Machine vision is a term that describes a broad set of technologies that execute computerized acquisition and analysis of images, particularly to serve automated processes where human visual perception and reaction would be impractical or impossible. Machine vision has been implemented for decades in diverse applications like agriculture, manufacturing, automotive production, and medical device assembly.
Besides being a mature technology, machine vision is a thriving global marketplace that has experienced consistent growth over many years and is projected to be worth more than $20 billion by 2030, according to analysts. This marketplace comprises manufacturers of components, including images and other sensors, cameras, optics, imaging systems, specialized computers, lighting devices and software. It also includes engineering companies, or “systems integrators,” that design, build and install complete turnkey machine vision solutions.
Machine vision can be described further as an engineering discipline. Engineers trained in electrical and mechanical engineering or computer science often enhance their skill sets with machine vision engineering.
What are some examples of machine vision?
Machine vision use cases generally fall into four broad categories.
Inspection. Perhaps the most common task of machine vision is quality inspection. A machine vision system can automatically detect defects, verify the correct assembly of components, or check colors and other features for every item produced.
Measurement. Machine vision is often used to perform in-line, non-contact measurements in an automated process, inspecting 100% of production, and identifying or rejecting parts with out-of-tolerance features. Some systems can achieve measurement precision to a few microns or better.
Location and Guidance. By coordinating the data and output of machine vision tools to real-world coordinates, a system can accurately locate parts or a part’s features, becoming the “eyes” of a robot or motion control system.
Identification and Classification. Machine vision is regularly used to identify or differentiate objects by:
- Visual characteristics
- Reading code symbology (as in 1D barcodes or 2D matrix codes)
- Reading numbers and characters using optical character recognition and verification
What are the basics of machine vision?
Even the most complex machine vision system can be deconstructed into three fundamental functions.
Image acquisition. Acquiring a high-quality image of the component, objects, or features of interest is the most critical part of any machine vision solution. Quality in automated imaging is defined as an image that provides the best possible contrast between objects of interest and the background, with sufficient resolution to extract the necessary data from the smallest possible features.
The selection of imaging components such as sensors, optics, and lighting and their design within a machine vision system can be complex. Fortunately, many applications can be addressed with machine vision systems featuring standard sensors, built-in optics and integrated lighting solutions. In today’s machine vision marketplace, though, components are available to address all aspects of acquisition, from simple to involved.
Image analysis. In machine vision, images are analyzed by processing pixel-level data using various algorithms, also called “tools” or “steps” in configurable machine vision components. Some algorithms perform simple tasks, though some advanced functions locate pre-trained models from geometric structures, detect subtle color differences, use deep learning to detect defects or classify objects, or even identify and locate objects in the point cloud of a 3D image.
Results integration. Finally, machine vision’s value to a process comes from integrating the results. This could be an automated rejection system for non-conforming parts, but the data can also be more effectively used to analyze machine or system performance overall.
What is the difference between machine vision and computer vision?
While “machine vision” and “computer vision” might sometimes be used to accomplish similar tasks and are sometimes used interchangeably, each term has a distinct functional and historical meaning.
Computer vision commonly refers to the science of using artificial intelligence (AI) techniques for the classification (or “segmentation” or “detection”) of objects using advanced neural networks and deep learning to make computers “see” in a perceptive way that mimics humans.
Machine vision is related to computer vision. However, the term refers not to a science but to the overall implementation of technologies that support discrete feature extraction, rule-based comparisons and deep learning to make decisions directly on image data.
Is machine vision difficult to implement?
Today’s machine vision components often include configurable general-purpose systems and application-specific solutions needing no configuration at all. That said, machine vision usually still demands some unique skills. However, an automation engineer can easily acquire these in the same way control engineers must learn how to use PLCs and configure ladder logic.
When the solution may be more complex, distribution and integration partners provide a valuable service within the machine vision marketplace and are ready to help. By default, many end users turn to a systems integration partner to mitigate risk and expedite return on machine vision project implementation.
In conclusion
What is machine vision? Machine vision is a thriving and valuable technology that enhances quality and productivity. It is an enabling technology that guides robots and allows computers to “see.” In nearly every industry, machine vision usage continues to grow and benefit diverse industrial processes.