A new configurable AI-powered 3D machine vision from SICK combines deep learning with precise height data analysis by means of its Nova foundation software. The industrial automation and sensor intelligence specialist says it has equipped its Nova machine vision platform with AI deep learning algorithms to enable ease-of-use and quick customisation to develop a range of new inspection capabilities.

The company explains that delivering configurable AI-powered 3D machine vision with SICK Nova across industries such as logistics, electronics, battery manufacturing, automotive, and consumer goods, improves quality control with intelligent inspection. The example-based approach, together with on-device training and a user-friendly interface, is designed to simplify solution development and provide a cost-efficient way of incorporating deep learning into quality control operations, with no extra equipment required.

SICK asserts that its AI-powered machine vision with 3D height analysis unlocks advanced quality-control applications that were previously difficult to achieve with traditional rules-based inspection. Key applications for the new solution include: package deformation inspection in sectors such as food and beverage and consumer goods, empty box or missing object detection in totes or matrix packaging for fulfilment and logistics operations, assembly verification and surface inspection for electronics and battery production, classification and tyre inspection, including 3D optical character recognition, for automotive applications, and completeness checks in manufacturing.

For example, AI-powered deformation inspection validates the appearance, integrity and function of carton packaging after filling, measuring height deviations of the shape and surface of packages to detect defects at high speed, even if there is no colour contrast. AI-powered 3D machine vision is engineered to see beyond the human eye to ensure faulty products are removed from the production and packaging process at an early stage.

A ‘teach-by-example’ method, using sample images, helps to enable easy setup, and the company’s AI for 3D solutions are customised and trained on an organisation’s own data. Users collect data, train models and execute inspection tasks directly on-device, with quick set-up of inspection tasks that enables fast, effortless batch changes.

Every pixel carries a height value and is captured by high precision technology, enabling reconstruction of 3D data and defect detection by a trained neural network embedded in the Nova AI software. The new solution is designed to provide colour and contrast independent inspection and improved quality control across industries, with detected defects displayed on an anomaly heatmap. Activated with licence on pre-defined devices or through an upgrade licence, the toolset also includes traditional rules-based machine vision software tools.

SICK says it works closely with customers to develop tailored solutions that address their key inspection challenges and detect specific anomalies, from minute electrical components to large packaged goods.