English 

Our view #6 - 2016 Q2

MADE Digital - Smart Production

Written by Jørgen Læssøe

During our 30 years as a vision supplier, JLI vision has repeatedly had to turn down applications where aesthetic control in some variation was needed. Typically applications have been turned down because it was impossible to define control metrics for a reject. We have therefore often asked ourselves: "What makes a reject from an aesthetic point of view? Is is possible to define this from a perception view-point?". Actually there is no research - at least to our knowledge - that have focused on what we as humans will se as an aesthetic reject. It probably varies from person to person, but there may also be some underlying metrics or methods that could be implemented in a vision system. In automated production vision based quality control is a key element. Vision based quality control can roughly be categorized in two types: structured and aesthetic.<\p>

The aim of structured quality control is to determine and check whether a product has the correct amount of holes and dimensions etc. often in order for it to be assembled with other parts. Creating a requirement specification is normally relatively easy and subsequently it will be easy to determine if a machine vision system comply with the specifications.

Aesthetic quality control aims at ensuring that the finished product respects the end users expectations. This means that it cannot contain too many scratches or too deep holes or a wrong color. Contrary to structured quality control it is rather difficult to define objective requirement specifications for aesthetic quality. In practice acceptable and rejectable products are selected and used as a reference for manual inspection.

Project description

The project aims at improving our understanding of what aesthetic quality is. Is it possible that this understanding can provide a method or a formula that subsequently can be used and implemented in general machine vision systems for automatic aesthetic quality control? Concretely the research is aimed at gaining:

  • General knowledge about the parameters that influences end-users judgment of the aesthetic quality 
  • Know-how about how this research can be translated into general machine vision solutions/algorithms capable of automating aesthetic quality control

Participants

  • Companies working with products where the aesthetic quality is essential
  • Aalborg University, institute working with computer vision, perception and user interpretation
  • JLI vision

About MADE Digital

MADE is an association that through research seeks to generate and implement new knowledge about production that provides danish companies with a global edge which is essential for the companies development and strength. Without deployment of knowledge it is without reel value - knowledge must be implemented. Should you want to participate and have a new angle on aesthetic control please contact JLI, we will be happy to introduce you to the appropriate network.

Made