The food industry retains enormous losses from misjudgment of food quality: every second fruit and vegetable produced will be thrown away due to mistaking its storage or shelf lifespan, or minor infections that have stayed unnoticed, and therefore, untreated. Employing machine learning and algorithms to determine the quality of the produce ensures more reliability for both offline and online grocery shopping, and a significant reduction in food waste, thus excluding the potential problem of good fruit being mistakenly thrown away, or expired produce being delivered to the customer.
A study conducted by the Food Marketing Institute and Grocery Manufacturers Association has shown that the average cost of a recall for food companies is about $10 million in direct costs, not to mention brand damage and lost sales. Microbiological contamination is reported to be the main reason for recall (47%) in 2016 in the U.S., the second being labeling issues (26%). Processing defects and physical contamination are responsible for 13% and 7% of recall cases correspondingly.
Today, it is possible to inspect the quality of a wide range of food products with a computer vision system. Both software and hardware parts of the system should be customized to the specific needs of a food company, including inspection goals and the type of product to analyze. Computer vision systems inspect food quickly, objectively, reliably and non-destructively, and they have the potential to take on many monotonous tasks traditionally performed by human inspectors. Revolutionizing quality control, such technologies also show great results in product grading and counting.
According to market analyses, all other things being equal, customers prefer apples with a maximum diameter of between 75 and 80 mm. However, people would have a hard time trying to accurately evaluate a fruit’s size with the naked eye, while a computer vision system can measure a precise diameter of an apple in a blink of an eye, literally. Food production companies that have already embedded computer vision solutions into the production line confirm this efficiency. For example, Ray Keller, General Manager at Apple King LLC, says that automated visual inspection provides “great grading results and 30% increase in apple production”.
Automated counting and sorting system based on image analysis can grade fruits, vegetables, nuts, oysters, etc. according to their shape, size and maturity (for fruits and vegetables), increasing the sorting speed by 10 times compared to humans. Automated visual check of a filling level and package labeling is another important application of computer vision in the food industry. Besides that, a visual system can check the freshness of a packed product with the aid of a special ink changing its color with time and at a different speed depending on the temperature.
Currently, in most depots and storage centers, ripeness of agricultural produce is still checked by hand but assessing the ripeness levels accurately is difficult even for professional inspectors. As tests conducted at AMAZON showed, assessors arrived at different results in 20% of the tests despite checking the quality of the exact same piece of fruit. This is why our startup team of machine learning experts joined forces with food experts in Russia to improve automation in the quality control of fresh produce.
The automated ripeness detection system consists of a conveyor belt that transports the food in containers to a particular sensor. The sensor looks like a normal camera, but it can capture information which is invisible to the human eye. We teach the machine what good and bad products look like by inputting new product variants on a daily basis. The products are photographed and made available to the machine in the shape of data. That way, the computer gradually understands the quality standards and becomes able to divide the fresh produce into five categories: "Premium", “OK”, “Damaged”, “Badly damaged” and “Expired”.
If defective products are detected early in the production, they can be rejected before any value is added. For example, it would be better to find a rotten apple before it was packed together with other apples.
Besides that, an automated visual inspection system can precisely count and measure the ingredients to prevent their excessive use. For example, it can control the amount of chocolate topping on a cookie, minimizing the give-away rate.
Thus, employing machine vision for food sorting on a region-wide scale would not only allow for a highly efficient food management and distribution, but also increase productivity, reach development goals, and spare time and labor of human workers who can concentrate on some more sophisticated tasks instead.
If you have any questions