Tag Archives: machine learning

Are Traasdahl, Crisp
FST Soapbox

How a History of Slow Technology Adoption Across Food Supply Chains Nearly Broke Us

By Are Traasdahl
1 Comment
Are Traasdahl, Crisp

The COVID-19 crisis has exacerbated existing disconnects between food supply and demand. While some may be noticing these issues on a broader scale for the first time, the reality is that there have been challenges in our food supply chains for decades. A lack of accurate data and information sharing is the core of the problem and had greater impact due to the pandemic. Outdated technologies are preventing advancements and efficiencies, resulting in the paradox of mounting food insecurity and food waste.

To bridge this disconnect, the food industry needs to implement innovative AI and machine learning technologies to prevent shortages, overages and waste as COVID-19 subsides. Solutions that enable data sharing and collaboration are essential to build more resilient food supply chains for the future.

Data-sharing technologies that can help alleviate these problems have been under development for decades, but food supply chains have been slow to innovate compared to other industries. By reviewing the top four data-sharing technologies used in food industry and the year they were introduced to food supply chains, it’s evident that the pace of technology innovation and adoption needs to accelerate to advance the industry.

A History of Technology Adoption in the Food Industry

The Barcode – 19741
We’re all familiar with the barcode—that assemblage of lines translated into numbers and letters conveying information about a product. When a cashier scans a barcode, the correct price pops up on the POS, and the sale data is recorded for inventory management. Barcodes are inexpensive and easy to implement. However, they only provide basic information, such as a product’s name, type, and price. Also, while you can glean information from a barcode, you can’t change it or add information to it. In addition, barcodes only group products by category—as opposed to radio-frequency identification (RFID), which provides a different code for every single item.

EDI First Multi-Industry Standards – 19812
Electronic data interchange (EDI) is just what it sounds like—the concept of sharing information electronically instead of on paper. Since EDI standardizes documents and the way they’re transferred, communication between business partners along the supply chain is easier, more efficient, and human error is reduced. To share information via EDI, however, software is required. This software can be challenging for businesses to implement and requires IT expertise to handle updates and maintenance.

RFID in the Food Supply Chain – 20033
RFID and RFID tags are encoded with information that can be transmitted to a reader device via radio waves, allowing businesses to identify and track products and assets. The reader device translates the radio waves into usable data, which then lands in a database for tracking and analysis.

RFID tags hold a lot more data than barcodes—and data is accessible in remote locations and easily shared along the supply chain to boost transparency and trust. Unlike barcode scanners, which need a direct line of sight to a code, RFID readers can read multiple tags at once from any direction. Businesses can use RFID to track products from producer to supplier to retailer in real time.

In 2003, Walmart rolled out a pilot program requiring 100 of its suppliers to use RFID technology by 2005.3 However, the retail giant wasn’t able to scale up the program. While prices have dropped from 35–40 cents during Walmart’s pilot to just 5 cents each as of 2018, RFID tags are still more expensive than barcodes.4 They can also be harder to implement and configure. Since active tags have such a long reach, businesses also need to ensure that scammers can’t intercept sensitive data.

Blockchain – 20175
A blockchain is a digital ledger of blocks (records) used to record data across multiple transactions. Changes are recorded in real-time, making the history unfalsifiable and transparent. Along the food supply chain, users can tag food, materials, compliance certificates and more with a set of information that’s recorded on the blockchain. Partners can easily follow the item through the physical supply chain, and new information is recorded in real-time.

Blockchain is more secure and transparent, less vulnerable to fraud, and more scalable than technologies like RFID. When paired with embedded sensors and RFID tags, the tech offers easier record-keeping and better provenance tracking, so it can address and help solve traceability problems. Blockchain boosts trust by reducing food falsification and decreasing delays in the supply chain.6

On the negative side, the cost of transaction processing with blockchain is high. Not to mention, the technology is confusing to many, which hinders adoption. Finally, while more transparency is good news, there’s such a thing as too much transparency; there needs to be a balance, so competitors don’t have too much access to sensitive data.

Cloud-Based Demand Forecasting – 2019 to present7
Cloud-based demand forecasting uses machine learning and AI to predict demand for various products at different points in the food supply chain. This technology leverages other technologies on this list to enhance communication across supply chain partners and improve the accuracy of demand forecasting, resulting in less waste and more profit for the food industry. It enables huge volumes of data to be used to predict demand, including past buying patterns, market changes, weather, events and holidays, social media input and more to create a more accurate picture of demand.

The alternative to cloud-based demand forecasting that is still in use today involves Excel or manual spreadsheets and lots of number crunching, which are time-intensive and prone to human error. This manual approach is not a sustainable process, but AI, machine learning and automation can step in to resolve these issues.

Obtaining real-time insights from a centralized, accurate and accessible data source enables food suppliers, brokers, distributors, brands and retailers to share information and be nimble, improving their ability to adjust supply in response to factors influencing demand.8 This, in turn, reduces cost, time and food waste, since brands can accurately predict how much to produce down to the individual SKU level, where to send it and even what factors might impact it along the way.

Speeding Up Adoption

As illustrated in Figure 1, the pace of technology change in the food industry has been slow compared to other industries, such as music and telecommunications. But we now have the tools, the data and the brainpower to create more resilient food supply chains.

Technology adoption, food industry
Figure 1. The pace of technology change in the food industry has been slow compared to other industries. Figure courtesy of Crisp.

Given the inherent connectivity of partners in the food supply chain, we now need to work together to connect information systems in ways that give us the insights needed to deliver exactly the rights foods to the right places, at the right time. This will not only improve consumer satisfaction but will also protect revenue and margins up and down food supply chains and reduce global waste.

References

  1. Weightman, G. (2015). The History of the Bar Code. Smithsonian Magazine.
  2. Locken, S. (2012). History of EDI Technology. EDI Alliance.
  3. Markoff, R, Seifert, R. (2019). RFID: Yesterday’s blockchain. International Institute for Management Development.
  4. Wollenhaupt, G. (2018). What’s next for RFID? Supply Chain Dive.
  5. Tran, S. (2019). IBM Food Trust: Cutting Through the Complexity of the World’s Food Supply with Blockchain. Blockchain News.
  6. Galvez, J, Mejuto, J.C., Simal-Gandara, J. (2018). Future Challenge on the use of blockchain for food traceability analysis. Science Direct.
  7. (2019). Crisp launches with $14.2 million to cut food waste using big data. Venture Beat.
  8. Dixie, G. (2005). The Impact of Supply and Demand. Marketing Extension Guide.
Megan Nichols
FST Soapbox

Machine Vision Training Tips to Improve Food Inspections

By Megan Ray Nichols
1 Comment
Megan Nichols

As machines become more intelligent, every industry on earth will find abundant new applications and ways to benefit. For the food industry, which has an incredible number of moving parts and is especially risk-averse, machine vision and machine learning are especially valuable additions to the supply chain.

The following is a look at what machine vision is, how it can play a role in manufacturing and distributing foods and beverages, and how employers can train workers to get the most out of this exciting technology.

What Is Machine Vision?

Machine vision isn’t a brand-new concept. Cameras and barcode readers with machine vision have long been capable of reading barcodes and QR codes and verifying that products have correct labels. Modern machine vision takes the concept to new levels of usefulness.

Barcodes and product identifiers have a limited set of known configurations, which makes it relatively straightforward to program an automated inspection station to recognize, sort or reject products as necessary. Instead, true machine vision means handlers don’t have to account for every potential eventuality. Machine vision instead learns over time, based on known parameters, to differentiate between degrees of product damage.

Consider the problem of appraising an apple for its salability. Is it bruised or discolored? Machine vision recognizes that no two bruises look precisely alike. There’s also the matter of identifying different degrees of packaging damage. To tackle these problems, it’s not possible to program machine vision to recognize a fixed set of visual clues. Instead, its programming must interpret its surroundings and make a judgment about what it sees.

Apples, machine vision
On an apple, no two bruises are alike. Machine vision technology can help. Photo credit: Pexels.

The neural networks that power machine vision have a wide range of applications, including improving pathfinding abilities for robots. In this article, I’ll focus on how to leverage machine vision to improve the quality of edible products and the profitability of the food and beverage industry.

Applications for Machine Vision in the Food Industry

There are lots of ways to apply machine vision to a food processing environment, with new variations on the technology cropping up regularly. The following is a rundown on how different kinds of machine vision systems serve different functions in the food and beverage sector.

1. Frame Grabbing and 3-D Machine Vision
Machine vision systems require optimal lighting to carry out successful inspections. If part of the scanning environment lies in shadow, undesirable products might find their way onto shelves and into customers’ homes.

Food products sometimes have unique needs when it comes to carrying out visual inspections. It’s difficult or impossible for fallible human eyeballs to perform detailed scans of thousands of peas or nuts as they pass over a conveyor belt. 3-D machine vision offers a tool called “frame grabbing,” which takes stills of — potentially — tens of thousands of tiny, moving products at once to find flaws and perform sorting.

2. Automated Sorting for Large Product Batches
Machine vision inspection systems can easily become part of a much larger automation effort. Automation is a welcome addition to the food and beverage sector, translating into improved worker safety and efficiency and better quality control across the enterprise.

Inspection stations with machine vision cameras can scan single products or whole batches of products to detect flaws. But physically separating these products must be just as efficient a process as identifying them. For this reason, machine vision is an ideal companion to compressed air systems and others, which can carefully blow away and remove even a single grain of rice from a larger batch in preparation.

3. Near-Infrared Cameras
Machine vision takes many forms, including barcode and QR code readers. A newer technology, called near-infrared (NIR) cameras, is already substantially improving the usefulness and capabilities of machine vision.

Remember that bruised apple? Sometimes physical damage to fruits and vegetables doesn’t immediately appear on the outside. NIR technology expands the light spectrum cameras can observe, giving them the ability to detect interior damage before it shows up on the exterior. It represents a distinct advantage over previous-generation technology and human inspectors, both of which can leave flaws undiscovered.

Tips on Training Workers to Use Machine Vision

Implementing machine vision into a productive environment delivers major benefits, but it also comes with a potentially disruptive learning curve. The following are some ideas on how to navigate it.

1. Take Advantage of Third-Party Training Courses
Don’t expect employees to hit the ground running with machine vision if they’re not familiar with the fundamentals of how it works. Google has a crash course on machine learning, and Amazon offers a curriculum as well to help companies get their employees up to speed on the technology and how to use it.

2. Get the Lighting Right
Having the appropriate intensity of light shining on the food product is essential for the machine vision cameras to get a clear photo or video. The most common types of lighting for machine vision are quartz halogen, LEDs, metal halide and xenon lights. Metal halide and xenon are better for larger-scale operations because of their brightness.

Train employees to check the amount and positioning of the lighting before each inspection station starts up for the day, so that no shadows obscure products from view.

3. Single Out Promising Subject Matter Leaders
Companies today don’t seem to have much confidence in how well they’re preparing their workforce for tomorrow, including future innovations. According to Deloitte, just 47% of companies in the world believe they’re doing enough to train their employees on the technologies and opportunities of Industry 4.0.

Machine vision does not involve buying a camera or two, setting them up, then slapping the “autopilot” button. As products turn over, and manufacturing and distribution environments change and grow over time, machine vision algorithms require re-training, and you might need to redesign the lighting setup.

Employers should find individuals from their ranks who show interest and aptitude in this technology and then invest in them as subject matter experts and process owners. Even if an outside vendor is the one providing libraries of algorithms and ultimately coming up with machine vision designs, every company needs a knowledgeable liaison who can align company needs with the products on the market.

Machine Vision Is the Future of Food Inspections

The market for machine vision technology is likely to reach $30.8 in value by 2021, according to BCC Research.

It is important to remember that neither machine learning nor machine vision are about creating hardware that thinks and sees like humans do. With the right approach, these systems can roundly outperform human employees.

But first, companies need to recognize the opportunities. Then, they must match the available products to their unsolved problems and make sure their culture supports ongoing learning and the discovery of new aptitudes. Machine vision might be superior to human eyesight, but it uses decidedly human judgments as it goes about its work.

Sasan Amini, Clear Labs

2020 Expectations: More Artificial Intelligence and Machine Learning, Technology Advances in Food Safety Testing

By Maria Fontanazza
No Comments
Sasan Amini, Clear Labs

2018 and 2019 were the years of the “blockchain buzz”. As we enter the new decade, we can expect a stronger focus on how technology and data advances will generate more actionable use for the food industry. Food Safety Tech has highlighted many perspectives from subject matter experts in the industry, and 2020 will be no different. Our first Q&A of the year features Sasan Amini, CEO of Clear Labs, as he shares his thoughts on tech improvements and the continued rise consumer expectations for transparency.

Food Safety Tech: As we look to the year ahead, where do you see artificial intelligence, machine learning and blockchain advancing in the food industry?

Sasan Amini: AI, ML, and blockchain are making headway in the food industry through advances in supply chain management, food sorting and anomaly detection, and tracing the origin of foodborne outbreaks. On the regulatory side, FDA’s focus on its New Era of Smarter Food Safety will most likely catalyze the adoption of the above mentioned technologies. On the private side, a few of the companies leading the charge on these advancements are IBM and Google, working in partnership with food manufacturers and retailers across the world.

Along those same lines, another area that we expect to grow is the use of AI and ML in tandem with robotics—and the value of new troves of data that they collect, analyze and distribute. For example, robotics for the use of environmental monitoring of potential contaminants, sorting techniques and sterilization are valuable because they ensure that end products have been through thorough testing—and they give us even more information about the lifecycle of that food than ever before.

At the end of the day, data is only valuable when you can transform it into actionable insights in real-time with real-world applications, and we expect to see more and more of this type of data usage in the year ahead.

FST: Where do you think food safety testing technologies will stand out? What advancements can the industry expect?

Amini: In 2020, technology is going to begin to connect itself along the entire supply chain, bringing together disparate pieces and equipping supply chain professionals with action-oriented data. From testing advances that improve speed, accuracy and depth of information to modular software solutions to promote transparency, the food safety industry is finally finding its footing in a data-driven sea of technological and regulatory advances.

Right now, legacy testing solutions are limited in their ability to lead food safety and quality professionals to the source of problems, providing insights on tracking recurring issues, hence having a faster response time, and being able to anticipate problems before they occur based on a more data heavy and objective risk assessment tools. This leaves the industry in a reactive position for managing and controlling their pathogen problems.

Availability of higher resolution food safety technologies that provide deeper and more accurate information and puts them in context for food safety and quality professionals provides the food industry a unique opportunity to resolve the incidents in a timely fashion with higher rigour and confidence. This is very in-line with the “Smarter Tools and Approaches” that FDA described in their new approach to food safety.

FST: How are evolving consumer preferences changing how food companies must do business from a strategic as well as transparency perspective?

Amini: Consumers are continuing to get savvier about what’s in their food and where it comes from. Research suggests that about one in five U.S. adults believe they are food allergic, while only 1 in 20 are estimated to have physician-diagnosed food allergies. This discrepancy is important for food companies to consider when making decisions about transparency into their products. Although the research on food allergies continues to evolve, what’s important to note today is that consumers want to know the details. Radical transparency can be a differentiator in a competitive market, especially for consumers looking for answers to improve their health and nutrition.

Consumers are also increasingly interested in personalization, due in part to the rise in new digital health and testing companies looking to deliver on the promise of personalized nutrition and wellness. Again, more transparency will be key.

FST: Additional comments are welcome.

Amini: Looking ahead, we expect that smaller, multi-use, and hyper-efficient tools with reduced physical footprints will gain market share. NGS is a great example of this, as it allows any lab to gather millions of data points about a single sample without needing to run it multiple times. It moves beyond the binary yes-no response of traditional testing, and lets you get much more done, with far less. Such wealth of information not only increases the confidence about the result, but can also be mined to generate more actionable insights for interventions and root cause analysis.

This “multi-tool” will be driven by a combination of advanced software, robotics, and testing capabilities, creating a food safety system that is entirely connected, driven by data, and powerfully accurate.

Ned Sharpless, Frank Yiannas, FDA

FDA’s ‘New Era of Smarter Food Safety’ to Focus on Traceability, Digital Technology and E-Commerce

By Food Safety Tech Staff
No Comments
Ned Sharpless, Frank Yiannas, FDA

“It’s time to look to the future of food safety once again,” declared Acting FDA Commissioner Ned Sharpless, M.D. and Deputy Commissioner for Food Policy and Response Frank Yiannas in a press statement released yesterday. Although progress has been made in implementing FSMA and with the development of the GenomeTrakr Network, the agency wants to move forward in taking advantage of the innovative technologies that will help make the food supply more digital, traceable and safer. With that effort comes the creation of a “Blueprint for a New Era of Smarter Food Safety”, which will speak to “traceability, digital technologies and evolving food business models”. Sharpless and Yiannas outlined the significant role that these components will play.

Digital technology in food traceability. Digital technologies could play a crucial part in rapidly identifying and tracing contaminated food back to its origin—changing the timespan from days or weeks to minutes or seconds. FDA intends to look at new ways that it can evaluate new technologies and improve its ability to quickly track and trace food throughout the supply chain. “Access to information during an outbreak about the origin of contaminated food will help us conduct more timely root cause analysis and apply these learnings to prevent future incidents from happening in the first place,” stated Sharpless and Yiannas. This means a shift away from paper-based systems.

Ned Sharpless, Frank Yiannas, FDA
(left to right) Ned Sharpless, M.D., FDA acting commissioner and Frank Yiannas, deputy commissioner of food policy and response. Image courtesy of FDA

Emerging technologies. Artificial intelligence (AI), distributed ledgers (no, they didn’t directly say “blockchain”), the Internet of Things, sensors and other emerging technologies could enable more transparency within the supply chain as well as consumer side of things. The FDA leaders announced a pilot program that will use AI and machine learning to assess food imports at the U.S. point of entry.

E-Commerce. “Evolving food business models”, also known as e-commerce, is growing fast and changing how consumers get their food. With food delivery introduces food safety issues such as those related to packaging and temperature control. FDA is exploring how it can collaborate with federal, state and local stakeholders to figure out ways to address these potential problems.

Sharpless and Yiannas emphasized the end-goal in keeping the food of American consumers safe. “So, welcome to the new era of smarter food safety that is people-led, FSMA-based and technology-enabled!”