Machine vision as a category is growing rapidly; the 3D machine vision market is expected to double in size during the next six years, and today, the technology is a vital component of most modern automation solutions.
There are many factors contributing to the increased adoption of this technology in a manufacturing context. First, demand for automation solutions in general has increased as manufacturers continue to grapple with labor shortages. Second, the cost has decreased dramatically—when cameras, sensors, robotics and processing power are inexpensive, they can be applied to more solutions.
Technology performance also is increasing, giving machine vision systems the ability to process large amounts of information in a fraction of a second. Finally, advanced artificial intelligence and machine learning algorithms are making the data collected from machine vision even more valuable, and manufacturers are realizing the power of those insights.
But what exactly is machine vision, and how can it be incorporated into automation solutions to produce better outcomes?
Giving Robots Eyes and a Brain
A vision system typically includes a series of disparate parts, including cameras, lenses, lighting sources, robotic components, processing computers and application-specific software.
The cameras are of course the “eyes” in this system—there are many types of cameras in use for machine vision, and each camera can be assigned for different application needs. There are also differences in how cameras are configured within an automation solution.
Static cameras are placed in a fixed position, have a more bird’s-eye view of the process unfolding below and can be used in scenarios where speed is imperative. There are also dynamic cameras, which are positioned on the end of the robot arm and much closer to the process, thereby delivering higher accuracy.
Computing power is also an important aspect of the vision system—essentially, the “brain” that helps the eyes conduct their work. The computation resources required for machine vision that leverage machine learning algorithms will be different than those needed for traditional machine vision applications. Many companies offer software libraries for implementing vision capabilities.
Some capabilities are designed for application users; others are intended for use by software programmers. Regardless, it’s software that gives machine vision advanced capabilities that have dramatic impact for manufacturers, with programs to control tasks and the ability to feedback valuable insights from the line.
Machine Vision Applications
The concept of replacing basic human capabilities with a vision-guided solution is gaining steam, as vision for assembly lines can be used in an increasing range of applications and processes.
A typical “electronics in a box” assembly process provides one example. This is a product category that spans many shapes and sizes and is relevant across a number of industries (including but not limited to medical equipment, power tools and home appliances). The assembly process for these hardware components consists of placing the electronics, such as a circuit board, into the housing, or box.
Many of the assembly steps involved in electronics in a box assembly can benefit from the use of machine vision because of the precision required.
Inspection. Machine vision can be used to inspect all top and bottom cover components as they enter the assembly line, looking for defects such as cracks in the metal—if these components are in poor condition, it can lead to quality issues for the assembled unit. With machine vision, cracks are detected quickly, and if they are larger than a specified size, the component is automatically rejected. In addition to cracks, color variations can also be inspected; a color camera can identify discoloration damages and reject faulty units.
Product tracking. Required in the automotive and healthcare industries, product tracking is used during the entire manufacturing process. So usually, the first vision task for circuit boards during the assembly processes is to read the product label, serial number or barcode. The product label identifies the specific unit so it can then be tracked throughout the entire assembly process.
The label location can be pre-defined and the vision system is used to read the label and communicate its findings to other factory systems. Applying labels is a common requirement in assembly lines and another perfect use case for machine vision, which can detect any obstacles on the surface and ensure perfect placement.
Installation. Adding components to a circuit board can require vision capabilities both for quality control and for positioning. As an example, for a varistor that needs to be added to a circuit board, machine vision would first inspect the varistor to verify that both legs are straight and not damaged. Vision can also inspect the varistor’s identification to confirm it’s the correct part, and the varistor’s orientation for proper assembly on the circuit board.
Finally, vision capabilities can identify the correct location on the circuit board where the varistor will be installed and soldered in place. During the soldering process, vision is used to monitor conditions such as the wetting area, lead lengths, solder balls, contact angles and solder fill.
Machine vision also aids in similar processes that require consistency and precision, like thermal paste dispensation application and metal shield insertion. Dispensing sealant is another example; a vision system can reference the expected path and compare it to the achieved result.
Final assembly. Fastening methods, such as screwdriving, also are frequently used to assemble products. This process can be more consistent with a vision system. Machine vision can be used to identify the screw hole and to navigate the robot with the screwdriver accurately to the specified location. Once complete, a vision system can verify that the screw has been fastened properly. If there is more than one screw, the vision system can be used to navigate to additional holes and implement the screwing process as many times as needed.
Machine vision can also be useful in component feeding, guiding detailed part assembly, part sorting and closed-loop activities such as DIMM card insertion using force feedback control. (any process where human inputs are not needed, and the system is designed to achieve known outcomes).
The ROI of Machine Vision
To achieve next-generation automation and fully realize the benefits of the technology, machine vision on an assembly line isn’t just “nice to have”—it’s vital. This technology enables diverse capabilities for inspection, robot navigation and quality control, and the use cases for machine vision are continuing to expand in scope and complexity as the technology continues to evolve.
That translates to tremendous potential for manufacturers, as vision solutions for automated lines can improve production capacity, production stability and production yield. As both the hardware and the algorithms involved with machine vision continue to improve, so does the ROI for manufacturers who invest in these solutions.
Assaf Eden is director of product management at Bright Machines (http://brightmachines.com), a full-stack technology company offering software- and AI-driven solutions to automate product assembly, manufacturing operations and production execution on the factory. He's based in Tel Aviv.