LMI’s edge-based AI solution

AI-Powered Anomaly Detection at the Edge

• Gocator 3D Smart Sensors - High-precision profile data capture
• GoMax ORIN Edge Processors - Advanced AI processing at the edge
• GoPxL Anomaly Detector Software - Intelligent defect detection without programming

• High-Resolution Requirements - Gocator 2610 (2.5μm resolution)
• High-Speed Applications - Gocator 2520 (up to 20 kHz)
• Wide Field of View - Gocator 2690 (up to 2m coverage)

For applications requiring exceptional surface detail capture or multi-layer inspection, the LMI Gocator 4000 and 5500 Line Confocal Sensor series integrate seamlessly with the GoMax ORIN and GoPxL Anomaly Detector.

The Gocator 4000 series offers coaxial line confocal technology with outstanding angular range (up to ±85 degrees) and zero shadowing effects, ideal for semiconductor and electronic component inspection. 

The Gocator 5500 series provides multi-layer capabilities, simultaneously generating 3D tomography, 3D topography, and 2D intensity data for transparent materials and complex structures. 

GoPxL Anomaly Detector Software

• Training Methods: 2D intensity map or 3D height map training 
• Training Requirements: Typically 10-20 good samples, 5-10 defective samples 
• Data Augmentation: Synthetic defect generation reduces sample requirements 
• Detection Capabilities: Surface anomalies, geometric deviations, texture irregularities
• Performance: Real-time detection at production speeds 
• Integration: Standard industrial protocols (EtherNet/IP, PROFINET, Modbus)
• Licensing: Licensed LMI Dongle required for LIVE operation (license-free evaluation in REPLAY mode)

• Minimum Detectable Feature Size: Down to 2.5 microns
• Detection Accuracy: Typically 90-99.8% (application dependent)
• Processing Speed: Real-time at production line speeds
• Inspection Throughput: Up to 20,000 parts per hour (application dependent)

• Communication: EtherNet/IP, PROFINET, Modbus TCP, TCP/IP, RS-485
• I/O: High-speed trigger inputs, pass/fail outputs, encoder inputs
• Integration Architecture: Direct sensor-to-GoMax or multi-sensor network via Master hub

Train Without the Cloud

• Develop detection models directly on the production line
• Maintain complete data privacy and security
• Update models in real-time as production conditions change
• No recurring cloud subscription costs

No CAD Files Needed

• The system learns from actual production parts, not idealized models
• Identify anomalies in complex geometries and natural materials
• Reduce implementation time with simplified setup
• Adapt to product variations without reprogramming

Beyond Basic Thresholds

• AI identifies patterns humans might miss
• Detect subtle defects traditional vision systems overlook
• Reduce false positives and false negatives
• Improve detection accuracy over time through continuous learning

Traditional machine vision systems excel at finding known defects with defined parameters. But what about the unexpected anomalies that don't match your programmed rules?

• EV battery manufacturers face costly recalls from microscopic welding defects
• Electronics producers struggle with subtle surface irregularities on complex components
• Building material inspectors miss product-compromising defects amid natural variation
• Automotive assembly lines need 100% verification of critical safety components

• Expert system optimization and integration guidance
• Custom technical solutions for your unique requirements
• Thorough feasibility testing and application validation
• Direct technical support from vision specialists

• Precision assembly and configuration for immediate readiness
• Optimized software settings for maximum efficiency
• Quality assurance testing for reliable performance
• Scalable solutions adapting to your changing needs

• Project qualification and planning for complex vision challenges
• Custom subsystem development tailored to your specific needs
• Bespoke software solutions for specialized requirements
• Collaborative approach combining industry and vision expertise