Machine vision & AI for advanced semiconductor manufacturing

MVTec provides powerful machine vision software with integrated deep learning technologies for precise and highly automated semiconductor manufacturing. Our software supports the entire process chain – from wafer production and front-end lithography to advanced packaging and back-end assembly – with robust tools for defect detection, code reading, precision alignment, 2D/3D position recognition, wafer OCR, and metrology. 

Machine vision and AI increase throughput, stabilize processes, and help manufacturers meet strict quality and yield requirements, even at high line speeds and submicron structures. They also reduce unplanned downtime by detecting defects or misaligned wafers before they can block, damage, or stop equipment. 

Machine vision across the
semiconductor process chain

1. Front-End Process

Deposition, coating, lithography, etching, ion implant, CMP, cleaning, AOI & testing

Typical machine vision tasks
  • High precision wafer and mask alignment
  • Surface inspection for cracks, scratches, and particles
  • Lithography pattern and reticle inspection
  • 2D/3D wafer metrology, including topography evaluation
  • Reliable wafer identification with OCR and Bar & 2D Code reading

2. Advanced Packaging

RDL layer creation, µ-Bumping, interconnect formation, die-attach, underfill, encapsulation 

Typical machine vision tasks
  • Sub-micron alignment between stacked wafers and dies
  • Fast inspection and sub-pixel-precise measurement of solder balls, bumps, pillars, pads, and RDL
  • Bump height and coplanarity inspection
  • Sub-surface TSV inspection
  • Detection of voids and cracks in complex 3D packages

3. Back-End Process

Dicing, bonding, molding, final AOI, handling & delivery

Typical machine vision tasks
  • Wafer alignment before sawing
  • Dicing inspection for edge chipping or cracks
  • Die-attach and bonding inspection
  • Completeness checks and end-of- line AOI

[Webinar] How to maximize the yield in wafer production

Watch the webinar to discover how advanced vision technologies are transforming wafer inspection and driving quality and throughput in semiconductor manufacturing.

To the webinar


Key semiconductor applications enabled by MVTec software

Wafer identification with Deep OCR and 2D Code reading

Wafers must remain uniquely identifiable throughout the entire production process. Typical markings include batch numbers, dates, manufacturer codes, and other process information. 

MVTec’s Deep OCR and code reading technology reliably reads characters and 2D codes even under challenging lighting, low contrast, or variable marking quality. Automatic grouping of characters into words improves recognition accuracy and reduces misreads of visually similar characters – for example, in wafer sorters and Equipment Front-End Modules (EFEMs).

Defect detection for scratches, cracks and particles

Wafer AOI with defect detection 

Scratches, micro-cracks, and particles directly impact yield and machine reliability. 

MVTec software combines rule-based image processing with deep learning-based methods like anomaly detection to detect all kinds of defects on wafers, dies, and packages.

The software delivers robust results even on highly structured or noisy backgrounds and supports high speed and semiconductor equipment. 

Redistribution layer
Redistribution layer

Redistribution layer inspection with 2D/3D metrology 

Redistribution layers (RDL) are critical for interconnects between chips and substrates. Their geometry must be inspected in the micron range to ensure compliance with design rules. MVTec provides subpixel-accurate pattern matching and metrology methods for precise alignment and 2D structure measurement down to 1/50 pixel. The flexible toolbox approach enables users to select the best combination of algorithms for each inspection setup and object.

Image source: Sawyer, Brett, Yuya Suzuki, Zihan Wu, Hao Lu, Venky Sundaram, Kadappan Panayappan, and Rao Tummala. 2015. “Design and Demonstration of 40 Micron Bump Pitch Multi-Layer RDL on Panel-Based Glass Interposers.” IMAPSource Proceedings 2015 (1): 379–85. https://doi.org/10.4071/isom-2015-WP24

Copper pillar bumps under magnification
Copper pillar bumps under magnification

Bump inspection with 3D vision 

Advanced packaging requires highly accurate bump structures for reliable electrical interconnects.

MVTec's 3D vision toolset supports surface-based 3D matching, 3D object processing, and 3D surface inspection for bump geometry and coplanarity checks. It handles data from various 3D sensor technologies, including electron microscopy, and provides high-speed, high-accuracy 3D metrology. Even arc and spherical surfaces can be reconstructed and analyzed precisely.

Image source: arthasarathy, Ravi, and Umut Tosun. 2023. “Defluxing of Copper Pillar Bumped Flip Chips.” IMAPSource Proceedings 2023 (DPC): 438–82. https://doi.org/10.4071/001c.90768

Precision alignment across the process chain

MVTec software performs alignment tasks in multiple semiconductor processes:

Lithography & photomask alignment

Shape-based matching aligns the current wafer position to existing patterns by precisely calculating displacement and rotation.

Probing, wafer dicing, die bonding

The orientation of the wafer can be derived from a single image. Operators in our software HALCON estimate rotation angles and ensure correct alignment before critical steps like probing or pick-and-place.

Determine position of the die after mounting
Determine position of the die after mounting

Die, ball, and wire bonding with position recognition 

Precise positioning and bonding are essential for reliable electrical connections. 

MVTec software provides subpixel-accurate 2D and 3D measurements and detects small positional deviations or voids, even in complex backgrounds. Cross-section profiles along local axial directions allow detailed bond analysis. 


Deep learning for complex inspections

MVTec combines classical rule-based algorithms with AI methods such as deep-learning-based classification, segmentation, and anomaly detection.

This enables
  • Robust defect detection in complex backgrounds
  • Stable decisions despite variations in illumination or structure
  • Fast adaptation to new products and process variations

This combination helps manufacturers reduce scrap, improve yield, and shorten time-to-production for new devices.

Our deep learning methods


[White paper] How machine vision automates and optimizes processes in the semiconductor industry

Download the white paper and learn how to sustainably increase the efficiency of your semiconductor value creation processes.

To the white paper


Why MVTec HALCON is ideal
for semiconductor applications

  • Full range of technologies from advanced blob analysis and matching to 3D vision and deep learning
  • High-speed optimized algorithms including parallel processing and hardware/AI acceleration
  • Sub-micron measurement and matching accuracy
  • Proven industrial reliability for 24/7 use
  • Hardware independence and seamless integration for long-term flexibility

More about HALCON