The Use of Self-identifying Targeting for Feature Based Measurement

Giuseppe Ganci
Geodetic Services, Inc.
1511 Riverview Drive
Melbourne, Florida USA
E-mail: [email protected]

Robert Clement
BMW Manufacturing Corporation
1400 Hwy. 101 South
Greer, South Carolina USA
E-mail: [email protected]

Coordinate Measuring System Committee July 2000, Dearborn, Michigan.

KEY WORDS:
Industrial Measurement, Coordinate Measurement, Photogrammetry,
Videogrammetry, Coded Targets,
ABSTRACT
A powerful, new concept is described for measuring geometric features. The concept was
originally developed for measuring the planes, edges and corners of the NC blocks that are
commonly used in automotive fixtures. These features in particular are very difficult to measure
accurately and efficiently. The concept greatly simplifies both the measurement and analysis of
these features. The concept has also proven very useful for numerous other measurement
applications.
The concept uses a new type of targeted adaptor called a Feature Target, herein abbreviated
as FT. Several different FT configurations are available depending on the type of feature
measured. The appropriate FTs are applied to the features that require measurement. Digital
photogrammetry is then used to quickly and easily measure them.
The paper describes the application of these targets in two case studies. The first study examines
their use in the measurement of an automotive fixture. The second study describes the use of the
targets for an in-line measurement application in an automotive production line.
INTRODUCTION
Photogrammetry is a fast, accurate 3-dimensional measurement method based on photography.
The new concept described in this paper uses what is called the single-camera or sequential mode
of photogrammetry. In this mode, a single high-resolution digital camera takes multiple pictures of
an object from different locations. These pictures are then automatically processed to yield the 3dimensional coordinates of points on an object. (Ganci
and Handley, 1998) The sequential mode is not real-time,
so it can only measure static objects and targeted points.
However, it is well suited to high accuracy measurement
of large, complicated objects since practically any
number of pictures can be taken and processed. In
addition, the photography is usually quick, so production
downtime is low and temperature effects are minimized.
The use of the sequential method of photogrammetry on
a typical measurement is illustrated in FIGURE 1.
Sequential photogrammetry has other attractive
attributes including high accuracy, portability, and the
FIGURE 1 - Sequential Photogrammetry
ability to measure in unstable or vibrating environments.
(Brown, 1998) All these attributes combined with the new concept described in this paper make
sequential photogrammetry a compelling choice for numerous applications in the automotive
industry.

Traditionally, three-dimensional inspection within the automotive environment takes place in a

Coordinate Measurement Machine (CMM) room. The CMM room will typically have several
stationary gantry-type CMMs dedicated to a certain class of dimensional inspection. This class of
inspection is most notably characterized by the need to bring the part to be measured to the
CMM. Such a constraint instantly disqualifies traditional CMMs from inspecting the tooling fixtures
that are located on the assembly line. These fixtures are in fact the most important measured items
in the factory, as they control the dimensions and fit-up of the sub-assemblies throughout the
manufacturing process. In-line production measurements are also impossible using CMM systems.
Clearly, a portable measuring system that can measure items in place is desirable for many
applications. Although numerous portable, in place measurement systems exist, they typically rely
on touching the desired features with some type of probe or measuring adapter that is usually held
in place by an operator. For many applications, access to the object is limited, making setup of
the instrument difficult. Furthermore, touching the desired features with the measuring device can
be difficult, tedious and prone to error.
For many automotive inspection tasks it is necessary to quickly collect and process feature data.
One limiting factor until now in the use of photogrammetry for feature measurement is the need for
point-of-interest targeting. For example, if a plane needs to be measured, then at least three
points (>three for redundancy) need to be applied to the surface that defines that plane. Similar
requirements apply to other geometric features such as circles and lines. In some instances, pointof-interest targeting cannot be used to define the feature. An edge is one example of a feature
that cannot be directly targeted. Typically, a special target adapter is
needed to define the edge, or the edge is determined indirectly via
the intersection of two planes. This is illustrated in FIGURE 2. The
example of an edge can be extended to a corner via the addition of
another plane and a line-line intersection to produce the corner.
Traditionally, target adapters have been an excellent way to bypass
some of the problems associated with measuring difficult features such
as edges and corners. The time penalty for these adapters comes at
the processing stage when they need to be identified, and their
measurement reduced, to yield the desired feature.
FIGURE 2 Plane plane
intersection to derive an
edge

An alternative to stick-on targeting when using photogrammetry is the

use of hand-held probes to touch points of interest. (Ganci and Brown,
2000) Some typical probes are shown in FIGURE 3. These touch-probes have a standard tip, similar
to the type used on conventional CMMs. In addition, each probe has five permanent targets. Two
or more cameras simultaneously view these targets and calculate their XYZ position. From the XYZ
position of the five targets, the probe-tip can be calculated. Probes have been successfully used
in automotive applications for many years. However, in some applications setting up the cameras
so they can measure the object is difficult. Furthermore, many features are hard to accurately
probe especially in the tight spaces and difficult conditions encountered on many measurements.
In so far, two approaches have been discussed, point-ofinterest targeting and touching by a probe. The targeting
approach uses sequential photogrammetry that can often
acquire the data very rapidly since photography is very
quick. However, sometimes the need to target the desired
features is difficult and time consuming. If, instead, a
probing tool is used, there is typically setup time for the
instrument, followed by the laborious probing and analysis
of each feature. To circumvent both disadvantages, the
focus of recent developments has been to reduce set-up
time and to also remove the requirement for probing
features. Substantial savings have been obtained via the
FIGURE 3 Assorted hand-held probes
use of targeting adapters known as Feature Targets (FTs).
This paper describes the principles involved and some of the characteristics of FTs. The
performance of the new targeting will also be examined through reference to two case studies.

CONCEPT
The principle of FTs is now explained. They work much like an ordinary target adapter but are
identified via the use of a "coded target". A coded target is made up of a unique pattern of
squares and a central dot. The patterns are automatically detected, identified and measured
through straightforward image processing techniques. (Fraser 1997a;b) By adding such a target to
an adapter, it is possible to identify that adapter in a measurement. There are obvious
advantages to automating the use and reduction of these adapters. Once a FT is identified, a
calibration file associated with it is used to determine the type and the function of that particular
target adapter.
The simplest FT is one for measuring a plane. This FT consists of a
flat piece of aluminum with an identifying coded target applied
to one side. The other side has magnets embedded in it to hold
the FT in place. Such a target is shown in FIGURE 4.
The
relationship between the targets on the face, and the plane
represented by the base of the adapter must be determined by
a one-time calibration as explained below.
The FT is placed on a flat surface such as granite. Normal retroreflective targets are placed in the area surrounding the FT
target. These targets will be used to define a "local" plane. A
FIGURE 4 Plane target
photogrammetric measurement is then made to determine the
location of the face targets relative to the local plane. Refer to FIGURE 5. With the relationship
between the targets and the local plane known, the coded target and its adapter now become a
feature target capable of directly measuring any plane to which it is attached. The relationship
between the face targets and the plane is shown graphically in FIGURE 6.
If a right-angle bracket appropriately calibrated is used instead of a flat plate, the resulting FT can
measure a plane at right angles to the face targets. The face targets on the adapter can be in
any relative orientation to the plane that needs to be defined. For example, it might be desirable
to have the FT at a 45 angle to the plane. In fact, the targets can be in any relative orientation
providing that they can be calibrated. Moreover, multiple FTs can be placed on a block, and the
feature planes they measure intersected to form a line as shown earlier in FIGURE 2.

FIGURE 5 Example of calibration set-up

FIGURE 6 Calibration file

At run time, a local transformation is completed for each of the FTs found. This transformation uses
the calibrated and measured values of the face targets to transfer the plane, line or feature
desired into the global coordinate system of the measured object.

There are four basic types of FTs. These are shown in TABLE 1.
Type

Example

Function

Plane

The contact plane is defined.

Edge

Two contact planes and the edge

derived by their intersection are
defined.

Corner

Three contact planes, two edges and

the
corresponding
corner
are
defined.

Circle

The center point of a circle with

known radius is defined.

TABLE 1 Basic types of FTs.

FT PROPERTIES:
The targets have a number of attractive properties. These are summarized in TABLE 2.
Fast

Measurement with FTs is much faster than with other methods. The setup time
for sequential photogrammetry is very low compared to other in-place
measurement systems. Using FTs to measure features is faster and easier in
many applications than probing the feature. Photography is usually very quick.
Finally, analysis is completely automated, so it is both fast and error-free.

Higher
Accuracy

The accuracy obtained is higher than normal probed data because the FT is
measured from multiple locations, and many points are used to define the
feature.
FTs can be configured and used to eliminate many line-of-sight problems.
The method is extremely versatile in both its use and application. FTs can be
configured to meet many measuring requirements.

Line of Sight
Versatile

Minimal
Downtime

The fast overall measurement and analysis time means critical production
downtime is minimized.

Enhanced
Targeting
Repeat
Measurement

FTs can measure features that cannot be directly targeted.

The ability to automate measurements through construction macros (Ganci
and Handley, 1998) makes FTs ideal for repeat production measurements.
TABLE 2 FT properties.

MEASUREMENT PROCEDURE
Use of the FTs for measurement of NC blocks on automotive fixtures is very simple. A target is
placed on each block or surface that needs to be measured. The target selection will depend on
the required data and the viewing angles to the block. For blocks with faces that are not
perpendicular it is customary to use a combination of FTs. For pin or circle measurements it is
necessary to specify the correct radius at the calibration phase. This figure typically comes from
the design data. Some sample FT solutions are shown in TABLE 3.
Feature
Plane

Sample FT Solutions

Example 1

Example 2

Example 1

Example 2

Example 1

Example 2

Edge

Corner

Complex
corner

Centerline

TABLE 3 Sample FT solutions.

CASE STUDIES Overview

To test FTs in some actual applications, two case studies were undertaken. The studies are
described in TABLE 4.
Case Study

Object

Case 1
Fixture
measurement.

Objectives

1.
2.

Use FTs to inspect the

fixture.
To complete the
inspection in less time
than the current
inspection system.

Fixture
Case 2
In-line car
body
inspection

1.

2.

BMW Z3 Axle Carrier

Use FTs to inspect the

underbody of a
vehicle during
assembly (in-line).
To complete the
measurement in less
than the production
cycle time of five
minutes.

TABLE 4 Case Studies Overview.

For these two case studies, the FT concept was an attractive alternative to other measurement
methods. In case study 1, analysis of production tooling requires a fast and efficient measurement
system. The driving factors are time and availability for analysis. Case study 2 presents the need for
immediate information for decision making during production. In both cases the need for
uninterrupted production schedules is paramount.
In the case of production tooling measurement, unavailability of tooling during after-shift hours due
to routine maintenance needs, and constantly changing production schedules due to high
product demand, have forced inspection personnel into working during scheduled break times.
This means finding a way to measure a tool during a normal 40-minute lunch break, with the goal
of eventually completing the measurement during a 10-minute coffee break.
The ability to flexibly analyze product deviation in-line is the focus of the second case study.
Increasingly, inspection is needed at the point of origin in order to efficiently determine a potential
problem and set corrective action. The problem facing dimensional inspection is flexibility and
time. The system needs to be set-up quickly and measure in unstable environments such as
constantly moving assembly lines. It also needs to complete the measurement between assembly
processes.

CASE 1 Fixture Measurement

The first case study involved the measurement of a
relatively simple object, namely a 1.8-meter long
panel-holding fixture.
The primary objective of the
measurement was to determine the location of key
features such as corners, edges and planes on the fixture.
FTs were placed on each of the desired features.
Wherever possible, one FT was used. For features without a
suitable target the necessary data was created using a
combination of FTs. A sample clamping mechanism is
shown in FIGURE 7. Four FTs are shown. The FT at the back
of the clamp is used to define a contact plane. The three
remaining FTs are used to define the hard corner of the
clamping surface. The fixture measurement required a
total of 40 FTs to measure all the desired features.

FIGURE 7 Sample clamp

After targeting, a total of 60 photographs were

taken of the fixture. The number of photos taken
depends on the complexity of the measurement
and accuracy requirements. The photography for
the fixture was completed in approximately five
minutes.
Camera-station locations for the
measurement are shown in FIGURE 8.
The following is a summary of the statistics from the
measurement of the fixture:
No. of photos
No. of FTs
No. of scales
Scale Agreement
Accuracy RMS (mm) X,Y,Z

60
40
2
0.01mm
X
0.010
Y
0.009
Z
0.008

FIGURE 8 Camera-station network

Some of the features measured on the fixture are shown in FIGURE 9.

corresponding FT analysis.

Also shown is the

FIGURE 9 Examples of features measured

To obtain a better idea of the time saving of FTs, the same measurement was completed using
conventional stick-on targeting and the multi-camera probe system. The probes were needed to
collect data on the features that could not be targeted.

A comparison of the two measurements is shown in TABLE 5.

Measurement type

FT Solution

Number of targets
Targeting
Probing
Target removal
Photography
Processing
Analysis
TOTAL TIME

40 FTs
5 minutes
3 minutes
2 minutes
2 minutes
3 minutes
15 minutes

Stick-on and
probes
320
25 minutes
15 minutes
20 minutes
2 minutes
3 minutes
30 minutes
95 minutes

TABLE 5 - Comparison of the two measurements.

It is clear from TABLE 5 that the FT concept is far faster. The automated analysis of the FTs especially
saves time while also eliminating measurement errors.
CASE 2 Production Measurement
In the second case study FTs were used
to measure the front-axle carrier of a
BMW Z3. The measurement was a pilot
study to examine whether in-line
measurements could be used to identify
cars with bushing angle problems.
Rectifying these problems early in the
production process would ultimately
result in significant scrap-value savings
further down the production line. The
test unit is shown in FIGURE 10.
The measurement was complicated by
the need to complete it within the time
the car would be idle at a station. A
total of five minutes was set aside for targeting,
photography, and teardown. Processing the images to
yield the required data was not necessary during the fiveminute time limit.

FIGURE 10 Axle carrier

Two operators performed the measurement.

The
targeting
and
teardown
was
completed
in
approximately two minutes. A total of 18 photographs
were collected in less than two minutes. The network
used is shown in FIGURE 11. The following is a summary of
the statistics from the measurement.
No. of photos
No. of FTs
No. of scales
Scale Agreement
Accuracy RMS (mm) X,Y,Z

18
8
8
0.01mm
X
0.008
Y
0.007
Z
0.017

FIGURE 11 Camera station network

Once the processing is finished, the desired features are automatically generated. In this case
study, six planes, eight circles and four lines were created. The feature generation is shown
graphically in FIGURE 12.

FIGURE 12 Key measurement features and their geometric reductions

All of the objectives of the measurement were met. This study showed how FTs could be used to
automate the process and significantly reduce the amount of time needed to complete such a
measurement.
CONCLUDING REMARKS
This paper and the case studies described herein show that measurement using the Feature Target
concept has the following useful characteristics.

Fast overall measurement so production downtime is minimized, and in-line

measurements within one car cycle are possible.
Automated analysis - eliminates the tedious, error prone analysis of features.
Simple system setup and operation the camera is small, portable and simple to set up,
photography is quick and easy, and FTs eliminate the tedium and difficulties of probing
features.
Works in unstable environments where other instruments cannot be used.

Consequently, Feature Targets are a powerful new tool for automating and simplifying the
measurement of geometric features; in particular the measurement of NC blocks in automotive
fixtures.
ACKNOWLEDGEMENTS
The authors wish to acknowledge the assistance of the BMW Manufacturing Corporation in
providing suitable case studies for this paper. We would also like to thank John Brown and Dr
George Blaha for their editorial comment.
REFERENCES
Brown. J., 1998. V-STARS/S Acceptance Results. Boeing Large Scale Optical Metrology Seminar,
Seattle.
Fraser, C. S., 1997a. Automation in Digital Close Range Photogrammetry. First Trans Tasmin Surveyors
Conference, 12-18
Fraser, C.S. 1997b. Innovations in Automation for Vision Metrology Systems. Photogrammetric
Record, 15(90): 901-911.
Ganci, G. and Brown. J., 2000. Developments in Non-Contact Measurement
Videogrammetry. Boeing Large Scale Optical Metrology Seminar, Long Beach.

Using

Ganci, G. and Handley. H.B., 1998. Automation in Videogrammetry. International Archives of

Photogrammetry and Remote Sensing, Hakodate, 32(5): 53-58