How do machine vision systems process the images they capture?
With the right combination of sensor, lens, lights, and other components in place, a machine vision system is ready to capture images. But that’s only the first step.
Once it acquires an image, the system needs to process and analyze the result to make a decision: to read a barcode, detect a defect, confirm that a kit contains all of its items, or measure a part.
Historically, the only way to get enough processing power to do image analysis was to send the image to an external PC. Depending on the size and complexity of the image, that could take time, potentially slowing down the line, requiring a separate PC dedicated to the task.
Today, highly capable vision systems have embedded processing, which speeds computation and reduces possible points of failure.
Once it acquires an image, the system needs to process and analyze the result to make a decision: to read a barcode, detect a defect, confirm that a kit contains all of its items, or measure a part.
Types of image analysis tools
Depending on the task, a vision system can leverage one or several specific image analysis tools.
Edge detection/extraction
In an image, an edge marks a distinct change in intensity. Such changes mark discontinuities in depth, orientation, and material, and allow for finding the boundaries of objects.
Pixel counting
A straightforward algorithm can count the number of pixels at each grayscale level in a region of interest or the overall image, providing information used by other functions.
Blob detection
Different regions can be identified, providing information used by other functions.
Pattern/template matching
Guiding and positioning functions use pattern matching or template matching on parts and products to determine object location and orientation.
Color analysis
Where object color is an important feature, there are several color tools, including color extraction and segmentation, and color matching.
Putting image information to use
The information derived from that image analysis isn’t much good if it stays in the camera. It needs to be shared with other systems that can put it to use.
To do this, machine vision systems use standard industrial formats or protocols to share data with other devices, usually via wired connections. That data can then be used in many ways:
It can tell a nearby programmable logic controller (PLC) to fire a piston to bump defective parts off the line.
It can pop up on a human-machine interface (HMI) display so an operator can check up on a production process and make adjustments.
It can be used with the factory’s process control system or a manufacturing execution system (MES), so the system can make adjustments to optimize the manufacturing process.
It can trigger sorting mechanisms in an automated sortation system to get a package on to the right truck.
Image-based data drives larger decisions, too
In addition to individual actions on single parts, products, or packages, the data generated through machine vision applications can also be aggregated and used to tackle larger business issues.
For example, businesses can use the data to find the root causes behind delays, defects, and missed shipments. Is one supplier using inferior raw materials for their part? Or is the problem caused by how parts are stored or shipped? Are workers being trained inadequately? Machine vision data can be analyzed to identify the source of the problem.
Other uses of machine vision data can include:
Learning that a certain type of part causes rejection of the assemblies it is installed in.
Locating and identifying parts, constantly updating inventory, and guaranteeing accurate order fulfillment.
Analyzing large data sets with an enterprise resource planning (ERP) system to make long-term operational decisions.
Reviewing failure rates and types to pin down where production errors are being introduced.
Types of communication protocols
There is a wide range of protocols used in automated manufacturing, from simple on/off signals sent by a direct connection, to sophisticated industrial protocols such as Ethernet/IP, Profibus, and DeviceNet, among many others.
While interoperability is common, meaning you can buy the best device for a specific job and be confident it can communicate with all your current equipment, it remains true that various manufacturers favor one protocol or another, and there are a lot of proprietary standards.
Machine vision systems can be configured to be interoperable with virtually any set of industrial protocols, providing their information in a form that is easily used to improve automated industrial operations.
Back to the start
Introduction to Machine Vision
Table of Contents
- Introduction to Machine Vision
- Types of machine vision and what they can do
- Rule-based vs AI-powered machine vision
- Machine vision hardware components
- The importance of lighting in machine vision applications
- Image analysis software
Download Whitepaper
mail_outline Contact Sales
pricing Get Pricing
settings Get Product Demo
