Harnessing AI for Security Screening in Public Venues

Region: USA

Background 

Intelligent detection systems are essential for securing public venues such as schools, stadiums, shopping centers, and hospitals with large crowds. The global concealed weapons detection market, valued at $1 billion in 2023, is projected to exceed $2 billion by 2030. Traditional security measures, such as passive cameras and metal detectors, have limitations, including slow queues and reliance on post-incident evidence. Crowded environments and varied contraband further complicate inspections. To address these challenges, the customer adopted an AI-powered security screening solution to enhance efficiency, reduce wait times, and improve user experience.

Project Challenges

Developing AI-powered security screening faces challenges, especially for open-air events like marathons and festivals, which require cost-effective, easily deployable, and seamlessly integrated security systems to protect crowds during short durations. For AI security screening applications, a hardware system capable of enhanced edge processing and equipped with versatile industrial I/Os for connecting equipment and sensors is essential. Beyond hardware configuration, developers must pre-train a model for object detection, making the choice of a suitable model for real-time security screening a critical consideration. Even after a pre-trained model is ready, it often needs to be converted into a specific format compatible with the target AI processor or GPU. Finally, ensuring the seamless deployment of the converted model onto the AI system is a crucial step. 

•　Systems offering LAN connectivity for cameras and DIO functionality for abnormal alarms or data collection. 

•　Easy integration of AI into existing systems without requiring additional hardware modifications. 

•　Support for a wide range of pre-trained models to enable effective weapon detection.

Solutions

The customer implemented Advantech's ARK-1221L Compact Edge Computer, paired with the EAI-1200, a Hailo-8-powered M.2 AI acceleration module, to enable real-time security screening. The EAI-1200, delivering 26 TOPS of AI processing power with low energy consumption, is designed for real-time edge inference tasks. This solution allows for the rapid detection of prohibited items such as ammunition, dangerous substances, sharp objects, and weapons.   

The ARK-1221L is an Intel Atom-based fanless box PC featuring a LAN port for two camera connections, M.2 expansion for EAI-1200 AI acceleration module integration, and digital I/O for data collection. Advantech further enhances this system with its Edge AI SDK, offering an inference benchmark tool, runtime SDK integration, and remote monitoring and management capabilities. Additionally, Hailo’s Model Zoo provides pre-trained models, the Dataflow Compiler supports model conversion, and TAPPAS offers pre-trained examples, ensuring seamless deployment and optimization for security screening applications.

Outcomes

• Save 50% on establishment costs and achieve a 10x faster detection rate compared to metal detectors.
• Quickly upgrade to AI-enabled security, enhancing real-time bag screening efficiency to handle up to 4,000 people per hour (ppl).