Machine vision types are broad classifications in themselves. In its simplest form, it is the process by which a machine determines the properties of an object or scene being viewed. Let’s break that definition down into four parts:
- The process
- Determines the properties
- Of an object or scene being viewed
- By a machine
Machine vision is a unique subset of computer science used in automation. It has been known to be incorporated into the following industries:
Aerospace, automotive, bio-pharmaceutical, medical device manufacturing, semiconductor equipment manufacturing, and packaging. In almost every industry, some type of machine vision is probably involved.
When the concept of machine vision was first introduced, it was referred to as a human with a camera who would view and recognize objects and then manually position the object accordingly (which is probably why we now call this person a machine vision engineer!). Over time, companies found that they were replacing their skilled workers with machines because they could outperform humans on repetitive, monotonous tasks. This realization allowed for rapid growth in the machine vision industry and has continued on this path ever since.
However, let’s dive into the machine vision types.
Table of Contents
Machine vision types
There are many different machine vision types available, but the two main tech ones are
- Structured Light Technology
- Stereo Vision Technology
Both of these have sub-machine vision types. Structured light types are further divided into laser-based and pattern projection types, while stereo vision is separated into passive and active subtypes. We will briefly touch on each of these types below.
Structured Light TechnologyStructured Light technology is subdivided into laser-based and pattern projection (illumination) methods. The first example of structured Light was using a pinhole camera to create an image with shadows instead of lines, giving shape rather than flat shading. This type of work led to the work of Dr. Thompsett, who, in the early ’80s, developed a more structured system using lasers and patented it in 1985.
The main problem with laser-based approaches was that they were costly. To make them cheaper, companies turned to pattern projection methods, which used LED or similar positional light sources instead of lasers to project a specific pattern onto objects.
The result was a system with the ability to mathematically interpret the projected pattern on an object and calculate its position or movement accordingly. This machine vision type is more popular because it’s easier to make, maintain, and use and is significantly more cost-effective.
- Stereo Vision Technology
Stereo Vision Technology can be divided into passive and active subtypes. Passive stereo vision means that special hardware is not required. In contrast, active stereo vision requires using infrared (IR) emitters and cameras to create a pair of images for analysis.
A passive system can use existing structures in an object or scene, such as windows or cracks between objects, to create comparisons between the images. This type of system was the first to be used, but it is less commonly found in the industry today because stereo vision systems that use emitters are more versatile and common.
The other subtype, active stereo vision, uses infrared emitters that project a pattern onto an object or scene with a specific frequency, which can have several uses. Many people think of this type of stereo vision when they hear “machine vision.” The systems are self-contained, meaning no additional hardware needs to be placed on an object or scene for analysis.
Machine Vision Applications
The main applications for machine vision are position measurement, dimensional control, assembly verification/inspection, identification (pattern matching), quality assurance/control (error detection), and alignment.
Position measurement is the process of measuring the actual position of an object as it moves and the object’s orientation at a single point in time. This is often used to create efficient automated systems that do not require human intervention but can also be used for simple tasks such as indexing a movable platform so that a fixed tool can reach a particular part.
Dimensional control is comparing a product dimension to a set standard. The machine vision system can then output a signal when a part exceeds the tolerance threshold to automatically adjust for proper sizing in production or provide feedback when needed during manual assembly to reduce errors and rework time.
In this case, parts can be inspected after they are assembled. Or when in assembly to ensure that they meet appropriate specifications. For example, if you cannot assemble a part in the correct spot but an incorrect fastening method, you can use it to assemble it. A machine vision system can detect when this occurs and provide feedback to prevent errors without halting the assembly line, which would slow production time.
This is a common task for machine vision. For example, you can program an automated system to look for specific shapes on a conveyor belt. And then move the part into storage once found or direct the part to another process. If you can not identify it. You can use pattern matching here to identify parts.
Quality Assurance/Control is identifying parts that fail to meet specifications. You can do this by looking for physical traits or irregularities on a part. Also, Measuring specific defects or characteristics. Or even identifying quality deviations from production procedures.