AI Enhanced Security
Intelligent Detection Systems:
By Eric W. Muise, Key Account Manager at Advantech with D.C. Smalley, General Manager North America at Hailo
Over the past decade, due to global conflicts, mass shootings, and increasing political violence, there has been a growing need for intelligent weapon detection systems to be deployed to all venues where people gather, learn, and socialize.
These systems are no longer limited to being used at airports or transportation hubs where passenger safety is the prime concern.
Today, intelligent detection systems are critical in protecting schools, stadiums, nightclubs, shopping centers, hospitals, and any public venue where a mass group of people could be a target. And open-air celebrations like marathons, festivals, and town-square celebrations introduce unique challenges to protecting crowds, whereas the security apparatus must be purposely deployed quickly for these events for a short time and removed just as easily.
Security screening is seen as an extension of the consumer experience. After spending money on a ticketed event, a bogged-down, slow-moving security line can impact that patron’s experience and cause frustration at the venue operator.
Now, walk-through weapons screening systems must detect concealed weapons without impeding foot traffic flow while deterring attackers and enhancing safety in crowded areas.
AI-enhanced systems are the best alternative to traditional metal detectors and camera systems. Metal detectors can clog entrances with slow-moving queue lines. Traditional passive security cameras are ineffective as a preventive measure as they are not routinely monitored and are better served in evidence-gathering after an incident.
An essential appeal of AI-based systems is that they are non-intrusive, allowing people to enter public spaces without the need to empty pockets, remove belts and coats, check bags, and other unpleasant procedures. The result is a fast, accurate, and less intrusive weapons detection system. The primary goal of this technology has always been to identify and detect weapons as a person passes through a screening checkpoint. With artificial intelligence, the flow of patrons passing through these screeners can be accelerated and not cause congestion and crowding. The real-time data analyzed by these weapons detection systems must also be accurate and prevent the false positives that can block innocent concertgoers from reaching their seats in time for the show.
These high-speed screening systems are already used at several sports arenas, entertainment venues, and theme parks, screening hundreds of thousands of visitors each year for deadly weapons. The global concealed weapons detection systems market was estimated at $1 Billion in 2023 and is expected to grow to above $2 Billion by the 2030s.
Although the hardware for these walk-through systems has hardly changed, the emergence of artificial intelligence and machine learning at the software level has supercharged their adoption in our society.
These new software requirements mean that purpose-built hardware must evolve to consider the increased edge processing. Advantech provides edge AI modules, AI inference systems, edge intelligence servers, and IoT gateways to accelerate the development of these detection systems.
Advantech is at the forefront of providing Edge AI solutions, and the ARK, embedded box PC, is designed to fit into this rugged edge AI space. A new generation of ARK provides system performance ranging from entry-level Intel ATOM solutions for edge computing to high-performance Intel XEON-D to provide inference server capability and anything in between. For example, the ARK-1221L is a compact, robust PC platform based on the Intel Elkhart Lake ATOM processor. ARK-1221 also provides the necessary I/O and expansion capability for add-on features like Wi-Fi, additional digital I/O, and AI modules.
Last year, Advantech announced an exciting strategic partnership with Hailo. Through this collaboration, Advantech is leveraging Hailo’s AI accelerators to develop exceptional performance and price-per-watt edge AI systems with AI acceleration modules.
Hailo’s product portfolio includes the Hailo-8 AI accelerator, which seamlessly integrates with Advantech edge platforms, empowering real-time deep learning inference tasks at the edge with 26 TOPS at a low power consumption of 2.5W, small size, and low cost.
Tell me a little bit about Hailo - background and product portfolio.
D.C. Smalley: Hailo is a chipmaker whose AI processors were specifically designed to run complex Artificial Intelligence applications on edge devices. Our processors have extremely high AI computing capabilities, enabling high-performance AI to run with the low power consumption required for edge devices such as cameras, NVRs, drones, autonomous vehicles, robots, and more. Hailo is working with 100s of customers using the Hailo AI processors to empower their edge applications in various industries, including automotive, security, industry 4.0, healthcare, and retail.
What is an AI accelerator?
Smalley: Hailo’s product portfolio includes the Hailo-15 line of AI vision processors, which are SoCs for cameras, enabling both AI-powered video analytics and AI-powered video enhancement, and the Hailo-8 AI accelerators. AI accelerators work as co-processors, specifically handling the machine learning and video analytics tasks. Due to their unique dataflow architecture, which is the product of a rethinking of traditional computer architecture, Hailo-8 AI accelerators enable smart devices to perform sophisticated deep learning tasks and pipelines in real-time, with minimal power consumption of typically 2.5W, at a tiny size, and affordable cost.
How is Hailo addressing gun violence with considerations of latency, accuracy, and privacy?
Smalley: Hailo’s customers use its processors for time-critical and accuracy-sensitive tasks such as weapon detection. The high AI capacity of the Hailo chips enables intelligent cameras to identify and detect guns or other weapons in real time and alert first responders and security personnel. AI enables very accurate detection, minimizing error rates and false alarms so that when an alert is generated, it is treated with utmost urgency. Additionally, having the AI algorithms running on the images directly at the edge and not in the cloud makes it possible to anonymize metadata to protect people’s private information.
“ENABLES INTELLIGENT CAMERAS TO IDENTIFY AND DETECT GUNS OR OTHER WEAPONS IN REAL-TIME AND ALERT FIRST RESPONDERS AND SECURITY PERSONNEL”
What are the trends you are seeing - more powerful AI on the edge, LLMs, and Generative AI?
Smalley: In recent years, we have seen tremendous growth in the requirement for processing AI at the edge across all industries. As intelligent cameras and video analytics become more prevalent, the requirement is for more robust applications that require higher computing power and AI capacity. To this end, Hailo has enhanced its product portfolio with Century PCIe cards, which enable up to 208 Tera Operations Per Second (TOPS) for very high-performance applications. We also see a trend of moving generative AI and large language models (LLMs) to the edge. While such applications as Stable Diffusion and ChatGPT were so far confined to the cloud due to the high computing power they require, new developments both in models and infrastructure now enable these models to run at the edge, enabling their usage even without network connectivity or expensive subscription, and at maximum user privacy.
Describe some of the other applications.
Smalley: Hailo also has many customers in the security business, ranging from intelligent transportation to smart cities and large-scale video management systems. Other customers are developing vision-based, AI-powered solutions for various applications in smart retail, optical inspection and industrial automation, medical imaging, and more.
