Machine vision has always been a business revolving around big data, acquiring and processing countless images measured in gigabytes, and then extracting the information needed to make decisions for a specific object or task. Applications such as remote sensing and network inspection generate massive amounts of data; gigabytes per minute quickly become terabytes, or even petabytes.
As data streams grow in size and volume, many industries are seeking offline computing and storage solutions, moving into the cloud. But will cloud computing be fast enough for machine vision applications? Is the quality of service sufficient for industrial applications? As machine vision expands beyond the factory floor, the answer to these questions is increasingly yes, even for industrial applications.
As data traffic increases and its volume grows, many industries are seeking cloud computing and storage solutions. Moving to the cloud is clearly a wise choice. But can cloud computing respond quickly enough for machine vision applications? Is the quality of service sufficient for industrial applications? As machine vision extends beyond the factory floor, the answers are increasingly affirmative, even for industrial applications.
Machine vision enters the cloud
“Without some form of cloud or Internet of Things (IoT) integration, it would be very challenging to collect and manage large amounts of image data in a highly disciplined manner,” said Darcy Bachert, founder and CEO of Prolucid Technologies. “The significant progress made in core cloud technologies over the past five years has made all of this possible.”
Industry giants like Google, Microsoft, and Amazon have already invested heavily in cloud technology, developing large-scale storage and analytics technologies that simultaneously ensure information security. "IBM developed a protocol called MQTT, specifically designed for interfacing with low-power distributed devices, aiming to ensure quality of service while guaranteeing data transmission of any type," Bachert said.
Beyond their massive storage and computing power, public cloud providers also offer machine learning and deep learning services. One example is TensorFlow, a framework widely used in machine vision for deep learning research and application development. From advanced disease detection to managing greater product diversity on the production line, deep learning has demonstrated its potential in every area.
These open-source tools, along with advancements in imaging and image-based models, mean that "instead of hiring a team of PhDs and data scientists, you can now leverage an ever-evolving dataset to train these models and extract value in a much simpler way," Bachert said.
Bachert estimates that half of the machine vision projects its company develops incorporate cloud components. The largest-scale vision and imaging processing implemented in the cloud is likely in the medical device industry. Prolucid is working with a client using ultrasound-based imaging equipment to acquire images and provide classification values such as population genetics and general location.
To protect patient privacy, Prolucid employs several security strategies when collecting and analyzing data from medical imaging devices. One step is to "de-identify" or remove personal information such as name, date of birth, and postal code.
In addition, Prolucid has developed a strategy to protect data security during transit or at rest, alerting customers to remediation when data breaches are detected at the data center and device levels, identifying other defects, and recovering data in the event of a catastrophic breach.
The combination of cloud and edge computing
In a manufacturing environment, machine vision in the cloud has raised concerns about internet bandwidth and latency, which could slow down inspection processes, cause data loss, and potentially pose safety risks to equipment and workers. “With machine learning, you still have a real-time inspection process,” Bachert said. “What changes is how you address it.”
For example, in defect classification applications, manufacturers might use the cloud to collect, classify, and validate datasets, and develop a machine learning model. This model is then taken down from the cloud and applied to a real-time process at the cloud edge. This means that real-time analysis can be performed at the network edge, close to the data source, such as in a manufacturing workshop.
“Therefore, we don’t need to worry about latency,” Bachert said. “In every system we’ve designed, the real-time process components can run whether or not there is a cloud connection.” This hybrid approach of cloud and edge computing represents a potential direction for machine vision integration.
The successful application of cloud computing in certain fields has spurred more manufacturing companies to try it out. "Ten years ago, no one would have foreseen the realization of self-driving cars, but now the public widely recognizes that the data used to operate these vehicles is collected and processed in the cloud," said Andy Long, CEO of CythSystems. "We've had conversations with customers who said, 'We don't know what we want to do, but our executive team told us we have to find a way to invest in this disruptive technology.'"
Deep expansion of IoT
As manufacturers seek to automate more processes, cloud-based machine vision manufacturing systems will play an increasingly important role. “We’ve done many assembly verification projects for our clients with the goal of providing a system that doesn’t require any programming, but instead uses artificial intelligence and cloud-based processors to do all the work,” Long said. “People who used to manually inspect parts are now responsible for training the system, teaching it how to identify good and bad parts. It doesn’t require any machine vision knowledge itself.”
Using the cloud to simplify machine vision implementation also gives manufacturers unprecedented freedom to use the technology for equipment testing and commissioning. “The front-end analysis is much faster than the programming you do in a traditional machine vision system,” Long said. “You can now experiment faster and more frequently to determine if the technology can solve a problem.”
Even if manufacturers refuse to analyze their imaging data over the internet, they will use the cloud in other ways—most notably for remote monitoring. For example, Omron Microscan Systems offers an interface called CloudLink that allows users to monitor real-time machine vision inspections via the web. Additionally, ImpactVision Technologies provides related machine vision products that enable remote monitoring of customers' vision system performance, changes to inspection standards, and maintenance.
The Triniti lighting controller from Gardsoft Vision is another excellent example demonstrating how IoT is extending not only to every corner of the factory but also to machine vision systems. The network-enabled Triniti controller provides intelligent and precise control over lighting systems and operations, including both fixed and variable data such as lighting attributes, model information, usage information, and optical and electrical characteristics. Compatibility with the GenICam and GigEVision standards allows for easy integration with other system components and facilitates the download of part numbers from the factory host.
“Parameters such as maximum operating temperature, duty cycle, and usage hours are crucial for proper maintenance or utilization of the luminaires,” said John Merva, Vice President of Gardsoft North America. “Triniti allows users to easily access information to make optimal decisions regarding the overall performance of the lighting and machine vision systems themselves.”
Use more cloud technology
Whether inside or outside manufacturing environments, machine vision is increasingly utilizing cloud-based off-site computing and storage solutions. In the past, these processes were performed internally for cost and security reasons. Now, all of that is quietly changing.
Machine vision systems can connect to remote cloud computing and storage devices, increasing the capabilities of any machine vision system, even in industrial applications. Applications such as quality inspection and sorting have greatly benefited from this trend.
Machine vision requires processing and analyzing massive amounts of image data, especially for industrial applications. Over the years, this data volume has ballooned to an unmanageable size, forcing companies to turn to offline cloud computing and storage solutions. Without cloud technology, machine vision systems are naturally limited by the amount of data they can process. Connecting to cloud solutions unlocks much of their commercial potential.
Increase machine vision connectivity
Cloud technology has significantly accelerated front-end analytics, enabling manufacturers to validate new, more connected solutions. The realization of the Industrial Internet of Things (IIoT) is driven by the speed and connectivity offered by cloud solutions. For example, remote monitoring is an early application of IIoT leveraging cloud technology to improve manufacturing operations. Machine vision inspection can be monitored in real-time via the network, allowing for performance analysis, maintenance, and modification of inspection standards.
The Internet of Things (IIoT) is expanding into every corner of the factory. Machine vision systems connected to the cloud offer unprecedented connectivity, enabling entirely new types of vision applications, especially in industrial inspection environments. Cloud technology is becoming increasingly important for machine vision systems.
Without cloud computing and storage, many of today's ongoing applications would be impossible. As more and more data is collected and processed in industrial environments, machine vision will become increasingly reliant on cloud technology to meet the demands of future smart manufacturing.
Disclaimer: This article is a reprint. If it involves copyright issues, please contact us promptly for deletion (QQ: 2737591964 ) . We apologize for any inconvenience.
