SICK has extended AI-powered inspection to its Nova 3D machine vision hardware for the first time, according to the company's Nova 2.15 release. The update makes the Nova Intelligent Inspection toolset license available on 3D devices, letting engineers combine AI Anomaly Detection and AI Object Detection with rule-based tools for height-data verification. The release processes 2D data only within the 3D toolset for now.
The AI for 3D toolset targets inspection tasks that rule-based machine vision has struggled with, including package deformation checks in food and beverage packaging, empty box or missing object detection in logistics totes, assembly verification and surface inspection in electronics and battery production, and tyre classification with 3D OCR on automotive lines. Every pixel captured by SICK's 3D sensors carries a height value, and a neural network embedded in the Nova software reconstructs that data to flag defects on an anomaly heatmap. Engineers running colorless or low-contrast lines will care most about that last part: deformation and surface defects invisible to color cameras become detectable through height data alone. Models train on-device from example images, with no added hardware required.
Nova 2.15 also introduces AI Blob Finder in dStudio, an instance-segmentation tool that detects, counts, and measures objects within a defined region using a trained neural network, returning pixel-level masks even when objects overlap significantly. The release adds class filtering and configurable width and height thresholds per class. The existing AI OCR Pretrained tool gains date verification, a threshold for the number of characters read, and support for additional characters, while the Ruler3000 line gets its own AI OCR Pretrained tool for 2D dot-print reading.
Three other new tools round out the release: a Network tool for viewing and changing Ethernet interface settings on the device, a Surface Fitter tool for inspecting curved 3D surfaces, and a Position Verifier tool that stores object positions found by a Finder or Object Detection tool as a reference for later comparison. The Dashboard interface now allows threshold values to be edited directly in Run-mode, with permission controls and a lock restricting interaction to pre-defined widgets.
The AI for 3D toolset is activated by license, either on pre-defined devices or through an upgrade license, and runs alongside SICK's existing rule-based tools rather than replacing them. Diego Quintana Tukasaki, market product manager at SICK, said the release targets "two of the highest growth product categories in machine vision: AI software and 3D imaging." How quickly SICK follows this 2D-only rollout with full 3D-native AI processing remains an open question for engineers evaluating the toolset today.
