Quality inspection and process monitoring using computer vision

Purpose

Visual quality inspection is an essential step in many manufacturing processes, which, if done manually, is often a time-consuming and tiring task with inconsistent performance. While automation through implementation of conventional computer vision systems is sometimes possible, variations in size, shape, and location of the defects is usually too difficult for these systems to handle due to their hand-crafted and rigid detection rules.

Visual process monitoring, such as continuous tracking or positioning of products moving on a conveyor, is associated with similar challenges.

However, with AI-based computer vision, where smart algorithms learn by themselves through training on large image or video datasets, automation of these tasks can now be achieved in a rapidly increasing number of cases.      

Our Service Offer

Our computer vision service offer consists of four steps, where step 1 and 2 are free of charge: 

1. Initial meeting: Introduction and a general overview of your company's needs and level of preparedness with respect to adoption of a computer vision solution.

2. Factory visit: RISE visits your factory to gain a first-hand understanding of your operational environment. This visit can optionally be combined with the initial meeting (step 1).

   After the visit, RISE will make an assessment of the complexity of your use case and come back to you with a suggestion of how to proceed. For very complex cases, we might suggest initiating a research project, led by RISE, in collaboration with your team and additional partners.

3. Workshop (Price: depends on agreed upon duration): We meet for a detailed discussion to produce 

• a problem and requirements specification:
  • outlining the types of defects/events that the vision system should be able to detect
  • prioritizing the defects/events
  • establishing the minimum performance criteria for the vision system
  • determining how to create an effective inspection/monitoring setup (e.g., whether there is a need for robots, dedicated lighting etc.)
• a data readiness analysis:
  • assessing whether your company is currently collecting data relevant to the computer vision task
  • evaluating how easily the existing data can be accessed and utilized
  • analyzing the quality of the current data to ensure its suitability for the project
  • determining whether there is a requirement for additional data, such as synthetic data 

The outcome of steps 1-3 will form the basis of an detailed implementation plan. In step 4 of our service we carry out this plan, which of course can be broken down into smaller parts (sprints) with clearly defined milestones, to reduce financial risk. 

4. Training and implementation of AI-based computer vision (Price: per agreement, Duration: per agreement): We implement machine vision solutions for defect detection or process monitoring by training state-of-the-art AI algorithms on image or video data partly or entirely collected from your process. A critical aspect of this training involves data labeling, a task that we undertake with input from your company's experts to ensure accurate categorization of, e.g., defects. 

   Note: in this step we also include cases where there already is a vision system installed, but where performance improvements are needed. As an independent research institute, our expertise is not limited to any specific system or software, allowing us the flexibility to work with a variety of platforms and technologies.      

Deliveries

What we can deliver:   

• A data readiness analysis that prepares you for implementation of computer vision
• Implementation of computer vision solutions in your production
• Improvements to existing solutions