Overview

LG Innotek a leading global electronics manufacturer headquartered in South Korea, operates across semiconductors and substrates, camera modules, and automotive electronic parts.

To support high-speed manufacturing, LG Innotek needed a more intelligent inspection approach capable of detecting microscopic defects, making real-time decisions, and working seamlessly with existing production systems - without compromising data privacy.

SotaVision enhanced the existing inspection workflow with AI-powered visual inspection, enabling high-precision defect detection and automated classification directly at the production line.

The Challenge

Inspecting Microscopic Defects at Production Speed

Defects such as insufficient solder, solder bridges, excessive solder, solder balls, billboards, and tombstones can vary significantly in appearance and may occur simultaneously on a single component.

Components move rapidly through the production line, requiring inspection to be completed within strict latency thresholds while maintaining extremely high detection accuracy.

Some defects occur in low-contrast areas or on reflective surfaces, making them difficult to distinguish using conventional image-processing approaches.

SotaVision Solution

01

Microscopic Defect Detection

Detect microscopic defects on compact moving components

02

Real-Time Latency

Meet real-time latency thresholds for on-the-line decision-making

03

On-Premise Data Security

Ensure absolute data security through fully on-premise deployment within factory systems only

04

MES & Traceability Integration

Compatible with existing MES and traceability systems

05

Optimized Image Acquisition

A tuned lighting and image-capture pipeline reliably surfaces defects on low-contrast and reflective surfaces.

06

Production-Speed Throughput

Inference runs at line speed so inspection never becomes a bottleneck on the production floor.

Key Results

99.99% - Defect Detection Accuracy

#01

Embed Gen AI, computer vision, or NLP directly into your operations or user experience

<50 ms - Processing Time per Unit

#02

Real-time AI inference designed for high-speed inline inspection.

0.1 mm - Smallest Defect Size Detected

#03

Enabling inspection of highly detailed defects on compact components.

20+ - Defect Types Automatically Classified

#04

Supporting a broad and evolving range of electronics manufacturing defects.

30% - Reduction in Manual Inspection Workload

#05

Reducing the need for manual verification and enabling more automated quality control.

Business Impact

Complex defect coverage

The AI identifies 12+ defect types - adhesive misalignment, air bubbles, soldering defects, and microscopic anomalies - even when several appear on one product.

Real-time inspection performance

Images are processed in under 50 ms per product, matching line speed while sustaining accuracy up to 99.99%.

Detection under challenging conditions

Combining Computer Vision, AI, and optimized image acquisition, defects are caught on low-contrast or reflective surfaces where traditional methods fall short.

LG Innotek