Automatic Anomaly Detection - Deep Learning for Surface Inspection

Machine Learning [ML] for Quality Assurance

Self-Learning Surface Inspection: Automatic Anomaly Detection in Textured Surfaces

Self-Learning Algorithm

  • This algorithm functions on live images directly without the need for training or labeling date in advance.
  • Boundary condition: when compared to the defect-free surface, which needs to account for a minimum of 51% of the whole inspected area, defects are small.

How Does it Work

For the complete evaluation of the entire image, local training of n classification is used. A majority voting decision is used to set-up the end result.

Integration and Deploying

The algorithm can be integrated into any machine vision software. As an example, it can be deployed in a Region of Interest (ROI) with the use of the anomaly surface’s size as criteria for measuring defect quantification.

Huge Advantages

  • There is no requirement for teaching in advance
  • As the algorithm can automatically adapt itself to any random or arbitrary surface, there are no necessary setting parameters.
  • The inspected surface can be evaluated in less than 50 msec on Core i3

Adaptive Surface Inspection: Additional Examples

Challenge

  • To identify faults in aesthetic technical and complex functional surfaces

Solution

  • Machine vision based inspection
  • Process fluctuation consideration
  • Process integrated 100% inspection

Customer Value

  • Inspection is reliable, fast, and contact-free.
  • Quality is documented

Deep Learning for Surface Inspection

Automatic anomaly detection in textured surfaces

EyeVision software now includes the Deep Learning Surface Inspector. This enables easy and dynamic detection of damages, impurities, and surface flaws. It has an adaptive surface inspection section and works on textured surfaces.

The most challenging thing is to identify faults on aesthetic and complex functional surfaces. By carrying out an inspection based on the DL Surface Inspector, a solution is achieved. This is based on machine vision under the consideration of process fluctuations.

Advantages

There are numerous significant advantages:

  1. Any machine vision software can integrate the algorithms.
  2. By, for example, using the size of the anomaly surface defect qualification criteria, it can be deployed in a Region of Interest.

Further Important Advantages:

  • There is no required teaching in advance
  • As the algorithm automatically adapts itself to all random surfaces, there are no required setting parameters
  • It takes less than 50 ms to inspect the surface on Core i3

This information has been sourced, reviewed and adapted from materials provided by EVT Eye Vision Technology.

For more information on this source, please visit EVT Eye Vision Technology.

Citations

Please use one of the following formats to cite this article in your essay, paper or report:

  • APA

    EVT Eye Vision Technology. (2019, May 13). Automatic Anomaly Detection - Deep Learning for Surface Inspection. AZoOptics. Retrieved on November 15, 2019 from https://www.azooptics.com/Article.aspx?ArticleID=1460.

  • MLA

    EVT Eye Vision Technology. "Automatic Anomaly Detection - Deep Learning for Surface Inspection". AZoOptics. 15 November 2019. <https://www.azooptics.com/Article.aspx?ArticleID=1460>.

  • Chicago

    EVT Eye Vision Technology. "Automatic Anomaly Detection - Deep Learning for Surface Inspection". AZoOptics. https://www.azooptics.com/Article.aspx?ArticleID=1460. (accessed November 15, 2019).

  • Harvard

    EVT Eye Vision Technology. 2019. Automatic Anomaly Detection - Deep Learning for Surface Inspection. AZoOptics, viewed 15 November 2019, https://www.azooptics.com/Article.aspx?ArticleID=1460.

Tell Us What You Think

Do you have a review, update or anything you would like to add to this article?

Leave your feedback
Submit