OptoInspect3D scanning technology for optical quality inspection

Speed, robustness, automation capability and a suitable measurement principle are the key properties of in-line metrology. The OptoInspect3D technology package developed at Fraunhofer IFF provides you a modular toolkit for implementing 3D scanning systems for specific applications.

OptoInspect3D includes methods and tools to design, configure and simulate optical scanning systems. Functions for rapid, timed and automatic 3D measurement data interpretation and geometrization are another key component. OptoInspect3D additionally includes tools to calibrate and adjust systems configured for specific applications from several sensors and motion components.

Our OptoInspect3D scanning technology can be used in any industry and is suitable for assorted process control and quality inspection tasks, such as inspection of geometric tolerances for dimension, shape and position and assembly and completeness verification. OptoInspect3D has been in use for over ten years and is continuously being refined.

Functions

Registration and alignment

Measurement data usually have to be registered before they can be compared with a CAD model or a reference measurement (golden sample). The OptoInspect3D-Inline library includes all the functions needed for this. Registration can be calculated between two point clouds or between a point cloud and a CAD model in the form of a triangle mesh. Several point clouds can also be simultaneously aligned and merged into one object. Then, a subsequent distance calculation between scan and model data shows whether the data sets fit together.

Filtering and preprocessing

Scanned point clouds frequently have to be preprocessed or filtered to allow sophisticated analysis. The functions provided for this enable removing outliers from the scan data and spatially segmenting smoothing and homogenizing point clouds. Features, such as curvatures or point densities, can be identified and normal vectors of the measurement data can be calculated for realistic visualization.

Best-fit geometric primitives

A frequent task in sophisticated 3D data processing is the fitting of geometric primitives in measurement data and the subsequent interpretation of differences. The OptoInspect3D Inline library supports a variety of such primitives, including lines, planes, circles, ellipses, spheres, cylinders, cones and tori. They can be approximated using various minimization criteria. Along with the least square displacement method (Gauss), approximations of the minimum zone (Chebyshev), minimum bounding (bounding geometry) and maximum inscribing (Pferch) are available.

OptoInspect3D Inline library features

• Numerous algorithms for the analysis of 3D measurement data and CAD comparison
• Support of various operating systems (Windows, Linux) and architectures (x86_64, ARM)
• Can be run on SoC systems, such as Raspberry Pi and NVIDIA Jetson
• Universal Type-C-compliant interface without external coupling
• Top performance through efficient data structures and multicore support
• Processing of large numbers of points
• PTB-certified accuracy
• Development and test environment with OpenGL-based visualization

Sensor simulation for synthetic reference data generation

OptoInspect3D Sim employs a model-based approach to enable the planning and implementation of assorted inspection tasks with optical sensors. Whereas conventional optical inspection systems normally draw on predefined reference data and corresponding inspection algorithms for variance analysis, the model-based approach is significantly more flexible whenever there is a need to respond to frequent changes of inspection requirements. Consistent use of the item under test’s CAD data and a model description of the inspection system enables precisely predicting the readings expected in the to-be state in the form of synthetic measurement data. These can be image or 3D data that adjust automatically to any change to the design of the item under test.

Comparison of the measurement data of the item under test with the data generated by the simulation using suitable metrics eliminates renders manual adjustment of inspection algorithms unnecessary. The simulation is optimized so that it already delivers suitable data.

Edge filters can be used to compare the informational content of synthetic and real camera images, for instance. The items under test are segmented, extracted and compared in both images. Our model-based approach permits precise segmentation of the image region analyzed, direct simulation of the expected edges and thus a robust decision.

Sensor simulation for view planning

An important task for flexible, for instance, robotic inspection systems is finding the ideal sensor position to precisely scan the desired quality features. Not only the visibility of the features to the optical sensor used but also its alignment to the product’s surface is crucial. This is the only way to meet scan conditions and prevent interference by adverse reflective properties.

OptoInspect3D Sim provides flexible and automatable view planning. Whenever the inspection task changes or new components are added, the corresponding path can be adjusted or created automatically.

Widely varying inspection system configurations can be analyzed very quickly by varying the sensor’s position to the product as a function of the available degrees of freedom, the maximum sensing range and, if necessary, other conditions, such as collision avoidance. The selection of a reliable configuration enables automated quality assessment of the generated synthetic measurement data based on predefined criteria. Manual teaching of sensor positions is no longer necessary.

Extensive practical tests usually cannot be performed for every product variant, either when configuring inspection systems or when operating them later. Our model-based approach in OptoInspect3D Sim delivers synthetic measurement data of the product regardless of its physical design. This enables drawing conclusions about the inspection system’s suitability for the intended use case. The inspection system can thus be designed and evaluated based on the simulation data before the item under test has even been manufactured.

OptoInspect 3D Sim library features

• Software library for the simulation of synthetic measurement data to calculate appropriate sensor positions based on CAD data and sensor parameters (view planning and viewpoint selection)
• Geometrically correct generation of synthetic measurement data for optical 2D, 2.5D and 3D sensors (e.g., camera, PMD, light sectioning and structured-light 3D scanning) including shading
• Simulation and data interpretation with various metrics and strategies
• Input: specification of sensor functions in the form of geometric and optical parameters and of the target object in the form of a CAD model
• Output: synthetic measurement data and location-based assessment of the data quality based on surface features and reflectance
• Top performance through full calculation in the graphics processing unit (GPU) and the use of advanced rendering techniques
• Available as a library and as an environment with a graphical user interface

Our software library OptoInspect3D Calib provides system integrators and sensor and equipment manufacturers a method and software functions for automatically determining sensor positions and motion axes (sensor positioning and scanning) for systems configured for a specific application.

The principle of the positioning procedure is based on a model description of the scanning system and the specification of the model’s parameters by taking measurements on a calibration block.

OptoInspect3D Calib library features

• Flexible procedure for calibrating a scanning system’s optical sensors and motion axes
• Transformation of the individual sensors’ data into a shared coordinate system for absolute measurements
• Traceability of the scan to a national length standard through reference measurement on an appropriate calibration block
• Calibration supersedes fine adjustment of components (e.g., when replacing sensors)
• Reproducible insertion of components is no longer necessary. The calibration procedure can be fully automated.
• Suitable for laser range sensors, laser light section and structured-light 3D scanning sensors and camera sensors