3D Inspection Made Easy with Machine Vision
3D Inspection with Machine Vision is Changing the Game in Manufacturing
As technology progresses, programmers, data scientists, optical experts, and more are working to introduce automation into more processes and further streamline them. While there are some ethical and economic quandaries attached to this, the budding infrastructure behind these changes is fascinating, and the many applications for them continue to amaze.
Here, we’d like to discuss how manufacturing lines are evolving to incorporate artificial intelligence (AI) systems that optimize processes such as quality assurance and hardware assembly.
The factory lines of the future are developed with state-of-the-art tech to detect defects and help maximize yield, which is especially poignant as the semiconductor shortage continues. Beyond that, these can produce massive amounts of data for storage, which requires innovative solutions in order to maintain.
We’ll talk about the need for tiered approaches to this storage a bit later on.
How Does Machine Vision Work With Computer Vision?
Machine Vision is a term to explain the way in which a machine “sees” something and interprets it, like when analyzing a component and detecting imperfections. According to Forbes, facilities have used machine vision technology since the 1950s. But it grew in popularity during the 80s and 90s.
This technology helped people maintain factory and assembly lines standards but has expanded even further in recent years. Soon, these systems will be fully automated and will utilize computers as they work to assemble the equipment and inspect it for quality and standards.
Forbes describes these as the key components of a machine vision system:
-
Sensors: These capture the image or video of the item so it can be inspected
-
Frame-grabber: The shutter or similar device that captures an individual frame for examination.
-
Cameras: The apparatus around the camera, including shutter and lens
-
Proper lighting: A necessity for the clarity of your image and focus on details that need to be seen.
-
Software and computing hardware: Hardware that can help facilitate the entire process, operating both the cameras and the forthcoming algorithms.
-
Algorithms that identify patterns: We’ll talk more about these later, but this AI will comb over the images to spot imperfections and faults that would make the equipment unacceptable.
-
Output to a screen of other mechanical components: In the event of human inspection, these terminals can act as means for the person to watch along and monitor the quality assurance process, possibly of several units at once.
An item is placed in front of a sensor, and a camera hones in on it, properly capturing every detail so that the frame-grabbers can capture a digital image of it for inspection. The software and algorithms will look at the samples of the items for patterns, and when it detects something that does not fit in with those patterns, it labels it as a defect and marks the product as failing inspection.
This is an important innovation for factory lines as rising labor costs and dwindling skilled human labor make it hard to staff these lines, so automation can act as a permanent solution.
Computer vision is the software and algorithm component of this process, but it has uses beyond machine vision. Artificial intelligence can examine digital materials such as images and videos for content to understand it better.
AI is the driving force behind features like Google’s ‘Search by Image’ or ‘content aware’ tools in applications like photoshop, which fill in portions of an image based on their contextual surroundings. If you’re pointing your phone at an object and asking for information about it, this is machine vision, and the software operating on the phone is computer vision.
As a whole, computer vision processes are driven by AI, which learns to look at components in order to evaluate them properly.
How Does The AI In Machine Vision Work?
Most of these solutions are built around the concept of a learning algorithm that can intuit the specifications of an item on its own. As Metrology.news explains,
“The [artificial intelligence] learns from the image data of non-defective products to quickly acquire the ‘expertise’ that inspectors develop over the course of many years. Although the sensitivity depends on the product, in many cases it can match or exceed the defect detection rate of human inspectors.”
In speaking to a machine vision manufacturer, Metrology found that algorithms could learn about these products at a much quicker rate than any person could, which was a boon for the accuracy in quality assurance and the required training time before an assembly line could be started.
Artificial intelligence is continually improving, and Metrology reported hearing that some results are 100% consistent for basic defects. However, there is still room to be made in more minute examinations, such as discovering if a component has scratches.
They explain that this has a surprising number of uses,
“Being able to detect surface scratches is beneficial for many manufactured parts, including plastic moldings, extruded plastic films, and optical parts made from glass or transparent polymers.”
Filters are being worked on to help with this form of defect detection, which is particularly essential in automotive manufacturing. However, the colors and finishes need to be incorporated into the AI.
Some software is conquering this hurdle with machine vision accurately detecting these design features on an individual basis to look past them and scan for scratches.
What Happens To The Data Used in Machine Vision?
Machine vision produces a staggering amount of data, and as Enterprise AI reports, this figure is expected to rise to 160 zettabytes by the year 2025. Of course, this number includes all machine vision applications, so it goes beyond manufacturing and into the realms of healthcare, agriculture, security, and even the high-quality machine vision lenses seen in autonomous vehicles.
Storing such a massive amount of data on a single centralized network can produce significant issues. This amount of data slows the processing of that information by the earlier-mentioned learning algorithms. The slowing down of processing can lead to minor manufacturing issues or even personal endangerment in autonomous vehicles.
Enterprise writer Plamen Minev recommends a modular approach to storing this data, segmenting it based on needed purpose, and implementing machine learning to help comb through footage as needed.
He writes,
“While distributed architectures have many advantages, they also introduce additional complexity. Selecting and deploying the appropriate storage and compute infrastructure at the edge together with centralized management is critical, and significantly impacts the overall system efficiency and cost of ownership.”
This tactic has two crucial benefits. The first is that it enables lower costs when storing low-priority data, as most data is collected and stored but never actually accessed. Specifically, he notes that network-attached cloud storage can cost upward of $3,351 per terabyte per year instead of $50 for lower-cost storage.
This strategy works by having all of the data combed by an AI while on cloud storage and having permanent backups stored on lower-cost storage. Any data that isn’t found to be of interest is removed from network storage.
The second key benefit is that better solutions may become available as both artificial intelligence and network storage solutions change. Having your data stored in a tiered system allows for better transitioning to emerging methodologies and platforms as they arise.
Conclusion
Machine vision is a powerful and innovative new solution to manufacturing that continues to increase, being adopted in more factories. This technology enables fully automated manufacturing lines to conduct their quality assurance without the need for human hands, removing the requirements for training and labor fees.
But, it requires smart solutions and premium equipment; if any one piece of the puzzle doesn’t stand up with the rest, it can drag the entire operation down. Ensure that you’re working with machinery that is built for your job, which sometimes must be custom-made by a manufacturer.
Sources:
Forbes - What is Machine Vision And How Is It Used In Business Today? | Forbes.com