SICK has expanded its Nova machine vision platform with artificial intelligence tools capable of analysing three-dimensional height data, targeting more complex quality-control and packaging inspection tasks across food manufacturing and other industrial sectors.

The new AI for 3D capability combines deep-learning algorithms with spatial measurements generated by SICK’s 3D vision technology. According to the company, the system can identify variations in an object’s height, shape and surface even where defects cannot be distinguished through colour contrast alone.

For potato processors, potential applications could include inspecting filled cartons for deformation, checking that packaged products or cases are complete and identifying dimensional or surface irregularities before products move further along the packaging line. The technology could also be applicable to completeness checks involving trays, totes and secondary packaging, although SICK has not announced a potato-specific installation.

The system uses sample images to train inspection models on a manufacturer’s own products. Data collection, model training and inspection can be performed directly on supported devices, which SICK said reduces the need for additional equipment and allows inspection tasks to be adapted when products or batches change.

Each pixel recorded by the system carries a height value, allowing the software to reconstruct an object’s three-dimensional form. A neural network embedded within the Nova software then evaluates the data and displays detected irregularities through an anomaly heatmap.

SICK said the approach could help manufacturers identify problems earlier in production and packaging operations, including deformed cartons, missing objects and incomplete packs. The Nova platform also combines the new AI functionality with conventional rule-based inspection tools.

“Introducing neural network AI capabilities into our high performance Nova platform expands 3D inspection beyond its traditional limits,” said Diego Quintana Tukasaki, market product manager at SICK.

“Seeing far beyond the human eye, it provides greater control over data and analytics to deliver highly accurate, configurable, easy-to-train, customised intelligent inspection solutions.”

The new functionality can be activated through licences on predefined SICK devices or added through an upgrade licence, the company said.