Searchable Video Intelligence Powered by Generative AI

23 October 23

6 mins read

5 Likes

Revolutionising Security: The Power of Generative AI in Video Analytics

In an age marked by escalating global security threats and a shortage of security personnel, video analytics (VA) has emerged as a vital tool to ensure public safety, secure critical infrastructures, and protect national assets.

According to Strategic Market Research, the VA market is expected to grow rapidly from US$5.54 billion in 2022, to US$35.6 billion by 2030. The Asia-Pacific region is expected to dominate with a growing CAGR of 28.5%.[1]

The increased deployment of surveillance cameras and the availability of high-resolution cameras across diverse industries has paved the way for more accurate and detailed video analysis. This surge has set the stage for real-time event detection, allowing for the identification of and alerts about vehicles, static objects, and people.

Powered by AI

In the area of VA, artificial intelligence (AI) has played a significant role in intelligence detection. However, the training of conventional AI models requires supervision, and they are often narrowly trained on pre-defined insights to identify specific instances, actions, or objects, such as identifying a fugitive’s facial features or markers of a traffic jam. This approach is not only costly and time-consuming, but lacks flexibility, as it is less adaptable to unforeseen situations.

The emergence of generative AI, with its ability to leverage both supervised and unsupervised learning, marks a revolutionary shift in VA. Unlike traditional models, generative AI identifies patterns and structures in data without extensive labelling, allowing users to run searches using natural language and obtain insights in seconds.

At the forefront of this innovation is ST Engineering’s AGIL® Vision, one of the first VA suites in Singapore to harness generative AI. Unveiled at the InnoTech Conference 2023 held in Singapore, AGIL® Vision has gained attention for its groundbreaking use of generative AI for searchable video intelligence.

AGIL® Vision addresses the limitations of traditional VA, allowing users to search vast video footage using natural language for generic fields like clothing colour, clothing brands, hair length, and even action or behaviour without the need for extensive pre-training of the system’s algorithms with many data samples.

The flexible search queries also extend to unattended bags, lost and found items, weapons, or even human behaviour, like the act of falling, fighting, running and more. This creates a robust tool that enables rapid and efficient investigation.

For example, simple queries like “a boy in a blue Nike shirt”, “a person climbing through a window”, or “a woman in a floral blouse” yield real-time results without the need for laborious manual video review.

This more natural and rapid search for information is made possible by AGIL® Vision’s deep learning model, which creates a semantic mapping between text—including colour, objects, activities, and behaviours—and visual representations, including logos.

The technology makes it possible for a more intuitive and quicker search of previously impossible search permutations, making AGIL® Vision an effective tool for any industry or organisation relying on video surveillance.

AGIL® Vision excels at locating lost individuals, spotting disputes or fire incidents, locating unattended baggage, or expediting healthcare responses in scenarios such as patient falls. Its ability to recognise suspects is achieved by combining information from additional sightings to create a comprehensive profile of the suspect’s actions on the property, including whether or not he met someone else or left a bag on the ground unattended.

Useful across industries

AGIL® Vision’s versatility extends across industries, simplifying and accelerating access to critical insights:

Public security and safety:
Faced with a wall of CCTV footage, operators can sometimes overlook anomalies. AGIL® Vision helps to secure critical infrastructure, manufacturing facilities, and warehouses against perimeter intrusion. Its automated detection ensures prompt identification of accidents or trespassing incidents.
Outdoor security surveillance and investigation:
The system can flag disturbing incidents such as accidents or fighting, locate missing persons, and identify wild animal sightings.
Condominium security and facility management:
AGIL® Vision can be a deterrent against premise intrusions, parcel theft, and inconsiderate behaviours, to bolster security measures in condominiums and for more effective facility management.
Construction workplace safety enforcement:
AGIL® Vision can help enforce perimeter security and safe work practices like the wearing of proper personal protective equipment such as safety helmets, vests, and boots. It can also deal with unexpected situations, such as locating misplaced equipment, identifying a flood situation, or supporting an injured worker.
Patient care and safety:
In healthcare settings, AGIL® Vision can detect instances of patients falling in hospitals or nursing homes. This enables a quick response, reducing the risk of fatalities. It also assists in monitoring patients with cognitive disorders, if they leave designated areas.
Student management:
AGIL® Vision can help to maintain discipline on school compounds by detecting misbehaviours like fighting, smoking, or vandalism.

Easy deployment

The solution’s easy deployment, offline functionality, and integration capabilities with existing infrastructure position AGIL® Vision as a hallmark in the VA landscape.

With a small form factor, AGIL® Vision fits seamlessly into security operations centres or any sector, regardless of size.

AGIL® Vision works offline and can be set up on-premise, without sending data externally and therefore, ensuring data privacy. It can also be deployed in the cloud, adhering to cybersecurity best practices. The integrated single-box solution facilitates plug-and-play connectivity with any camera and video management system, with minimal set-up required.

ST Engineering’s rich experience in implementing video analytics solutions positions us as a leader in optimising AGIL® Vision’s potential. As the capabilities of AGIL® Vision continue to unfold, it is clear that the solution transcends the current scope of use cases. With AGIL® Vision, the possibilities are limitless, marking a transformative leap in the area of security and surveillance.