Chapter 1

What Is Automated Optical Inspection?

Automated inspection and defect detection systems use AI to inspect manufacturing parts for failures and defects. They are used across industries to detect flaws on manufactured surfaces such as metallic rails, semiconductor wafers, and contact lenses. Here are some examples of how companies are using automated inspection systems:

  • Airbus—to detect defects in multiple elements of the aircraft
  • Musashi Seimitsu Industry—to inspect bevel gears used in automotive parts
  • Korea Railroad Research Institute—to detect surface defects, missing parts, and cracks in railway facility components such as rails, sleepers, and fasteners
  • Kansai Electric Power—to assess creep damage on steel pipe welds

Why Use Automated Optical Inspection?

If you have ever worked on a vision inspection system for quality control or defect detection, you know how challenging that process can be.

To take a basic example: A manufacturing belt carrying hex nuts and an inspection camera that captures images of those parts.

Fig. 1

If the camera captured four images, the operator could find the defective nut just by looking at them.

Fig 2

But how about now?

Fig 3

Once images run to thousands or millions, it takes an automated system to separate the defective nuts from the good ones. 

Automated Optical Inspection with MATLAB

This ebook shows how you can use MATLAB® to develop a deep learning network to detect and classify different types of anomalies. You will learn about the three main stages of the defect detection workflow:

  1. Preparing data, including denoising, registration, and labeling
  2. Building and training a deep learning network
  3. Deploying the network to multiple hardware platforms such as CPUs and GPUs

This workflow is iterative:

Iteration and refinement

Once you deploy your system, you get more data. You can use that data to refine and tune your models, redeploy them, and get more accurate results.