With the arrival of Industry 4.0, machine vision plays an increasingly important role in smart manufacturing. This article explains how machine vision works and why it is an appropriate choice for process automation and quality improvement.
What is machine vision?
Definition: Machine vision uses optical, non-contact sensing devices to automatically capture and interpret images of real scenes to obtain information for controlling machines or processes. It uses computers to emulate human visual functions, extracting information from images of objects, processing and interpreting that information, and applying the results to inspection, measurement, and control. Key characteristics of machine vision are high speed, large information throughput, and versatile functionality.
Example: Bottle fill-level inspection
Machine vision systems are designed to automatically extract information from digital images for process control or inspection. The illustration below shows a typical workflow.
To illustrate, consider a fill-level inspection system used in a brewery. As each beer bottle passes a sensing station, the sensor triggers the vision system to emit a strobe and capture an image of the bottle. After the image is acquired and stored in memory, vision software processes and analyzes the image. The system then issues a pass or fail result based on the actual fill level. If a bottle fails the check, the vision system sends a signal to a diverter to remove the bottle from the production line. An operator can view rejected bottles and ongoing process statistics on a display.
Figure 1: Beer bottle fill level inspection example
Measurement and robot guidance
Machine vision systems can also perform dimensional measurements, such as determining spark plug gaps or providing positional information to guide robots during assembly. The example below illustrates how a vision system can be used for pass/fail detection of an oil filter (right) and for measuring the width of a center shaft head on a bracket (left).
Figure: Real-time inspection and measurement on a production line
Binary inspection results
- If the product meets the required criteria, the inspection result is "pass".
- If the product does not meet the required criteria, the inspection result is "fail".
Advantages of machine vision
Although human vision excels at qualitative interpretation of complex, unstructured scenes, machine vision offers advantages in speed, accuracy, and repeatability for quantitative measurement in structured scenes. On production lines, a machine vision system can inspect hundreds or even thousands of parts per minute. With appropriate camera resolution and optics, machine vision can detect fine details that are invisible to the naked eye.
Because inspection is non-contact, machine vision prevents damage to parts and avoids maintenance time and costs associated with mechanical wear. Reducing human involvement in manufacturing processes also improves safety and operational efficiency. Machine vision can prevent contamination in cleanrooms and protect workers from hazardous environments.
Types of machine vision systems
- Smart cameras
- Embedded-based systems
- PC-based systems
Typical system components
- Image acquisition: light source, lens, camera, frame grabber, mechanical platform
- Image processing and analysis: industrial host, image processing software, graphical user interface
- Decision and execution: signal output unit, mechanical actuators
