Seeing Computer Vision in a New Light

April 28, 2023 | 5 min read

Barry Clark, Director of Robotic Software, Bright Machines

Modern technology is ushering the manufacturing industry into a new period of innovation, and at the center is the rapidly growing field of artificial intelligence (AI). More manufacturers are turning to technological advancements, advanced analytics, and data-driven solutions to improve efficiency, streamline production, and make operations more cost-effective.

In particular, computer vision – which has been studied for more than a half century – is experiencing a new spike in advancements. Like other branches of AI, computer vision began as a field of research that sought to mimic human behavior – in this case, perception – in software and robotics. Over time, it has evolved into an interdisciplinary field dedicated to algorithms and software that allow computers to automatically extract and analyze useful information from digital images and video.

The Basics of Computer Vision

Unlike the related topics of machine vision and digital image processing, the goal of computer vision is broader, enabling a “full-scene understanding” of imagery or a 360-degree view. If we imagine the field of AI as the entirety of the brain, then computer vision would be the visual cortex (where visual information is processed). While this comparison is accurate enough, it fails to capture the full complexity and potential of this rapidly growing field.

Digital image processing refers to a broad branch of signal processing applied to images. It includes the general application of pattern recognition, feature extraction, and classification and represents a set of tools used in computer vision. Machine vision is a subset of computer vision concerned with the engineering or industrial use of vision for automatic inspection, process control, and robot guidance. Computer vision is a larger field than machine vision and is comprised of more fields than just signal processing, including machine learning.

Driving Efficiency and Flexibility

Machine learning and, in particular, deep learning are large driving forces for technological improvements in computer vision. And as machine learning becomes more prevalent across industries with each passing year, so do unique applications of computer vision.

The applications are growing, with uses ranging from consumer electronics, autonomous driving, and robotics to retail and smart agriculture. In agriculture, for instance, computer vision running on images taken from overhead drones can help farmers and farm equipment identify weeds, disease, and even estimate soil makeup and moisture content. AI-enabled farm equipment can accurately pick ripe crops without harming them while moving past crops that need more time to mature.

For the manufacturing and automation industry, computer vision has the potential to transform production and quality control and allow for greater efficiency and flexibility.

Transforming the Factory Floor

We are starting to see computer vision implemented more broadly in two categories: inspection and navigation. While many companies strive for solutions, the inspection process typically ends with a human operator verifying the final result. The next wave of computer vision promises to automate this process, allowing it to happen at various points along the production line. By catching errors earlier and more consistently, computer vision will help reduce costs and improve assembly efficiency. It also has the potential to increase uptime, reduce manufacturing flaws, and decrease the amount of custom hardware required.

Similarly, computer vision for navigation has largely been avoided unless absolutely necessary in the past. As the new wave of computer vision reaches the factory floor, many of the problems experienced with navigation in the past (i.e., drift over time, sensitivity to environmental conditions, etc.) will disappear, making its use in assembly automation more ubiquitous.

Another benefit is the way computer vision can help bring manufacturing closer to the consumer, also known as reshoring (or nearshoring). According to the 2021 Thomas Industrial Survey, the pandemic and its impact on supply chains have increased interest in reshoring. Of the manufacturers surveyed, 83% indicated that they were likely to reshore, an increase from 54% in March 2020.

In addition to supply chain management, automation and machine learning applications, including computer vision, can significantly offset labor costs. In this vein, computer vision technology can enable “micro-factories” or smaller operations that produce quality products closer to the consumer base.

Barriers to Implementation

Naturally, the most prevalent barriers to implementing computer vision are similar to other emerging technologies: cost, complexity, and environmental sensitivity. It remains expensive to upgrade factory operations with computer vision, from the cameras to the computer systems it requires. Moreover, incorporating the technology into the assembly line requires the services and oversight of skilled experts. While there are many computer vision experts in the field today, they are far from prevalent, and most industry technicians don’t have the necessary training. However, this challenge may also represent an opportunity for reskilling and upskilling in the workforce.

But with current advancements, these issues will become problems of the past. And once these and other barriers are overcome, widespread implementation on the factory floor will follow.

Driving Resilience with Computer Vision

As computer vision continues to grow and improvements are made, the associated costs of implementing it will come down significantly. Similarly, as the technology becomes more mainstream, training and expertise in the field will as well. In the meantime, organizations that take the initiative to train employees to service computer vision software and hardware will be ahead of the curve as the technology becomes more widespread.

Ongoing research and development will also address technical issues, allowing for systems that are easier to use, more capable, less sensitive, and more adaptive. This work is well underway and is being driven by both new technology and consumer demand.

Above all, the growing reliance of industry on machine learning will fuel the demand for more sophisticated visual systems. As computer vision becomes more common, cost-effective, and easier to use, it will become a regular part of automation. For manufacturers, systems that can not only see but also recognize and understand will become an indispensable part of assembly solutions.

To learn more about our capabilities in building the backbone of AI, visit Bright Machines.

Learn more

The Role of Upskilling in an Automated Future
Reskilling

The Role of Upskilling in an Automated Future

Empowering Women: A Spotlight on Bright Machines' Female Engineers
Bright Ideas

Empowering Women: A Spotlight on Bright Machines' Female Engineers

A New Paradigm for the ‘AI Backbone’
Manufacturing Innovation

A New Paradigm for the ‘AI Backbone’