AI CCTV systems operating in real-time, across vast areas.
AI applied to physical security of the future however will not merely be limited to cameras; although they will most likely continue to be a key element. In addition to this, governance, integration into legacy systems, and scalability will continue to dictate the future of AI and CCTV. Whether it be implementing facial recognition, detecting suspicious activities, or enabling sensor fusion.
Where there are challenges, there will also be solutions. If data is not accessible, create it, if legacy technology is not up to the job, replace it, and if scalability is unmanageable, rethink the design underpinning the flow of information. There are many ways to better implement and utilise AI in physical security that goes beyond CCTV and video.
For example, audio sensors, CBRNE, access control systems, and LiDAR systems aid in detection. Combining online activity with physical sensors could enable the collection of critical data, at critical locations at the right time. Finally, these sensors have more to offer society than surveillance, in a world driving towards autonomy these sensors may be a key facilitator in the deployment of autonomous robots and vehicles.
About the Author:
Share this: