Content uploaded by Richard Leach
Author content
All content in this area was uploaded by Richard Leach on Jun 17, 2021
Content may be subject to copyright.
euspen’s 20th International Conference &
Exhibition, Geneva, CH, June 2020
www.euspen.eu
Development of an all-optical dimensional measuring system
Richard Leach1,2, Waiel Elmadih1, Samanta Piano1, Mohammed A. Isa1, Danny Sims-Waterhouse1,2, Nicholas
Southon1, Wahyudin P. Syam1
1Manufacturing Metrology Team, University of Nottingham, UK
2Taraz Metrology, Nottingham, UK
richard.leach@nottingham.ac.uk
Abstract
We present the novel design of an all-optical dimensional measuring system (AODMS) for measuring the geometry and surface
texture of small-scale components. The system is designed to operate in a cube of 100 mm sides, with micrometre or sub-micrometre
measurement uncertainties. The AODMS includes a four-axis motion system for mounting and moving the sample to be measured,
a photogrammetric system for coordinate measurement and motion system tracking, a combination of coherence scanning
interferometry and focus variation microscopy for texture measurement and a metrology frame fabricated using additive
manufactured lattice structures with internal resonating bandgaps for vibration isolation. The paper will discuss the development of
the AODMS, the experimental realisation of the instrument and the first steps in its validation.
Coordinate metrology, surface topography, information-rich metrology
1. Introduction
To address the need for three-dimensional (3D) measurement
of the geometry of complex milli- to micro-scale components,
there have been a number of developments of tactile micro-
coordinate measuring machines (CMMs) – see [1] for a recent
review. However, the commercial success of such CMMs has
been limited due to their delicate mechanics, complexity of use,
contact nature and slow measurement speeds. To address these
limitations, several optical surface topography measuring
instruments have been equipped with multi-axis motion systems
to allow them to act as CMMs (e.g. [2]), but for geometry
measurement they require multiple stitching operations and can
be slow. Optical probes have also been integrated with tactile
CMM platforms (e.g. [3]) or robot arms (e.g. [4]), but these are
usually for geometry measurement only and often have
limitations in terms of object accessibility.
In this paper, we present the novel design of an all-optical
dimensional measuring system (AODMS) for measuring the
geometry and surface texture of micro-scale components. The
system is designed to operate in a cube of 100 mm sides, with
micrometre or sub-micrometre measurement uncertainties.
This new system is designed to be fast and produce dense point
clouds; characteristics that are not shared with tactile
instruments [1]. It is important to state up front that the AODMS
is not designed to be a state-of-the-art coordinate measuring
system and/or surface texture measuring instrument – rather it
is a platform to demonstrate the concepts of “information-rich
metrology” (IRM) [5]. When manufacturing a product, we have
information about the product before we start manufacture. We
usually have computer models, information about the materials,
and we often know what to look out for in terms of defects. This
“a priori” information can be used to enhance the measurement
process by focusing on what exactly needs to be measured, so
decreasing the time to do it. Most of the above information
becomes available at product development and at
manufacturing process planning, and we are asserting that such
information may also bring benefit to metrology.
Several examples of IRM will be investigated using the
AODMS, but first, we need a highly stable platform with multi-
scale sensing capabilities. This paper concerns the design and
development of such a platform.
2. System design
The core design principle of the AODMS is to combine two
scales of optical measurement to create a system for
measurements of the geometry and surface texture of small-
scale components. Additively manufactured lattice structures
are incorporated into the design to aid in vibration isolation,
with the aim of improving the performance of the optical
measurements.
Figure 1 Computer render section view of the AODMS, with the central
cube being the measurement volume possible with both texture and
form sensors.
The optical texture sensor (see Section 2.2) is mounted on an
aluminium ring, which places the thermal centre close to the
optical axis of the surface texture measurement. The ring also
places all three photogrammetry cameras (see Section 2.1)
equidistant from the centre of the measurement volume. The
ring is kinematically coupled with the lattice legs (see Section
2.5), using micrometers and three ball and vee-groove kinematic
couplings [6]. A computer rendered cut-away depiction of the
system is shown in Figure 1 and Figure 2 is a photograph of the
assembled system.
Figure 2 Photograph of the AODMS.
The position of the measurement stage is tracked with the
photogrammetry system (see Section 2.4). This allows us to use
low accuracy, low cost (but high repeatability) motion stages,
and anchors the measurement stage co-ordinate system to the
optical co-ordinate system.
2.1. Form measurement
The AODMS is capable of performing geometry (form)
measurement using the multi-camera system. The form of a
sample is reconstructed with photogrammetry; a passive
triangulation-based technique in which the 3D location of
corresponding image features between two or more images can
be triangulated [7]. Typically, SIFT features are used as they are
considered the most effective feature detection method due to
their robustness to changes in perspective [8]. The multi-camera
setup, along with the rotation stage, allows the system to
capture many images of a sample to be measured from a wide
range of positions. Smooth objects can be measured by taking
advantage of illumination with a stationary laser speckle pattern
(not shown) [9].
The photogrammetry system is based on an open-source
photogrammetry pipeline called OpenMVG (Open multi-view
geometry). However, OpenMVG is based on self-calibration and,
therefore, only produces point clouds with an arbitrary scale
factor. In order to apply a scale factor to the form
measurements, a pre-calibration procedure is used in order to
determine the true metric location of the three cameras with
respect to each other. The pre-calibration procedure consists of
imaging of a calibrated checkerboard artefact as described by
Zhang [10]. The pre-calibration procedure provides both a
characterisation of the camera intrinsic and extrinsic parameters
as well the relative distances between the cameras. The
calibrated camera-to-camera distances allow the
photogrammetry point cloud to be appropriately scaled and
metric information about the sample form to be evaluated.
Further scaling and uncertainty estimation for the
photogrammetry system are described elsewhere [11]. Figure 3
shows a point cloud measured with the photogrammetry
system, where the measured ball diameter is 12.096 mm with a
standard deviation of 0.060 mm. Of course, traceability for these
measurements is still part of the work in progress, although we
have recently published on the how to establish uncertainty for
the photogrammetry measurements [11].
Figure 3 Initial photogrammetry measurement.
2.2. Texture measurement
The optical setup for the surface texture measurement is a
coherence scanning interferometer (CSI) sensor [15]. The optical
setup consists of a Kohler axial illumination system and a
microscope equipped with a 10× Mirau objective. Figure 4 shows
the optical design and setup. The optical sensor is mounted on a
precision linear stage with an axial resolution of 20 nm.
From the CSI sensor, two types of raw data can be obtained:
CSI data as the primary raw data type and focus variation (FV)
data [13] as the secondary raw data type. The purpose of using
two types of data is to combine the benefit of CSI for measuring
smooth and highly reflective surfaces and the benefit of FV for
measuring rough surfaces with high-slope angles. These two
types of raw data can be extracted from a single measurement
with the optical system.
2.3. Data fusion
As well as using checkerboard targets to provide high-accuracy
localisation of the motion stage, the SIFT features described in
Section 2.1 can be used to track the location and orientation of
the sample. The feature information captured during the
photogrammetry measurement process can be used to locate
the photogrammetry point cloud within the surface texture
measurement coordinate system. By locating the
photogrammetry point cloud in the same coordinate system as
the surface texture measurement, the two data sets can be
combined and fused in order to generate a single data set
covering both sample form and surface texture information.
In order to ensure that the coordinate systems of both the
form and surface texture measurement systems are accurately
aligned, some common features visible in both measurement
systems are required. The optical targets shown in Figure 5
provide an good solution to this, as the intersections between
the squares of the checkerboard can be seen in the surface
texture measurements and can be triangulated to sub-pixel
accuracy by the photogrammetry system. By measuring at least
three checkerboard intersections with both systems, the relative
rotation and translation between the two coordinate systems
can be determined.
The outcome of this fusion of form and texture information
allows for the generation of sample measurements with
improved dynamic range of spatial frequencies. Additionally, the
methodology allows for the sample to be removed and replaced
in any orientation and still allow the two measurements to be
registered. This ability to manually move the sample to any
orientation and still be able to register texture measurements
negates the need for any highly complex and high-cost precision
lateral motion stages (see Section 2.4).
Figure 4 The optical design and setup of the CSI sensor.
2.4. Motion system
The AODMS has positioning capacity in the lateral
𝑥
and
𝑦
directions up to 100 mm. In addition, motion of the texture
sensor measurement head in the
𝑧
-direction allows positioning
of the measurement range in a desired region. The fourth
degree of motion is for the multi-view imaging at various
rotational positions
𝜃
as illustrated in Figure 5. Machine vision
cameras are used for the purpose of photogrammetric 3D
measurement, positional tracking for registration and geometric
characterisation of the rotation axis.
There are three machine vision cameras (see Section 2.1) in
the setup of the AODMS for multi-view monitoring of the
position of a measured sample. The use of multiple cameras
improves the tracking accuracy and the visibility of optical
targets that can be occluded by the measured sample. Four
checkerboard-patterned optical targets are placed at the
corners of the measurement plate shown in Figure 5. The
tracked positions are used for stitching and fusing surface
texture measurements at different
𝑥𝑦
positions.
The position and orientation of the rotation axis, about which
images are acquired for photogrammetry, is evaluated through
a geometric characterisation process. This is necessary in order
to accurately relate the image-based photogrammetric
coordinate measurements to the actual dimensions and position
of the measured sample. The details of the stage
characterisation process using the camera data are given
elsewhere [14].
Figure 5 Measurement stage showing optical targets and motion
directions. Note that the workpiece metrology frame shown in Figure 2
had not been assembled at the time of this photograph.
2.5. Vibration isolation
Design of the support frame and workpiece metrology frame
of the AODMS used additive manufactured (AM) lattice
structures to mimic the vibration isolation behaviour of elastic
bandgap structures [15]. Bandgap structures provide enhanced
vibration isolation, in comparison to solid structures, with
vibration transmissibility of up to -66 dB as reported elsewhere
[16]. We targeted the low-frequency vibration range from 50 Hz
to 150 Hz, which corresponds to the vibration of a typical
laboratory environment (below 50 Hz we rely on the use of a
massive concrete structure on which the instrument is
mounted). However, it is challenging to use existing bandgap
structures to isolate this low-frequency range. This is because
the resulting unit cells size would be larger than the envelope of
the AODMS and the geometrical features of the lattices would
be impossible to manufacture with current AM machines. The
support and workpiece metrology frames of the AODMS share a
novel lattice structure design for low-frequency vibration
isolation. This design incorporates an energy absorption
mechanism in the form of a cubic solid structure with stiffeners,
as shown in Figure 6.
Figure 6 Design of lattice unit cells for low-frequency vibration isolation.
The support frame and workpiece metrology frame are shown
in Figure 7. For an illustration of the ability of the structures to
isolate vibration, a vibration test was set up, integrating a
mechanical shaker, a piezoelectric accelerometer and a laser
vibrometer. The transmissibility of vibration through the
support frame was measured, referencing acceleration data
from the accelerometer and vibrometer, as shown in Figure 8.
Figure 7 Support frame and workpiece metrology frame as
manufactured from Nylon 12.
Below the targeted isolation region, the response is slightly
below 0 dB. The first resonance of the structure is designed at
the start of the isolation frequency range, at 50 Hz. Above 50 Hz,
the energy absorption mechanism is in action and results in wide
bandgap behaviour from 50 Hz to 1000 Hz. The transmissibility
reaches as low of -62 dB. Note that whilst this is the lowest
bandgap frequency reported in the literature to date [16], the
use of bandgaps usually has the detrimental effect of amplifying
low-frequency vibration (as can be seen in Figure 8). For now,
we rely on the high-mass passive isolation of the large concrete
support table (and isolated laboratory floor) to reduce this effect
and we continue research to push the lower end of the bandgap
frequency to 0 Hz.
Figure 8 Experimental testing results for the support frame.
To avoid the thermal expansion of the Nylon-12 having a
significant effect on the metrology loop, the lattice structures
are sandwiched between low-expansion Invar sheets which are
rigidly connected to each other with Invar strips (shown in Figure
2 and Figure 5).
3. Current status of AODMS
At the time of writing, the full system had only just been
assembled and initial tests are still underway. Some early data
from the registration and fusion process are described
elsewhere [17] and a paper on the enhanced stage tracking is in
press [18]. Further results will be presented at the 20th euspen
Annual Conference and Exhibition.
Acknowledgements
This work was supported by the Engineering and Physical
Sciences Research Council [grant number EP/M008983/1].
Thanks also to Dr Rong Su, Dr Ian Maskery and Ahmed
Mohammed (University of Nottingham), Dr Wu Jianwei and Dr
Bo Zhao (Harbin Institute of Technology), Fabrizio Medeossi
(University of Padua), and Phil Krause (now Cranfield University).
References
[1] Thalmann R, Meli F, Küng A 2016 State of the art of tactile micro
coordinate metrology Appl. Sci. 6 150
[2] Zangl K, Danzl R, Helmli F, Prantl M 2018.Highly accurate optical
µCMM for measurement of micro holes Proc. CIRP 75 397-402
[3] Xiong H, Pan M, Zhang X 2010 The development of optical fringe
measurement system integrated with a CMM for products inspection
Proc. SPIE 7855 78551W
[4] Du H, Chen X, Xi J, Yu C, Zhao B 2017 Development and verification of
a novel robot-integrated fringe projection 3D scanning system for large-
scale metrology Sensors 17 2886
[5] Senin N, Leach R K 2018 Information-rich surface metrology Proc.
CIRP 75 19-26
[6] Leach R K, Smith S T 2018 Basics of Precision Engineering (CRC Press)
[7] Luhmann T, Robson S 2011 Close Range Photogrammetry: Principles,
Techniques and Applications (Whittles Publishing)
[8] Lowe D G 2004 Distinctive image features from scale-invariant
keypoints Int. J. Comp. Vis. 60 91–110
[9] Sims-Waterhouse D, Piano S, Leach R K 2017 Verification of micro-
scale photogrammetry for smooth three-dimensional object
measurement Meas. Sci. Technol. 28 055010
[10] Zhang Z 1999 Flexible camera calibration by viewing a plane from
unknown orientations Proc. 7th IEEE Int. Conf. Comp. Vis. 1 666–673
[11] Sims-Waterhouse D, Isa M A, Piano S, Leach R K 2020 Uncertainty
model for a traceable stereo-photogrammetry system Prec. Eng. in press
[12] de Groot P 2011 Coherence scanning interferometry. In: Leach R K
Optical Measurement of Surface Topography (Springer: Berlin)
[13] Helmli F 2011 Focus variation instruments. In: Leach R K Optical
Measurement of Surface Topography (Springer: Berlin)
[14] Isa M A, Sims-Waterhouse D, Piano S, Leach R K 2019 Kinematic
error analysis of stage tracking using stereo vision Proc. ASPE, Pittsburgh,
USA, Oct. 151-156
[15] Syam W P, Jianwei W, Zhao B, Maskery I, Elmadih W, Leach R K 2018
Design and analysis of strut-based lattice structures for vibration
isolation Prec. Eng. 52 494-505
[16] Elmadih W, Chronopoulos D, Syam W P, Maskery I, Meng,H, Leach
R K 2019 Three-dimensional resonating metamaterials for low-
frequency vibration attenuation Sci. Rep. 9 11503
[17] Leach R K, Sims-Waterhouse D, Medossi F, Savio E, Carmignato S, Su
R 2018 Fusion of photogrammetry and coherence scanning
interferometry data for all-optical coordinate measurement Ann. CIRP
67 599-602
[18] Isa M A, Sims-Waterhouse D, Piano S, Leach R K 2020 Volumetric
error modelling of a stereo vision system for error correction in
photogrammetric three-dimensional coordinate metrology Prec. Eng. in
press
