Why Computer Vision In Manufacturing Works Best As Part Of An Industry 4.0 Ecosystem

Computer vision is quickly emerging as an integral part of the modern connected factory.

Though forms of computer vision have been in use in manufacturing for decades, recent advances in machine learning and image processing have enabled new manufacturing use cases. No longer limited to structured, repetitive tasks, AI-assisted computer vision platforms are capable of functioning in increasingly complex environments. They work continuously and in conjunction with operators, leading to improved efficiency, fewer errors, and better data.

But not all machine vision platforms will provide the same benefits. Computer vision solutions are most effective when integrated strategically into the smart factory, enhancing both digital and human performance on the line.

At Tulip, we’ve found that the best computer vision solutions share two key features:

1. They augment the operator–that is, they work with operators to make them more effective, accurate, and efficient
2. They function as part of an Industry 4.0 ecosystem.

In other words, in order to get the full benefits from computer vision, organizations should consider two things: 1.) How the technology will enable their operators and engineers, and, 2.) How their other technologies will work with a computer vision solution to unlock the technology’s full benefits.

[Free Ebook] Download our Ultimate Guide to Quality Management Systems →

What is machine vision?

Machine vision has historically referred to a subset of manufacturing applications for computer vision technology. Since the 1980s, systems of cameras, lights, reflectors, and software have helped organizations automate and optimize vision-based tasks. With robust applications in quality assurance, compliance, and inventory management, versions of machine vision have been crucial to manufacturing efficiently at scale. For example, machine vision systems can:

• Monitor the fill level of bottles on the bottling line, detecting deviations from standards and automatically initiating corrective action.
• Detecting quality defects in machined parts too subtle for the human eye
• Processing continuous visual information at a superhuman rate

These systems, though, are limited to repetitive, structured scenarios. Their cost–between cameras, software, and tightly controlled lighting–can be prohibitive. They were among the first systems to use early machine learning in manufacturing, but they require specialists to program and can be inflexible once deployed.

Broadly speaking, computer vision describes a number of methods by which computers extract, process, and analyze information from visual inputs. The applications of this technology are wide-reaching. Computer vision is behind the systems that:

• Help driverless cars process their surroundings
• Read features like age, gender, activity, and location from user photos on social media
• Guide operators through complex assemblies in a factory

In contrast to established machine vision systems, modern computer vision systems are more flexible. More than processing structured information in controlled settings, these new systems have an improved ability to function in semi- and unstructured scenarios. And they’re not limited to repetitive, automated processes.

Industry 4.0 computer vision systems are better able to work with operators, responding to their actions and enabling new forms of digital-human interaction.

With this in mind, how can organizations get the most out of a computer vision system?

Humans are central to manufacturing. Therefore, the best computer vision solutions work with operators to make them more effective, accurate, and efficient.

Modern computer vision solutions can achieve operator augmentation in several ways.

For one, Industry 4.0 computer vision solutions don’t just track operators as they execute a task. They also respond.

Such responses can take many forms. Computer vision systems can be trained on a manufacturer’s operations to recognize all of the gestures in a process, guiding an operator through complex work instructions as they accomplish each step. Computer vision can assist with quality checks in-line as the operator works by providing oversight at the level of the process. Some systems can even detect and classify operators' actions over time, giving engineers increased visibility into the timing and execution of every step.

What emerges from this physical-digital-physical feedback-loop between operator and computer vision are processes that are more efficient, less prone to error, and more visible through thorough, real-time data collection.

One constant in the discussion so far is that computer vision works best as part of a suite of integrated technologies.

Augmenting the operator requires more than a computer vision system collecting data on performance. Rather, it needs a framework like visual work instructions or SOPs to maximize efficiency. It needs other IoT-connected devices to facilitate communication and data exchange with an operator. And it can benefit from integration with other operations apps.

One of the promises of computer vision is continuous, real-time visibility into processes. It’s in the name. But the data collected by a computer vision system is more meaningful when placed in context with data from machines and other, non-visual operator data. While computer vision can provide a new source of data, it still requires a complete picture for context and validity.

In short: The best benefits of computer vision unlocked when the system can send signals to respond to and influence the actions of an operator in real-time. And for computer vision to interact with an operator in real-time, it needs supporting tech. It needs an ecosystem.

In its current state, computer vision is an important part of the digital factory. It gives operators and engineers a tool to collect better data, empower operators, and error-proof assemblies.

Those considering a digital transformation should keep in mind that computer vision works best when it augments the operator, and when it functions as part of an Industry 4.0 ecosystem.