Editor’s note: Devendra Chandani is co-founder and head of US at Intello Labs, based in New Jersey. Headquartered in Gurugram, India, Intello Labs uses machine learning tech to grade the quality of agricultural produce. The views expressed in this article are the author’s own.
Fresh fruits and vegetables are a critical ingredient for food companies that make anything from juices and smoothies through to sauces, pastes, and pulps.
The characteristics of fresh produce differ by variety and season, unlike with many other raw materials. This variability creates challenges in the manufacturing process, and in delivering the quality customers expect in finished products.
Measuring quality in fresh produce
Defining fresh produce quality parameters and their measurement for acceptance is more complex compared to other raw ingredients. Uniformity, ripeness, color, gloss, and absence of defects are just some of the components of quality. Evaluating them can be subjective — in other words, done manually — or objective, when handled by a machine.
But the visual appearance of fruits and vegetables is universally accepted as the first quality determinant.
Color is an important parameter because it directly affects the appearance of the final product. If there is color variance in fresh produce, corrective actions in the manufacturing process become necessary to ensure the perception of quality in the final product.
In short, failure to meet fresh produce color criteria can adversely impact cost and quality in the finished product. This applies to a wide range of food industry items.
Obstacles to identifying color
The biggest challenge to determining the color of fresh produce is heterogeneity. A fruit or vegetable may not be one color, but different shades of the same color. A single piece could also feature different colors blending into one another.
A good example here is tomatoes. A single tomato may have red, orange, and light green colors and still be acceptable as an ingredient. Another example is apples, which may have different shades of red depending on variety.
Traditional solutions
The color heterogeneity in fresh produce makes it difficult to set and measure acceptable criteria. Right now, the food industry utilizes a battery of equipment to measure color in both ingredients and finished products. Most often, this equipment includes colorimeters, spectrophotometers, and image-based color identification machines.
Though made for myriad food (and non-food) ingredients, these traditional instruments mostly fail in measuring acceptable color criteria for fresh produce:
- Colorimeters are best for gauging homogeneous solutions or solids. When faced with heterogeneous solids they, at best, can indicate an average color. This often results in inaccurate identification.
- Spectrophotometers are most appropriate for input in a liquid medium and for homogeneous products.
- Equipment that processes high-quality images in a controlled environment appears to be the ideal choice, at least among existing solutions for fresh produce. These go beyond average color measurement to offer percentage composition of colors.
The latter can detect, for example, that a displayed tomato is 80% red and 20% light green. But when shown five separate tomatoes together, where one appears light green and the other four red, such instruments fail. Since the combined composition is still 80% red and 20% light green, this will often lead to the one light green tomato getting accepted erroneously.
Innovative solutions
In the last few years, artificial intelligence (AI) has widened the lead it has over all prior technologies for accuracy in visual identification of fresh produce. It has also surprised us in the plethora of tasks it excels at within image processing. It has beaten previous benchmarks in object detection, classification, and segmentation.
To explain these three tasks, let’s take the example of tomatoes again:
- Object detection is the ability to identify different objects in an image. In this case, all the individual tomatoes that are presented.
- Classification is the ability to organize an image into one of many classes. In this instance, the classification of a tomato can be 80% red, 30% red, and so on.
- Segmentation is the ability to distinguish the composition of an image into the different features that are present in it. For example, the tomatoes, the table they are on, the piece of paper next to them, and so on.
Two factors enable AI to achieve excellence in these tasks. First, a large dataset of images labeled to meet acceptable and non-acceptable parameters of fresh produce. Second, sizable amounts of computing ability in the form of powerful microprocessors.
In a nutshell, AI fixes the issues riddling color for fresh produce accurately and much better than ever before.
Advantages of AI
AI-based solutions have the unerring ability to eliminate color issues in a heterogeneous scenario and color acceptance criteria for almost all fresh produce. But that’s not the sole benefit. It can do so in a wide variety of lighting conditions.
Covid-19 has caused plenty of volatility in fresh produce prices. Read more here
On top of that, with cloud processing, the solution becomes highly portable. It means one can conveniently access this powerful technology through a handheld device like a mobile phone.
In addition, cloud computing allows easy integration of the solution with the rest of the user’s applications, such as email and messaging. These qualities make AI-based solutions far lighter on a user’s pocket than the bulky equipment traditionally used for color identification.
Color and quality control in fresh produce: Traditional solutions vs AI
September 4, 2020
Devendra Chandani
Editor’s note: Devendra Chandani is co-founder and head of US at Intello Labs, based in New Jersey. Headquartered in Gurugram, India, Intello Labs uses machine learning tech to grade the quality of agricultural produce. The views expressed in this article are the author’s own.
Fresh fruits and vegetables are a critical ingredient for food companies that make anything from juices and smoothies through to sauces, pastes, and pulps.
The characteristics of fresh produce differ by variety and season, unlike with many other raw materials. This variability creates challenges in the manufacturing process, and in delivering the quality customers expect in finished products.
Measuring quality in fresh produce
Defining fresh produce quality parameters and their measurement for acceptance is more complex compared to other raw ingredients. Uniformity, ripeness, color, gloss, and absence of defects are just some of the components of quality. Evaluating them can be subjective — in other words, done manually — or objective, when handled by a machine.
But the visual appearance of fruits and vegetables is universally accepted as the first quality determinant.
Color is an important parameter because it directly affects the appearance of the final product. If there is color variance in fresh produce, corrective actions in the manufacturing process become necessary to ensure the perception of quality in the final product.
In short, failure to meet fresh produce color criteria can adversely impact cost and quality in the finished product. This applies to a wide range of food industry items.
Obstacles to identifying color
The biggest challenge to determining the color of fresh produce is heterogeneity. A fruit or vegetable may not be one color, but different shades of the same color. A single piece could also feature different colors blending into one another.
A good example here is tomatoes. A single tomato may have red, orange, and light green colors and still be acceptable as an ingredient. Another example is apples, which may have different shades of red depending on variety.
Traditional solutions
The color heterogeneity in fresh produce makes it difficult to set and measure acceptable criteria. Right now, the food industry utilizes a battery of equipment to measure color in both ingredients and finished products. Most often, this equipment includes colorimeters, spectrophotometers, and image-based color identification machines.
Though made for myriad food (and non-food) ingredients, these traditional instruments mostly fail in measuring acceptable color criteria for fresh produce:
The latter can detect, for example, that a displayed tomato is 80% red and 20% light green. But when shown five separate tomatoes together, where one appears light green and the other four red, such instruments fail. Since the combined composition is still 80% red and 20% light green, this will often lead to the one light green tomato getting accepted erroneously.
Innovative solutions
In the last few years, artificial intelligence (AI) has widened the lead it has over all prior technologies for accuracy in visual identification of fresh produce. It has also surprised us in the plethora of tasks it excels at within image processing. It has beaten previous benchmarks in object detection, classification, and segmentation.
To explain these three tasks, let’s take the example of tomatoes again:
Two factors enable AI to achieve excellence in these tasks. First, a large dataset of images labeled to meet acceptable and non-acceptable parameters of fresh produce. Second, sizable amounts of computing ability in the form of powerful microprocessors.
In a nutshell, AI fixes the issues riddling color for fresh produce accurately and much better than ever before.
Advantages of AI
AI-based solutions have the unerring ability to eliminate color issues in a heterogeneous scenario and color acceptance criteria for almost all fresh produce. But that’s not the sole benefit. It can do so in a wide variety of lighting conditions.
On top of that, with cloud processing, the solution becomes highly portable. It means one can conveniently access this powerful technology through a handheld device like a mobile phone.
In addition, cloud computing allows easy integration of the solution with the rest of the user’s applications, such as email and messaging. These qualities make AI-based solutions far lighter on a user’s pocket than the bulky equipment traditionally used for color identification.
Join the Newsletter
Get the latest news & research from AFN and AgFunder in your inbox.
Related Stories
The Week in AgriFoodTech: Mondelez invests in cocoa tech, a $350m regen ag fund launches
Voltiris leverages spectral filtering to ‘transform the greenhouse into a solar field’ and cut costs for growers
The Week in AgriFoodTech: AI-powered protein, plant-based food dyes, and maggot robots
Japan rising through the ranks as an agrifoodtech hub after 58% increase in YoY funding
Get the latest news and research from AFN & AgFunder in your inbox.
Follow us:
Sponsored Content
Sponsored
Sponsored post: The innovator’s dilemma: why agbioscience innovation must focus on the farmer first
Editor's Pick
XPRIZE ‘Feed the next billion’ finalists learn there is no $10m grand prize: ‘It feels dishonest’
Frankly Speaking
Trump’s tariffs won’t help US agrifood industry, says ex-Congressman Charlie Dent: ‘There are no winners’
Data Snapshot
From novelty to necessity? The evolution of insect farming
Investor Insight
🎥Foodtech investing: ‘You can’t escape the fundamentals. You have to produce something that adds significant value’
Meet the Founder
BioDefense seeks to disrupt seafood market with ‘tasteless, odorless, and invisible’ coating that can boost shelf life 2-3x
Research & Data
Japan rising through the ranks as an agrifoodtech hub after 58% increase in YoY funding