Deploy AI-Powered Anomaly Detection into Your Production Line
Hydra3D+ is a 832 × 600 pixel resolution Time-of-Flight (ToF) CMOS image sensor tailored for versatile 3D detection and measurement. Hydra3D+ includes a brand-new 10 µm, three-tap pixel, which provides very fast transfer times down to 10ns and displays high sensitivity in the NIR wavelength. Its powerful on-chip HDR combined with an on-the-fly flexible configuration, enables the best trade-off between application-level parameters. The sensor embeds an innovative on-chip multi-system management feature enabling multiple ToF systems to work simultaneously without interference. With Hydra3D+, customers can easily achieve fast 3D measurements without motion artefacts in all distance ranges and in all light conditions.
Operates in real-time without motion artefacts
The sensor includes a brand-new 10 μm, three-tap pixel, which provides very fast transfer times, excellent modulation contrast and high sensitivity in the NIR wavelength. This enables the sensor to operate in real-time without motion artefacts for customers seeking reliable 3D detection and the highest levels of 3D performance, including high depth resolution, high speed and flexible operation conditions, all without system interference.
At all distances and in any condition
Deploy AI-Powered Anomaly Detection into Your Production Line.
Anomaly Detector integrates with all Gocator Line Profiler, Snapshot, and Line Confocal Sensors when accelerated with GoMax. The firmware is available for download from the LMI Website free-of-charge. Purchase of a licensed LMI Dongle is required to run Anomaly Detector with a LIVE sensor. License-free evaluation is available with firmware running on GoMax in REPLAY mode. Anomaly Detector is supported on GoMax ORIN and ORIN+, our latest smart 3D vision accelerator platform.
We invite you to test out Anomaly Detector sample scenarios from our new Scenario Downloads platform.
Common Use Cases
Part Inspection
Find defects and irregularities on machined or casted parts with complex geometry. Train an AI model to identify good and bad parts directly on the production line without a cloud connection, a CAD file, or complex thresholds typically used with traditional tools.
Surface Inspection
Find defects and irregularities on machined or casted parts with complex geometry. Train an AI model to identify good and bad parts directly on the production line without a cloud connection, a CAD file, or complex thresholds typically used with traditional tools.
Key Features
Train on 2D intensity or 3D height map data
Scan parts with intensity and surface data and select the best option at time of training. Use a 3D visualizer to view complex parts and improve labeling accuracy.
Preview functionality without a license
Evaluate anomaly detection with a GoMax in replay mode and train and perform inference without a license.
Generate synthetic and augmented data
Reduce the number of images required for training with integrated generation of synthetic defects and augmented frames that improve the performance of your model.
Store new production data directly on GoMax
Use GoMax to store production data directly to a project archive that can be used to quickly train a new model and improve performance of your application.
Integrated with GoPxL
Benefits
Detect features of varying shape and size
While traditional tools are excellent at blob and segmentation, they require tuning of thresholds, often specific to each part. With Anomaly Detector, the user is not required to manage detection thresholds. Training relies on providing a dataset of OK and NG parts and creating a detection model specific to the parts in the dataset.
Integrated AI modelling workflow
Model training and related datasets are managed directly inside of GoPxL tools, so you spend less time moving files and datasets compared to using a separate application for training models.
Train on the production line
Train data directly on GoMax and avoid the time, cost, and data security concerns related to moving data to the cloud or local PC. Training and inference uses the same license, allowing for models to be updated in production without an additional, development license.
Use Prediction to label new production data
After initial training, rely on assisted labeling to make iteration fast and easy.
Measure anomalies and define acceptable thresholds
Surface anomalies are passed to subsequent tools for measurement and gauging. Users can pass anomalies of a certain size or shape, depending on what is considered acceptable to the end user.
Measure anomalies and define acceptable thresholds
Use the Python-based script tool to add custom logic or pull measurement thresholds from a local file. Power users can leverage the Python GDK to train models using open source and proprietary tools for subsequent deployment.