optolab, university of brescia ligm, university of padua validation of the measurement performance...
TRANSCRIPT
Optolab, University of Brescia LIGM, University of Padua
Validation of the Measurement Performance
of a 3-D Vision Sensor by means of a Coordinate Measuring Machine
Giovanna Sansoni1, Simone Carmignato2, Enrico Savio2
1Laboratory of Optoelectronics,DEA - Dept. of Electronics for the Automation
University of Brescia, Italy2Laboratory of Industrial and Geometrical Metrology,
DIMEG – Dept. of Mechanical Engineering and Management
University of Padova, Italy
Optolab, University of Brescia LIGM, University of Padua
Optical-Contact CAD
The optical sensor: OPL-3D
View alignment
Single point clouds, Multiple point clouds, 3D desctiptive models
Volumetric fusion
CAM RP Re-styling
Development of a novel methodology for the reverse engineering of complex, freeform surfaces, combining three-dimensional vision systems and Coordinate Measuring Machines
BS
Context of the work
MI
PD
Optolab, University of Brescia LIGM, University of Padua
‘rough’ CAD model
The Optical-Contact CAD
Contact digitization
‘accurate’ CAD model
Data elaboration
Point clouds
Descriptive models
CAM RP Re-styling
Optolab, University of Brescia LIGM, University of Padua
The optical sensor: OPL-3D• Optical active triangulation
• Time-space coding of the meas. volumeFringe projector
Video camera
Assembly system
Optolab, University of Brescia LIGM, University of Padua
Point clouds
Optolab, University of Brescia LIGM, University of Padua
Multi-view integration of point clouds
Optolab, University of Brescia LIGM, University of Padua
Embedded ‘Check’ procedure (1)
• Performed after the calibration
• Based on the use of control markers
Calibration master
Optolab, University of Brescia LIGM, University of Padua
Embedded ‘Check’ procedure (2)
• Master dependent– Calibration masters for computer-vision
applications often priviledge low-cost aspect rather than the accuracy
• Designed for single view acquisition– Unable to grasp any information about the
influence of the alignment for multi-view acquisition
What about the accuracy of
OPL-3D?
Optolab, University of Brescia LIGM, University of Padua
…Monitoring the accuracy
MeasuremeMeasurement nt
UncertaintyUncertainty
Hardware:
principle of measurement,
mechanical stability..
Environment:
illumination,
temperature
Vibrations…
Extrinsic factors:
Operator skill,
Surface cleanliness
Measurement time
Clamping system
Measuring strategy:
measuring field,
calibration procedure
coordinate system
Object:
Surface finishing,
Colour and texture,
accessibility
Data processing:
Filtering,
Registration of multiple views
Algorithm sophistication
No standard available!
Optolab, University of Brescia LIGM, University of Padua
Evaluation of the accuracy of OPL-3D: substitution approach
Accuracy evaluation of OPL-3D
Point cloudsAccurate CAD Model
CAD-Point cloud Deviations
CMM Uncertainty
Calibrated Object
CMM used as the link of the traceability chain
Full optical digitization by means of OPL-3D
Optolab, University of Brescia LIGM, University of Padua
The calibrated object•Stainless steel, long term stability
•Volume: 230mm x 100mm x 20 mm
•Free form surface
•Representative of a typical measurement application
•Spheres for accurate alignment
•White painted•Substitution approach valid only for measurement tasks similar to that one represented by the calibrated object•Reasonable estimate of the accuracy for a wider range of applications
Optolab, University of Brescia LIGM, University of Padua
The accurate CAD Model•Touch probe CMM• MPE = 2.2 + L/300 µm (L: mm)
CA
D m
odel
from
the
CM
M
Accuracy of the CAD model
Optolab, University of Brescia LIGM, University of Padua
The optical acquisition
From the Check 100m
From the Check 70m
Single view acquisition: 240mm x 180mm
Multiview acquisition (3 v)120mm x 89mm
Optolab, University of Brescia LIGM, University of Padua
CAD-Point cloud Deviations
• Performed by Imageware Surfacer (SDRC Corporation)
• Evaluation of local deviations between the accurate CAD model from the CMM and the point clouds from OPL-3D
• Criticality of the alignment of the point clouds to the CAD model– Best fit using all measured points – Best fit using the sphere points– Best fit using the points on the freeform surfaces
Optolab, University of Brescia LIGM, University of Padua
CAD-Point cloud Deviations:single view
Optolab, University of Brescia LIGM, University of Padua
CAD-Point cloud Deviations:multiple views
Optolab, University of Brescia LIGM, University of Padua
Bottlenecks
Scanning Number
Planarity (mm)
1 0,2485
2 0,2470
3 0,2460
average 0,2472
•Calibration master •Alignment errors
•Need of skilled alignment strategies
•Need of view- fusion
•Need of accurate polygonal models