How Machine Vision Continues to Increase Industrial Automation Efficiency

In April 2021, it was announced that the Robotics Industries Association (RIA), AIA – Advancing Vision + Imaging (AIA), and the Motion Control and Motors Association (MCMA) would merge under one umbrella as A3, the Association for Advancing Automation. Instead of focusing on separate components and businesses, the new association would look at automation from a holistic standpoint. Walk any automation tradeshow floor in 2023, and you might see why.

While attendees will see plenty of companies showcasing their innovations in these individual areas, many demonstrations — including several from an influx of newer companies — are highlighting technologies that incorporate the latest developments in robots, machine vision, and motion control together. (Autonomous mobile robots, for one example).

These individual technologies serve unique, valuable purposes on the factory floor and beyond. For instance, “blind” robots can still automate many repeatable, pre-programmed tasks across different industries, but pairing a robot with machine vision creates a much more flexible automation system with significantly more capabilities. As automation systems continue to evolve and progress, machine vision will remain a key enabler. Let’s look at some of the recent ways that machine vision has helped drive automation technologies forward.

3D Improvements

While 3D imaging has long served the industrial automation space, recent developments have enhanced existing applications while opening the door to new ones. New or enhanced 3D capabilities include lower noise, higher resolution, texture images in RGB, higher accuracy, and the ability to capture images of objects in motion at fairly high rates.

In addition, certain 3D imaging implementations have become less expensive and easier to use. For example, several of the application-specific automation systems on the market today — such as those purpose built for bin picking and general pick and place, palletizing and depalletizing, and logistics sorting — leverage a single 3D RGB-D camera. It captures a 2D color image and a depth measurement that can be combined to create RGB-D images, which are in turn used to help guide the robot’s movements.

Elsewhere, other 3D advancements have addressed some of today’s automation challenges, including:

Time-of-Flight (ToF): ToF cameras and sensors have made major strides in recent years, allowing them to meet requirements in logistics, autonomous robots, and other challenging factory automation applications.

High dynamic range imaging: 3D imaging systems today offer high dynamic range capabilities never seen before. This suits applications involving high- or low-reflectivity surfaces, such as automotive assembly, or in logistics and palletizing applications, where the objects or parts may vary greatly and become difficult to image with a fixed exposure time.

Factory-calibrated 3D profilers: Targeting ease-of-use for customers who require 3D vision capabilities, factory-calibrated, fully integrated 3D profilers offer intuitive setup and operation for imaging applications. This can range from automotive inspections involving small electronic parts or large automobile parts to packaging tasks including container and fill-level inspection and location, sortation, and volumetric measurements.

An AI Boost

Another interesting development in recent years has been the intersection of AI and 3D, along with the general niche that has been carved out for AI methods including deep learning and machine learning techniques. First, many of the application-specific systems mentioned above that leverage 3D imaging also deploy AI techniques as a complimentary but powerful tool that adds flexibility. For example, AI algorithms can help these systems individually identify highly variable items, allowing the robot to make a high-speed pick based on a training set — a task where traditional machine vision algorithms would struggle. These items may include anything from small consumer goods in a warehouse to individual chicken breasts moving at high speeds on a conveyor line.

Of course, AI has found its niche within machine vision beyond 3D imaging. Across many different industry verticals, there are still many manufacturing processes that rely heavily on human visual inspection. Augmenting a machine vision system with machine learning or deep learning techniques can allow the system to make subjective inspection decisions where a human may otherwise be required. Examples include anomaly detection, defect detection, classification, assembly verification, and more, which help companies get the most out of their automated inspection systems.

AI advancements have also made robotic automation easier and simpler to deploy. One of the most important value propositions for AI in the industrial automation space is the diversification of who can automate, according to Juan Aparicio, President, Robota Labs, who cited a recent MIT report on the future of work. This report found that very few robots existed at small to medium-sized manufacturers. As AI becomes easier to deploy — with several suppliers now offering low-code or no-code tools for AI development — some smaller manufacturers have started turning to AI as the tools become easier to use, which is, hopefully, a trend that will continue.

Easing Automation Adoption

Advancements in areas including sensors, cameras, industrial computing, lighting, and robotics will undoubtedly help drive automation forward and allow new applications to emerge, but in many ways, the next wave of improvements will be driven by software. This doesn’t just mean software that will offer new tools and capabilities targeting industrial automation, but software that eases integration and ultimately makes application development and deployment for the end user much easier.