Frozen Pizza Sorting and Inspection
Automatically confirm pizza type, proper ingredients, and absence of physical contaminants
Graphical programming environment for deep learning-based industrial image analysis
Frozen pizza is an extremely competitive consumer food market. The ability to consistently produce appealing, symmetrical pizzas with minimal waste is key to survival. Pizzas are manufactured in large volumes by dropping sauce, cheese, and a variety of toppings as precooked dough disks move past on a conveyor. Pizzas are then fast-frozen in a tunnel freezer or spiral freezer.
The production rate is high, so placement and volume are not precise. Before packaging, each pizza must be visually inspected to ensure it meets various criteria, including the type of pizza, the number of pieces of particular discrete ingredient such as pepperoni, and the absence of any physical contaminants.
Like most food items, pizzas vary widely in appearance. Pepperoni slices can appear anywhere in the circle and the distribution of other ingredients can also vary significantly. Any number of undesired ingredients from other pizza varieties can appear, as can other physical contaminants such as bits of plastic. Conventional machine vision has trouble distinguishing between acceptable variations and unacceptable ones. It also struggles with the wide range of possible defects.
Cognex AI-enabled technology is an ideal solution for automating frozen food inspection applications. It has different tools that train on sets of images of different pizza varieties and pizza topping distributions and learn to make the distinctions necessary to pass only pizzas that meet requirements.
The part location tool trains on a small set of images of pizzas with the desired number of pieces of pepperoni or other toppings. It then identifies, locates, and counts pepperoni slices, even when they are touching or overlapping, while ignoring variations in the underlying cheese and sauce. If the number of topping pieces falls outside prescribed limits, the pizza is rejected.
The defect detection tool trains on a set of images of acceptable pizzas, even those with a large variety of toppings. It then detects any physical contaminants, including toppings not appropriate to this pizza type, while ignoring variations in the underlying cheese and sauce. Consumers do not, after all, expect or want to find a mushroom on the meat supreme pizza they just bought for dinner.
The classification tool trains on a labeled set of the full range of possible pizza varieties that are produced in the facility. It then distinguishes each variety and can sort them so that they are placed in their appropriate packaging and accurately entered into inventory. All of these tools used together help manufacturers automate processed food inspections ensure only the highest quality products leave the facility and end up on customer’s kitchen tables.