Overview

This document details LNC Technology's AI-based vision systems designed for industrial applications such as visual inspection, workpiece positioning, and mold inspection. The product lineup includes AI-2000, AI-1800, AI-WG1600, and MI-5000, each optimized for specific manufacturing inspection and automation tasks.

Product Specifications

AI-2000 and AI-1800

• CPU: Quad-core A75
• Memory: 4GB storage
• Connectivity: 4 USB ports, Gigabit Ethernet
• Display: HDMI output
• Size: 115 x 99.5 x 57.8 mm
• Camera: 8 million pixels (color not supported)

MI-5000 Mold Inspection System

• Display: 10" industrial touch screen
• Storage: 32GB MicroSD
• Network: 100/1000 Mbps Ethernet (RJ45)
• USB Ports: 2
• Camera: 1.2 million pixels CCD, 1/3" sensor, 30 FPS, GigE interface, zoom lens (f12mm-36mm)
• I/O: 3 inputs, 2 outputs
• Protocol: TCP/IP

Hardware Accessories

Includes industrial-grade camera holders, light holders, white light sources, magnetic stands, hooks, light controllers, and adapter plates designed for robust integration.

Software Features and Capabilities

• Camera controls: zoom, move, capture, sensitivity adjustment, and point management
• Multi-camera support: MI-5000 supports up to 2 cameras for large mold monitoring
• Multi-area monitoring: overlapping zones with independent sensitivity (scale 1-100)
• Alarm system: notifications via email/cell phone and machine stop triggers to prevent defective production
• Masking function: excludes non-monitored areas to reduce visual clutter

Industrial Applications

Mold Inspection (MI-5000)

Protects injection molds by detecting residual materials, mold position, and appearance consistency. Interfaces with gantry robots and molding machines to halt production on abnormalities.

Surface Defect Detection (AI-2000)

AI-based recognition detects defects like color deviation, cracks, burrs, dimensional errors, foreign materials, and texture anomalies across materials such as plastic, wood, glass, metal, stone, and fabric.

Wood Grain Recognition (AI-WG1600)

Deep learning identifies and classifies 27 wood grain types with high accuracy, surpassing traditional optical methods.

Workpiece Positioning (AI-1800)

Edge detection and coordinate calculation correct workpiece orientation and position, enabling precise machining, especially for transparent materials like glass.

System Integration and Workflow

1. Data collection under stable lighting
2. Labeling and AI model training using LNC defect marking tools
3. Deployment of trained AI modules
4. Real-time inspection with feedback to control systems

Advantages Over Traditional Systems

• Higher accuracy recognizing complex images and defects mixed with backgrounds
• Lower hardware/software requirements compared to traditional AOI systems
• Faster setup and training of recognition models
• Robustness to environmental changes and defect variability

Case Studies

Sponge Strip Positioning

AI-2000 accurately detects sponge strip hole centers for gantry robot positioning, reducing cycle time to 40 seconds for four stations and improving accuracy over manual inspection.

Badminton Surface Defect Detection

AI visual recognition replaces manual inspection to detect stains, short shots, overshoot, reducing labor and improving defect detection rates.

Glass Workpiece Edge-Finding and Positioning

AI-1800 corrects positioning deviations of transparent glass workpieces, enabling precise machining despite initial misalignment.

Best Practices and Recommendations

• Maintain stable lighting during image capture for high accuracy
• Use LNC labeling tools to create comprehensive training datasets covering all defect types
• Set appropriate sensitivity levels per monitoring area to balance detection accuracy and false alarms
• Utilize multi-camera setups for large or complex inspection zones
• Integrate alarm and machine stop functions to prevent defective production continuation

Limitations and Considerations

• Environmental factors like lighting changes can still affect detection despite AI mitigation
• Initial training requires sufficient labeled image data and effort
• System performance depends on proper integration with existing machinery and control systems

Traditional vs. LNC AI-2000 Defect Detection Systems

Limitations of Traditional Methods

• Highly sensitive to external conditions (lighting, background, color, surface curvature)
• Accuracy drops sharply with new or varying defects
• Requires programming skills for software re-optimization when environments or defects change
• Complex installation and hardware setup
• Long development times due to hardware readjustment and programming rewrites
• Complex backgrounds and textured surfaces confuse detection

Advantages of LNC AI-2000

• Simplified hardware with powerful software enables quick setup and faster development
• Machine learning training without programming skills using LNC tools
• Efficient training with tens to hundreds of images
• Deep learning tolerates flaw variations and converts new flaws into learning data to improve recognition

Critical Considerations

• Complex backgrounds and multiple surface patterns still challenge accuracy
• Continuous learning and adaptation are essential for maintaining accuracy in changing environments

Contact and Support

LNC Technology Co., Ltd offers support via email ([email protected]) and phone (+886-0800-561888), with offices in Taiwan and China.

©2021 LNC. Technical data subject to change without notice.