Real-Time AI Video Analysis at the Edge
16 camera streams. One industrial computing platform. Real-time AI processing. Multi-camera applications generate large amounts of video data and require considerable processing capacity.
With the right combination of industrial computing and dedicated AI acceleration, multiple camera streams can be analysed locally without relying entirely on remote servers or cloud infrastructure. At Automation Xperience, Arcobel demonstrated this approach by processing 16 Logitech HD camera streams on one IPC platform using a single Axelera AI Metis PCIe accelerator.
Multi-Camera Edge AI
360-degree surround monitoring
- Sixteen HD cameras positioned in a 360-degree surround array across four monitoring zones.
- AI-powered person detection counted people per zone with immediate visual warnings when limits were exceeded.
- Demonstrated converting multiple video streams into real-time information via local AI inference.
Why Process Video at the Edge?
Local computing performance
- Low latency: Fast local analysis enables rapid responses when objects or events are detected.
- Reduced data transfer: Transmit selected images or alerts instead of continuous full streams.
- Local intelligence: Runs AI inference close to the source, independent of external infrastructure.
- Scalable processing: Dedicated AI acceleration handles demanding multi-camera workloads.
Demonstrated Edge AI Hardware Platform
The setup showcased high-throughput local AI processing using the following hardware architecture:
| Component | Configuration |
|---|---|
| Processor | Intel® Core™ i3-12100 |
| Motherboard | ASRock Industrial IMB-X1714 |
| Memory | 8 GB DDR5 |
| Storage | Samsung 990 PRO 1 TB NVMe SSD |
| AI accelerator | Axelera AI Metis PCIe Card, 4 GB |
| Cameras | 16 × Logitech HD cameras |
| Arrangement | 360-degree surround camera array |
The ASRock Industrial IMB-X1714 provides an expandable industrial motherboard platform with DDR5 support, multiple PCIe slots, and extensive I/O options. Complementing this, the Axelera AI Metis PCIe Card combines four AI cores with 4 GB of memory to deliver up to 214 INT8 TOPS of theoretical peak performance (actual performance varies by AI model, video resolution, frame rate, and software pipeline).
Applications for Multi-Camera AI Vision
Depending on operational requirements, this technology can be adapted for object detection, classification, movement tracking, and event monitoring across various industries:
- Crowd and occupancy monitoring
- Industrial safety and perimeter monitoring
- Access control and infrastructure security
- Production and process quality control
- Traffic analysis and movement tracking
The specific AI models and computer vision techniques are selected based on what your system needs to detect, classify, or track.
Choosing the Right Platform
Key requirements to evaluate
- Number of cameras, resolution, and frame rate.
- Required AI models, accuracy, performance, and latency.
- Camera, network, storage, and I/O interface needs.
- Form factor, power consumption, and environmental conditions.
Application-Ready Systems
End-to-end hardware configuration
- Industrial CPU boards, processors, and GPU/Edge systems.
- Axelera AI accelerator cards and industrial storage/memory.
- Camera, networking, and customized I/O solutions.
- Complete lifecycle management and long-term product availability.
There is no one-size-fits-all Edge AI platform. The performance of an Edge AI system depends on how its components work together—the processor, AI accelerator, memory, storage, interfaces, and software pipeline must be configured directly around your application. Arcobel combines industrial and embedded hardware with deep system configuration experience to build platforms tailored to your exact operational requirements.
Discuss Your Edge AI Application
Are you developing a multi-camera vision system? Tell us what you need to detect, how many cameras you use, and where the platform will operate. Arcobel can help determine the right computing architecture for your application.
Contact ArcobelOr call directly: 0412 660 066 | [email protected]
Real-Time AI Video Analysis at the Edge
16 camera streams. One industrial computing platform. Real-time AI processing.
Multi-camera applications generate large amounts of video data and require considerable processing capacity. With the right combination of industrial computing and dedicated AI acceleration, multiple camera streams can be analysed locally without relying entirely on remote servers or cloud infrastructure.
At Automation Xperience, Arcobel demonstrated this approach by processing 16 Logitech HD camera streams on one IPC platform using a single Axelera AI Metis PCIe accelerator.
Discuss your Edge AI application with Arcobel.
Multi-Camera Edge AI Demonstrated in Practice
Sixteen HD cameras were positioned in a 360-degree surround array and divided across four monitoring zones.
AI-powered person detection counted the people in each zone. When a predefined occupancy limit was exceeded, the system generated an immediate visual warning.
The demonstration showed how multiple video streams can be converted into real-time information using local AI inference. Depending on the application, similar technology can be used for object detection, classification, tracking and event monitoring.
Demonstrated Edge AI Platform
| Component | Configuration |
|---|---|
| Processor | Intel® Core™ i3-12100 |
| Motherboard | ASRock Industrial IMB-X1714 |
| Memory | 8 GB DDR5 |
| Storage | Samsung 990 PRO 1 TB NVMe SSD |
| AI accelerator | Axelera AI Metis PCIe Card, 4 GB |
| Cameras | 16 × Logitech HD cameras |
| Arrangement | 360-degree surround camera array |
The ASRock Industrial IMB-X1714 provides an expandable industrial motherboard platform with DDR5 support, multiple PCIe slots and extensive I/O options.
The Axelera AI Metis PCIe Card combines four AI cores with 4 GB of memory and delivers up to 214 INT8 TOPS of theoretical peak performance. Actual performance depends on the AI model, video resolution, frame rate and complete processing pipeline.
Why Process Video at the Edge?
Low latency
Local video analysis enables the system to respond quickly when a relevant object, person or event is detected.
Reduced data transfer
Instead of continuously transmitting complete video streams, the system can send selected images, alerts or analysis results. The exact benefit depends on the system architecture.
Local intelligence
AI inference runs close to the application and does not depend entirely on external computing infrastructure.
Scalable processing
Dedicated AI acceleration extends the inference capabilities of an industrial computer and supports demanding multi-camera workloads.
Applications for Multi-Camera AI Vision
This technology can be adapted to applications such as:
- Crowd and occupancy monitoring
- Industrial safety monitoring
- Access and perimeter monitoring
- Production and process monitoring
- Object detection and classification
- Movement tracking and event detection
- Traffic and infrastructure monitoring
The required AI models and computer vision techniques depend on what the system needs to detect, classify or track.
Choosing the Right Edge AI Platform
There is no one-size-fits-all Edge AI platform. The number of cameras is only one part of the system design.
Important requirements include:
- Number of cameras
- Resolution and frame rate
- AI models and required accuracy
- Inference performance and latency
- Camera, network and I/O interfaces
- Storage requirements
- Form factor and power consumption
- Environmental conditions
- Required product availability
These factors determine whether the ideal solution is based on an industrial motherboard, compact Edge AI computer, GPU platform, AI accelerator card or another architecture.
From Components to an Application-Ready Platform
The performance of an Edge AI system depends on how its components work together. The processor, AI accelerator, memory, storage, interfaces and software pipeline must be configured around the application.
Arcobel combines industrial and embedded hardware with experience in system configuration and multi-vendor integration. Depending on the requirements, a solution can include:
- Industrial CPU boards and processors
- Axelera AI accelerators
- GPU and Edge AI systems
- Industrial memory and storage
- Camera, networking and I/O solutions
- Customized industrial computing platforms
- Lifecycle and availability management
The result is a platform designed around your application instead of forcing your application into a predefined hardware configuration.
Discuss Your Edge AI Application
Are you developing a multi-camera vision system?
Tell us what you need to detect, how many cameras you use and where the platform will operate. Arcobel can help determine the right computing architecture for your application.