In a groundbreaking collaboration, HTS and Deepsensing, supported by STM and leading universities, have unveiled an innovative solution for Automated Optical Inspection (AOI) under the Horizon-funded EdgeAI
project. Designed specifically for the digital industry, this cutting-edge system leverages advanced Edge Computing technologies to deliver efficient, localized, AI-driven defect detection.
Currently showcased through a state-of-the-art demonstrator, this solution highlights how Edge AI makes high-precision inspections both accessible and cost-effective for small and medium-sized enterprises (SMEs).
Key Features and Innovations
This project revolutionizes AOI by drastically reducing complexity and cost, empowering SMEs to integrate advanced inspection systems without major infrastructure investments. Its simpe architecture consists of three main hardware blocks as illustrated below:
Key highlights include:
- Compact Conveyor Belt Integration: Featuring STM32 or NVIDIA Jetson processors for real-time testing and high-resolution cameras to detect and classify faults with precision.
- Advanced AI Models: Leveraging YOLO, ResNet, and Transformers algorithms, these models are trained on a multi GPU workstation using real-world production data. Periodic updates are seamlessly delivered to edge devices via internet or 5G channels.
- Localized Edge Deployment: Defect detection is performed directly on Edge AI hardware, minimizing latency and energy consumption while limiting cloud dependency to essential model updates.
Meeting High Standards
The system is designed to fulfill stringent functional and non-functional requirements:
Functional:
- Real-time defect detection for PCB and PCBA boards.
- Continual learning to accommodate new defect patterns.
- Seamless integration with existing production environments.
Non-Functional:
- Energy-efficient operation, optimized for low power consumption.
- Scalable architecture supporting parallel Edge units.
- Affordable implementation tailored for SME budgets.
Unmatched Performance:
- 99% Detection Accuracy: Ideal for complex defect types.
- Sub-10ms Processing Speed: Ensuring high throughput.
- Reliability: Comparable to centralized solutions like the Omron VT-S1080, meeting KPIs for electric vehicles and other demanding applications.
Real-World Applications
SMEs can leverage the innovative AOI system by installing a conveyor belt equipped with a high-resolution camera and a testing board in their labs. Fault models are periodically updated via a GPU server, ensuring seamless updates through internet or 4G/5G connectivity.
While specifically designed for SMEs, the modular design also serves mid-to-large digital manufacturers. Its flexible layout supports fault detection during prototype development or smaller production runs.
Furthermore, multiple parallel AOI stations can process variable customer orders simultaneously, maintaining workflow efficiency and adaptability without disruptions.
Compact AOI: The Next Frontier
Looking ahead, HTS and Deepsensing aim to redefine AOI further with a portable version available at the end of the projetct. This compact system, equipped with high-resolution cameras, NVIDIA or STM32 processors, Wi-Fi, and 5G connectivity, will enable seamless model updates from remote GPU servers. The result? A powerful IoT-enabled device capable of delivering exceptional testing performance anywhere within a manufacturing facility.
A Game-Changer in Manufacturing
This innovation underscores the transformative potential of industry-academic collaboration, providing an open-source framework that fosters further innovation. With deployment scheduled for 2025, Edge AIpowered AOI is poised to revolutionize automated inspection for SMEs and beyond.
Stay tuned as this breakthrough technology reshapes the future of quality control worldwide