2024 won't be business as usual; the landscape is rapidly evolving, revealing an intriguing future. The security industry is undergoing a remarkable transformation in video technology, driven by the increasing application of artificial intelligence (AI).

In this article, Rahul Yadav, Chief Technology Officer for Milestone Systems, explores this AI-driven future where, instead of humans watching video, software does the watching, and humans make decisions. These trends are enhancing operational performance and opening new opportunities in the sector. Journey with us as we delve into four AI-driven trends and explore how the symbiotic relationship between AI and human oversight will redefine safety and security solutions in 2024.

Data-driven video technology

The security industry experienced an accelerated impact from AI in 2023, specifically through the application of computer vision techniques to video technology in surveillance applications. As a result, the industry's trajectory in 2024 is overwhelmingly focused on data-driven video technology.

Data-driven video technology uses AI to combine video data with other types of data and extract actionable insights. This is disrupting the security industry, but it’s not removing people from the solution, it’s putting people at the centre of the solution. Software is now the tool that identifies objects, recognizes patterns and generates actionable insights from video data. People act as human-in-the-loop, using their intuition and judgment to verify the insights and make informed decisions.

This is driving a strategic shift in video surveillance, moving beyond passive observation, it is evolving into a proactive tool for intelligent action. Data-driven video technology encompasses several AI-driven trends that are creating new and potentially valuable opportunities both within security and beyond security. We will look at four of these trends in more detail.

Trend 1: Game-changing video analytics software

Basic video analytics, such as object detection and counting in a box, are already extensively employed in safety and security applications. To envision the future of security, we can draw inspiration from self-driving cars. These vehicles are already leveraging advanced video analytics to identify and track objects, even predicting how to evade them, all in real-time.

Affordable computing power is paving the way for these advanced video analytics with detection, tracking and prediction to enter the security industry. Some of these, although still in development stages, are becoming available now, while others go beyond anything we expect to see in applications any time soon.

By extracting contextual information from video data, these advanced techniques can interpret what’s happening in a video scene (a series of frames) and use this to generate actionable insights for humans. Here are some of the techniques that will be game-changers for the security industry:

Segmentation: enhancing our comprehension of scene dynamics, providing a sophisticated understanding of the unfolding events.

Recognition combined with image enhancement: improving the quality and resolution of video recordings, making it possible to identify objects and behaviour, such as walking, jogging, and running.

Detecting human interactions: recognizing and understanding the intricate ways in which humans interact with one another and their surroundings.

Anomaly detection: empowering humans to make informed decisions about highlighted incidents.

Prediction: looking to the future, the rapid advancements in large vision and language models (LVM) hold immense potential to enhance operational performance in the field of security. Moreover, the introduction of generative AI will facilitate a deeper understanding of humans by providing detailed textual descriptions of objects, their behaviour, and their interactions. Keep an eye on this space for exciting developments!

Incorporating a human-in-the-loop is vital for the successful implementation of these advanced techniques. While future video analytics software will have the ability to detect and alert for specific behaviours, it is the human operators who ultimately review the video recordings and make informed decisions regarding necessary actions. This process provides valuable feedback, allowing the software to continuously enhance its capabilities with each input. As the software receives more feedback, it becomes smarter at making accurate predictions, ultimately leading to improved performance.

Trend 2: Synthetic Data

To interpret video scenes accurately, video analytics software requires large amounts of accurately labelled training data. However, if the data has poor labelling or limited scope, such as portraying all people as walking, and no examples of people in wheelchairs, then the data is biased. Software trained on such biased data will not only inherit the bias, resulting in less effective solutions but also produce solutions that are less ethical.

Synthetic Data, which is artificially generated rather than sourced from the real world, holds great promise in addressing bias issues. By introducing diversity into training data, Synthetic Data effectively mitigates bias. It also provides the added advantage of precise labelling from its inception, eliminating any inaccuracies that may result from human error in manual labelling. Additionally, it safeguards individuals' privacy and avoids consent-related concerns that arise from utilizing real consumer information without permission or compensation.

Trend 3: Edge AI

In 2024, we will see a major acceleration in AI development on the edge (AI in devices like cameras and sensors). Until now, AI tasks were processed either in the cloud or in a limited way on local devices, but now there's a middle ground. Thanks to Nvidia and Intel, two key trends have emerged.

First, the edge is becoming more capable of handling AI tasks independently, reducing reliance on cloud resources. This enables faster and more efficient processing because AI-driven applications can operate closer to the data source. Today, there are many devices at the edge, like smart cameras and IoT devices, that can analyze and respond to data in real-time.

Second, having AI at the edge is cost-efficient. It reduces reliance on cloud resources, saving bandwidth costs and reducing latency. This is especially beneficial for security tasks that require real-time monitoring. The cost-efficiency of edge AI is making it an attractive option for the security industry.

The combination of enhanced capabilities and cost-efficiency makes edge AI a compelling security solution for the future. In 2024, we can expect further advancements in edge AI, unlocking more sophisticated applications in devices.

Trend 4: Responsible Technology

As AI drives the shift in video surveillance from observation to action, Responsible Technology is emerging as a prominent trend in 2024 and beyond. Future generations are watching how tech companies will approach AI in a responsible way. For them, innovation is no longer solely about who can innovate the fastest; but who can innovate responsibly. Consequently, tech companies must integrate Responsible Technology principles into the way they develop, the way they sell and the way their customers use their technology.

This trend was revealed in a 2023 global survey of 150 technology decision-makers that revealed their intention to exclude potential vendors based on their approach to technology usage. The majority (85%) of technology buyers expect responsible use of AI, video analytics, and video surveillance to be a prerequisite for engaging with tech vendors in the future.

The survey highlights Responsible Technology as a key priority for decision-makers and an essential business requirement. In the next three to five years, Responsible Technology will become a license to operate.

Be ready for the future

While data-driven video technology will continue to shape the roadmap for the security industry, it is not eliminating the human factor; rather, it is putting people at the centre of the solution. Software now serves as the peripheral tool for monitoring, analyzing, and understanding video scenes. At the center, people play a vital role as human-in-the-loop verifying analysis and making informed decisions.

The human element is crucial for the intelligence of the AI-driven future because it depends on high-quality feedback for learning. Human oversight and expertise maximize the value of AI-driven security solutions, helping to foster a safer world.

The future of video surveillance, shaped by AI and human expertise, is set to revolutionize safety and security. Embrace data-driven video technology and the strategic shift it is driving, and make sure you are ready for these four AI-driven trends coming your way in 2024.