0% found this document useful (0 votes)
37 views40 pagesUnit 1
Uploaded by
pranja07Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
37 views40 pagesUnit 1
Uploaded by
pranja07Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 40
ne Course | course Structure | Pre-Requlate
Title of & L-T-P: 3-0-2 MEMECI9
Course No” Reverse Engineering
MEMEE29
COURSE OUTCOMES (COs):
‘After completion of this course, (Re students are expected to be able to demonstrate
fol ttitudes:
and attituteverse engineering and Concurrent
knowledge, skills,
and. the cone
Jds and techniques for rev’
surements.
f computer ai
followins
tee anders ept of Re
Engineering.
To understand the metho
To understand the theory of meas
To understand the application o}
geometrical modeling.
5 eeemeyerotand the applications of reverse engineerin : :
‘COURSE CONTENT:
‘UNIT-I
UNTeduction:/Introduction of Reverse and concurre
Concurrent engineering. Advantages and ‘applications/
Methodology and techniques for reverse engineering:
mermeeninge computer vision, and reverse engineerin,
verse engineering,
ded reverse engineering and
BON
ant engineering. Elements of
‘structured
Imaging, Scanner pipeline.
~UNIT-II
Theory of measurements: Linear, angular, curved surfaces, methods of advanced
‘Cnordinate Measuring Machine. Elements 10 Data
Measuring devices,
accumulation, retrieval.
UNIT-I
| Cilzputer-alded reverse engineering: 2D and SD Graphics, concer of various
Computer-aided revermeune, models, Wireframe surface and solid modeling
techniques. Visual realism and graphics tools.
INIT-IV
Rapid Prototyping: Concurrent engineering, Need for rapid ptorseyy: Post.
curing, Data retrieval from CAD, Apparatus for - quality measurement (C! ™M).
UNIT-V
‘Auto-CAD and UNIGRAPHICS, CAD/CAM
Applications of reverse engineering:
imerfaces, process planning, and computer-aided production, planning systems.
\
Capacity planning, Case Studies.
Le
| SUGGESTED READINGS:
| 1, Reverse Engineering: An
Springer.
Industrial Perspective - Raja and Fernandes,
s and Applications - Noorani, Wiley,
2. Rapid Prototyping: Principle:
3. Rapid Manufacturing: An Industrial Revolution for the Digital Age -
Hopkinson, Hague and Dickens, Wiley.
| 4. Reverse Engineering: Technology of Reinvention - Wang, CRC Press.
5. Reverse Engineering: Mechanisms, Structures, Systems & Materials -Messler,
l McGraw Hill.
Ten 21ve- By eePm/ CIT Chapter 1
Introduction to Reverse Engineering
Vinesh Raja, University of Warwick, UK
Abstract
his chapter introduces readers to the tern
ated techniques that can be used
reverse engineering, (KE), and to the
4 for scanning physical parts. Io addition, the
Fresents the process of reverse engineering and the strategy for scanning, and «
the scanned data into a 3-D surface oF solid model
1.1 Introduction
1h today’s intensely. com
stantly seeking new way:
that meet all customer
petitive global market, product enterprises are com
10 shorten lead limes for new product development
expectations, In general, product
Figure 1.1, Product development cycle 1 Introduction to Reverse Engineering
in CADCAM, rapid prototyping, and a range of new technologies that provide
business benefits. Reverse engineering (RE) is now considered one of the tech-
nologies that provide business benefits in shortening the product development
cycle. Figure 1 | below depicts how RE allows the possibilities of closing the loop
between what i “as designed” and what is “actually manufactured”
1.2. What Is Reverse Engineering?
gineering is the process of designing, manufacturing, assembling, and main.
taining products and systems. There are two types of engineering, forward engi
neering and reverse engineering. Forward engineering is the traditional process
of moving from high-level abstractions and logical designs to the physical im:
plementation of a system. In some situations, there may be a physical part!
product without any techpical details, such as drawings, bills-of-material, or
without engineering data. Mer doping ‘an existing part, subassem-
bly, or product, without drawings, documentation, or a computer model is
known as reverse engineering Hes ener also defined as the process
of obtaining a geometric CAD model from 3-D points acquired by scanning)
diguizing exsting parts! he process of digitally capturing the physical
enities of a compan as tevesecrpincering (RE) soften defined
by researchers with respect to their specific task (Motavalli & Shamsaasef 1996)
gtbellavet al. (1994) described iascaneanyaie>g part based
on an original or physical model without the use of an engineering drawing
Yau et al(1993) define RE, as the “process of retrieving new geometry from
a manufactured part by digitizing and modifying an existing CAD model”
Reverse engineering is now widely used in numerous applications, such as
‘manufacturing, industrial design, and jewelry design and reproduction For ex:
ample, when a new car is launched on the market, competing manufacturers
may buy one and disassemble it to learn how it was built and how it works, In
software engineering, good source code is often 2 variation of other good source
code. In some situations, such as automotive styling, designers give shape 10
their ideas by using clay, plaster, wood, or foam rubber, but a CAD model is
needed to manufacture the pi ganic in shape,
designing in CAD becomes more challenging and there is no guarantee that the
CAD representation will replicate the scummy
Rev ring provides a solution to this problem because the physical
models the surce of information forthe CAD model. This is as referred toas
te physical-to-digtal process depicted in Figure 1.2. Another
engineering is to compres dt development sely
competitive global market, manufacturers are co _
: nstantly seeking new w
shorten lead times to market a new product.
refers to recently developed technologies and techniques that assist manufactur
crs and designers in meeting the demands of shortened product development
time. For example, injection-molding companies need to shorten tool and die Physical Cd
Figure 1.2. Physial o dal process
evelopment time drastically easing he diners
physical product or clay mock-up can be quickly captured in the digital
remodeled, and exported for rapid prototyping/tooling or rapid man)
using multi-axis CNC machining techniques
1.3 Why Use Reverse Engineering?
Following are some of the reasons for using reverse engineering:
+ The onginal manufacturer no longer exists, but a customer needs the rod
Uuct, eg, aircraft spares required typically after an aircraft has been in sem
for several years.
The original manufacturer of a product no longer produces the prods
the original product has become obsolete.
* The onginal product design documentation has been lost or never ex:ste
+ Creating data to refurbish or manufacture a part for which there are => ~4>
data, or for which the data have become obsolete or lost.
* Inspection and/or Quality Control-Comparing a fabricated part to sts CA
description or to a standard item
+ Some bad features of a product need to be eliminated ¢g., excessive wear
‘might indicate where a product should be improved.
‘+ Strengthening the good features of a product based on long-term usage
‘Analyzing the good and bad features of competitors’ products,
Exploring new avenues to improve product performance and features
Creating 3-D data from a model or sculpture for animation in games and
movies
(7 Creating 3-D data from an individual, model or sculpture to create, scale or
reproduce artwork.
4” Architectural and construction documentation and measurement.
‘Fitting clothing or footwear to individuals and determining the anthrs
try of a population. 1 L Introduction 10 Reverse Engineering
© Generating data to create dental or surgical prosthetics, tissue engineered
body parts, or for surgical planning,
‘* Documentation and reproduction of crime scenes.
The above list is not exhaustive and there are many more reasons for using
reverse engineering, than documented above.
1.4 Reverse Engineering-The Generic Process
The generic process of reverse engineering is a three-phase process as depicted in
Figure 1. three phases are scanning, point processing, and application-
specific geometric model development. Reverse engineering strategy must con-
sider the following:
2 30
8 Scanning
. Point Mtiple
a loud Sean
8 | Processing ‘Nigment
=
Zz
2
Folygonisation
g
s
2
=
19 the generic process 1.5 Phase 1-seanning
© Rea
on for reverse engineering a part
Number of parts to be scanned single or multiple
q
ple or complex
Part material-hard oF soft
+ Part finish-shiny or dull
# Part size-large or sm
+ Part complexity
+ Part geometry-organic or prismatic and internal or external
# Accuracy required-linear or volumetric
1.5 Phase 1-Scanning
‘This phase is involved with the scanning strategy-selecting the correct scannin
technique, preparing the part to be scanned, and performing the actual scannin
to capture information that describes all geometric features of the part such a
steps, slots, pockets, and holes. Three-dimensional scanners are employed 1
scan the part geometry, producing clouds of points, which define the surfac
geometry. These scanning devices are available as dedicated tools or as add-on
to the existing computer numerically controlled (CNC) machine tools. There ar
two distinc! types of scanners, contact and noncontact
1.5.1 Contact Scanners
‘These devices employ contact probes that automatically follow the contours «
a physical surface (Figure 14). In Ithe current marketplace, contact. prob
Figure 1.4.Conac scanning touch probe. Originally published in Rapid Prototyping Casebook, Dor
4. Ryal, Cand Wimpenny, 01,2001, © John Wiley and Sons Limited Reproduced with permission of
data with
1 £0,025 to 0.2mm
space of times
iJata describing SUFFaces:
«The typical tolerance of nonce! ining is with
oonac systems have Pople generating
Jo the axis of the 16).
tact sca
aser (Figute
This cre-
‘and hence some
der before
+ Some not
which are parallel
# Nonco!
ates problem
Surfaces must be pre
scanning
employ light i!
the light impinges 0” St
pared wth a temporary
yy surfaces
ting of fine PO
Figure 1.5. Optical scanni
Tae 5. on DDN Cog pulsed in apd Pt
2s ts Sota Reece Coch, poral 1A
ih pemisin, Lo Phase 2 Point Prox essing
TaahPrate a
Figure 1.6. Vera faes-touch probe vesus a laser. hig publ In Rap Prot Casebook
McDonald, JA, yal, J. and Winpenny, 1, 2001, © John Wily and Sons Lined. Reproduced wath
permission
These issues restrict the use of remote sensing devices to arcas in engineering,
Where the accuracy of the information generated is secondary to the speed of
data capture. However, as research and laser development in optical technology
continue, the accuracy of the commercially available noncontact scanning de
vice is beginning to improve.
The output of the scanning phase is point cloud data sets in the n
‘ent format, Typically, the RE software provides a va
as raw (X, ¥, Zvalues se}
conven
such
yy of output fo
ed by space or comm,
1.6 Phase 2-Point Processing
This phase involves importing the point cloud data, reducing the noise in the
data collected, and reducing the number of points. ‘These tasks
using a range of predefined filters. It is extremely important that the tsers have
very good understanding of the filter algorithms so that they know which filter is
the most appropriate for each task. This phase also allows us to multiples
scan data sels, Sometimes, it is ‘cessary to take multiple scans of the
ensure that all required features have been scanned. This involves rot
part; hence each scan datum becomes very crucial. Multiple scan planni
direct impact on the point processing phase. Good datum pl
scanning will reduce the effort required in the point processing phase
avoid introduction of errors from merging multiple scan data A wide
commercial software is available for point processing,
\_The output of the point processing phase is a clean , merged, point cloud
set in the most convenient format his Phase also supports most of the propri
tary formats mentioned above in the scanning phase. 4.7 Phase 3-Application Geometric Model Development
the same way that developments in rapid prototyping and tooling technolo
hhorten dramatically the time taken to generatg physical rep:
oes are helping
resentations from CAD models, current RE technologies are helping to reduce
the time to create electronic CAD models from existing physical representations
he need to generate CAD information from physical components will arise
sequently throughout any product introduction process
The generation of CAD models from point data is probably the most com
plex activity within RE because(potenrs {ting algorithms pre required to
fenerate surfaces that accurately represent the three-TOTe
powusfosg usw to9'0 Moesns9¥ | soot xs0zt soz una sya] qin Butsuos yuiod-o1 104
z | wumszo0:z+| —_o9xostxooz oz-xan
3 ‘uur gy'0 0 sda saumypeur ND)
2 aerate xO OST sr-xan ‘oqoud Guguueos &
: pourwesSong) se Z-A:4 ur youd ueos] 09% €02*S0F OE-XId F214) “dio vO PuEION sun Buisuos an8oeUY
3 ww 9zzr9 sonun2>y ost) xa aqsomY
: raueyy) ——-urugigr9:beana9y ott) xa aquaso1ony| dog worssousu,
5 wus 9800
s [eves sonny oocs-o071 | wnuneja wsrvoury suze pequey2ous
& ww Te OF ‘2qoud 188un-yonor e
EB 01 960°0 4oeInD>V 002€-0021 | aBerueapy wuyosey| sevBojouy>a,, ore
2 poods pur | T
" uonendo| —‘wonnyosasoesmay| (uy) 2umnjo, eon Jurdwop
2yempiey 3 eD-2wwO>jexdh-spoyiau DEW) LE Ae
25). (b) Faro
achine-CRA
5/20 sean
mploys
of con.
r 4.2 Reverse Engineering Hardware 37
Disadvantages:
(i) slow data collection and
(11) distortion of soft objects by the probe.
3.2.2. Noncontact Methods
(Cin noncontact methods, 2-D cross-sectional images and point clouds that repre-
sent the geometry of an object are captured by projecting energy sources (light,
sound, or magnetic fields) onto an object; then either the transmitted or the
reflected energy is observed. The geometric data for an object are finally calcu-
lated by using triangulation, time-of-flight, wave-interference information, and
image processing algorithms. There is no contact between the RE hardware and
an object during data acq
There are different ways to classify RE hardware that uses noncontact
RE methods for data acquisition. These classifications are based on the sensor
technologies (Tamas et al. 2005) or data acquisition techniques (Alain 1999;
Rocchini et al. 2001) employed, Figure 3.4 presents a classification of noncontact
RE hardware based on data acquisition techniques.
‘The advantages and disadvantages of noncontact methods compared to con-
tact methods are as follows.
Advantages:
i no physical contact,
ips Aigitizing of substantial volumes;
fil) ood accuracy and resolution for common applications;
v
Non-contact
— *
Relecve
Tinsmisve
vA SN, Vi \
a
Non-optica Optical MRI
[me ON
Passive Active
Microwave radat 4 a | \NO* Coherent
Ree \ tsar
Shape rom shading Shape rom metion Time of fight
Shapelromsteeo Shapefomfocus Sucre ight Moid eects
Figure 3.4. RE hardware classification-noncontact methods sae Enpncerng Harare go Sostwae
erat panical touch probes m3
1 aletailed object®s where mee
slish the task.
say vo dete
hy to sean highh
page to aso
pemireats ent, oF reflective surfaces and
soible Inmitati anspar
yet accuracy
: vrcial RE hardware sit
she 4.2 presents some typical commercial FF mn
sahads tor data acquisition, The following > duce the mo:
ay alable noncontact RE data acquisition te
ections into
chniques.
1.22.1 Optical Techniques
1 Tranqulation
ion to determine
.d that employs
devices (CCD-
soot laser scanners use straightforward geometric tianguay
hie surtace coordinates of an object. Triangulation is @ metho
deena nal angles between light sources and photosensitive
harge coupled device camera) to calculate coordinates.
{gure 3.5. shows two variants of triangulation schemes Us
double CCD camera. In a single camera system,
sing CCD cameras:
‘a device transmits
angle and
alight spot (or line) on the object at a defined angle. A CCD camera detects the
wor oyecr Objet
1 fo a
~~ p Pi Light source
ti
i + ml:
; _#
@
he
yd
i!
a
4 « co
Figure 3.5. angulation methods a single and (b) double camera arrangement 3.2 Reverse Engineering Hardware
position of the reflected point (or line) on the surface. In a double camera sys-
tem, two CCD cameras are used. The light projector is not involved in any
measuring functions and may consist of a moving light spot or line, moving
stripe patterns, ora static arbitrary pattern (Bohler eral. 2001).
‘The principle of the triangulation method is shown in Figure 3.5a. A high-
nergy light source is focused and projected at a prespecified angle (8) onto the
surface of an object. A photosensitive device senses the reflection from the illu-
minated point on the surface, Because the fixed baseline length (L) between the
light source and the camera is known from calibration, using geometric triangu-
lation from the known angle (), the focal length of the camera (F), the image
coordinate of the illuminated point (P), and fixed baseline length (L}, the posi-
tion of the illuminated point (Pi) with respect to the camera coordinate system
canbe calculated as follows (Park and DeSouza 2005):
FL
Po Fuand
X=L-Ztand
‘The measurement errors in P and @ can be determined from the following
equation:
Fy 28g
FL L
The error in the Z measurement is directly proportional to Z' but inversely
proportional to the focal length and the baseline length. Therefore, increasing
the baseline length can produce higher accuracy in the measurement. For prac-
tial reasons, the baseline length cannot be increased at will, and itis limited by
the hardware structure of the scanners. Therefore, triangulation scanners are
commonly used for scanning small objects over short distances.
Ifthe single-point or sheet-oflight pattern is used as the light source, the tri
angulation scanner is mounted on the travel platform so that it can produce mul-
Uple surface scans. Triangulation scanners-are supplied both as complete sys
tems and as self-contained scanning heads for mounting on standard touch
probe arms or CMMs,
3221.2 Structured Light
In structuredight techniques (Park et al. 2001; Pagds et al 2003; Caspi et al
198 Page et al 200; Seymon ea 2002; Chen and Kak 198 wa ea
2004; Salvi et al. 1998; Kiyasu et al. 1995; Grin etal. 1992; Morano et al. 1998),
‘ah pile is projected ats known angle ont the surface of interest and an
mage ofthe resulting pattern reflected by the surface, is captured. The image
‘hen analyzed to calculate the coordinates of the don flares teresa
ca isht pater can be (i) a single point; i) a sheet of ligt (nels and
Strips grid or more compler coded light (Caspi ef al 198) (Figure 30) 0
ing
uw
”
ud
».
crate nin
The most commonly used pattern is a sheet of light that is aaa i
uta light beam, When a sheet of light intersects an object, nell ea
formed along the contour of the object. Ths Line is detected and the 112: <0 a
dinates of hundreds of points along the line are simultaneously calcula ty
tion. ‘The sheet of light sweeps the object as the linear slide carrying
scanning 3 n the X direction while # sequence of images js Laken
by the camera in diverete steps, An index number kis assigned to each of the
images in the order they are taken. Therefore, each k corresponds to the
X position of the sheet of light. For each image kya set of image coordinates (1 j)
the pixels in the illuminated stripe is obtained, The triples (i, j, 4)'s are the
Fange image coordinates; they are transformed to (x, y, 2) world coordinates
ssing a calibration matrix
To imp:
ve the capturing process, a light pattern containing multiple strips is,
nto the surface of an object. To distinguish betw
st be coded approximately so that the corresponde
without ambiguity (Park ef al. 2001; Pages et al. 2003; C.
al 200% Szymon ef al 2002; Chen and Kak 1987; Joaq
eal 1998; Kiyast ef al. 1995; Grin et al, 1992; Mor
Structured light systen
laser systems,
different strips,
¢ problem is solved
pi et al. 1998; Page
1 et al, 2004, Salvi
no et al 1998)
wing strong adva
sulted
ges of human beings:
s compared to
favoring structured light sys.
tems for digitizing. imay
» theda
acquisition is very fast (up
to millions of points per second)
8) color textuy
le
ns do not use a laser
G4 JE<
Single dot Sheet of ight
WES HRY
Coded strip patterns ar ia
‘nformation is av
hi) structured light syste
Gad patterns
Flgure 3.6. Different light paternsu
i structured ight techniques © Reverse Engineering: Hardware and Software
3.23 Destructive Method
The RE destructive method is useful for reverse engineering stvall and complex
Syects.on which both internal and external features ate scanned. A CNC milling
mach which are then gathered by
# CCD camesa The scanning software automatically converts the digital bitmap
nned. The company, CGI Inc,
nd calls this technology cross sectional scanning
une exposes 2 D cross sectional (slice) image
g¢ to edge detected points, as the patt is s¢
produces a destructive system,
8S) (CGI 2005).
In RP processes, the part is built layer by layer based on 2-D slice data
Pham and Dimov 2001) The destructive RE process is the reverse of this. To
model the part, 2-D slice images of the part are gathered by destroying the
art layer by layer
The data acquisition procedure of a destructive system is presented in Fig-
ere S11. The disadvantage of this method is the destruction of the object. How-
vet, the technique is fast. The accuracy is acceptable; the repeatability is
20.0127 mm (CGL systems). The layer thickness is from 0.0127-0.254 mm, The
method allows capturing internal structures. In addition, a destructive system
an work with any machinable object, induding parts made from aluminum
alloys, plastics, steel, cast ton, stainless steel, copper, and wood.
NS
~- #
Figure 3.1. procedure fr data acquisition ina destructive RE system (a) The pat and matrix combina
tion is embedded in a contrasting colored plastic main. (b,¢) The partis machined layer-by-layer to
expose the cross-sectional images.) The newly exposed surface ofthe pan. A color reproduction of his
figure canbe seen inthe color section pages 219-230), Chapter? oe
Methodologies and Techniques
for Reverse Engineering-The Potential
for Automation with 3-D Laser Scanners
Mong) Absd:, University of Tennessee, USA
David Page, Andreas Koschan,
Aostract
apiet we present methosiboye
neering (PF
ote an
ARE) to describe thas process. With this defin
adinional approsch to RE asing coordinate
4 the chapter where we exp
mn begins with data acqu
ach For these
segs the main Soc
wo CAMs. This erp
Weanners as 4 promsion 3P
and
= s, we explain
ig rates and contrast those 10 CMM
for creating CAD models using these
perlormanct Wen. we present ing pipe
includes tasks such at new repsstration, surface integration
noise removal, and other functions. This chapter
igning an RE
5 help the reader understand tbe portance
fameters have on modeling accuracy
impact tha! ranous pi
\A41 Computer-aided Reverse Engineering
Reverse engineering (RE) has many meanings to many different people. As we
bepan this chapter, we first focus our attention on the specific meaning that we
niend to address and then extend our attention to the automation of RE, which thodologies and Techniques for Reverse Engineering
mputer aided reverse engineering (CARE). To this end, this chapter
CARE in general and then explores 3-D laser scanning in depth as an
8 CARE technology from the computer vision community, as intro-
Varady er al (1997), Bernardini et al. (1999), and Page et al. (2003).
ter details the methodologies and techniques associated with computer
sniming and notes specific challenges for the CARE problem.
2.1.1 What is Not Reverse Engineering?
‘Plane of engineering has a different definition for RE. Computer engi-
Computer scientists, for example, refer to RE when they speak of de-
{he algorithmic functionality of a software package when they have no
owledge of the original software design. Engineers and programmers
‘(0 plevclop functional block diagrams of the software through interac.
n the interface and to
develop high-level code descriptions from raw
code. This software defi
tion is not the scope of our RE discussio:
example of RE that might be familiar-but also outside the scope of this
ner workings of a machine to figure out what
also a systems level a
mbles the item of interest to develo
Ship of components or to gain
the components. As w
nineeing ON V4
tsntng al wos Conpaes 30
ane medel
x ! it~
)
Le f ) ‘
Dee e t
NN CAD
Computer-aided design
Figure 2 "Congyyraie teverse engineering (CARE) process sso
Jogies and Techniques for Reverse Engineering
stand the CARE steps, consider the stages shown in Figure 2.1 from
wet w oght (The first step in the CARE process is to make measurements at
Porwdesiong the surface of the brake. Each point has an x, y, and z coordinate
Maung we point in 3-D space. For a given object, a CARE system will measure
Junaid nv even thousands of such points depending on the nature of the ob.
Wok and the type of CARE system. The collection of these points is known as
ve pow ihn example appears in the second picture from the left in Fig.
© most applications, the point cloud is a sufficient description of the
obiyecr wrever, higher levels are possible as the remaining two pictures on the
wee Som The third picture from the left is a feature description of the object,
Sees cata The final picture on the right is a full and complete CAD description
ss ect For this description, the CARE system uses the point cloud and the
Sees features to fit surfaces for modeling the entire geometry of the object
* See industrial and military applications, CARE offers many advantages
te Se {26 design, manufacturing fabrication, and field support of a part
‘of coverpo< «For example, CARE allows rapid inspection and validation in real
wpeisne Production line based on the original CAD designs. A production
’ 2 quickly evaluate tolerances relative to the CAD models. Such
fe ooles tighter control over the manufacturing process or may aid fu
~=Sewens of the component. As another example, because CARE creates
ee mesels, the technology is ideal for electronic dissemination of informa
Sefacturing center in Asia can transmit as-built models of compo-
gn enter in North America or vice versa. Although many thou
smiles separate these two centers, CARE enables more efficient
The North American team can transmit their CAD design to Asia
ec mail, and the Asian team can return a CARE model of a fabricated
eve Previously, such electronic sharing was a one-way endeavour where
=! prototype would require expensive courier-based delivery back to
“® Serica. Finally, archival problems of field modifications and out-of.
= sen components are another area for which CARE systems offer promise.
sweets often encounter situations where nontechnical personnel develop
ed smportant modifications of their designs but only after they are de
« Sech field-operations personnel are inadequately equipped to communi-
see modifications to the engineers. These modifications are typically ad hoc
- “+ or no documentation, such as CAD drawings. Thus, engineers have
pe {they wish to capture these modifications and incorporate them into
© Sewans. However, the potential exists for even nontechnical personnel to
& 2 5-D fax of the modification to the engineer. Additionally, engineers
2 similar problem when asked to modify an out-of-production
ope for which CAD documentation does not exist or has been lost. The
w= Sas the tedious task of generating a CAD model without a CARE system,
seer ss the promise of CARE to allow both technical and nontechnical indi-
8 generate engineering quality CAD models of existing objects quickly
2 cally
Ake
7
© se CARE system has detected surface edges and creases from the point > (“270
pat tae
cleus foi 2.2 Computer Vision end Reverse Engineering 15
lf a CARE system automatically captures the geometric structure of an object,
2nd stores the subsequent shape and topology ipformation as a CAD model, the
nen question might be how to achieve this goak How can a CARE system auto-
cally measure the geometry of an object? In industry, the most common
answer is through 2 coordinate measuring machine (CMM). Although CMMs
= viston Scluton that this chapter explores, they are the
ndustry standard for a discussion of CARE. So, in the next
section, we quickly review the capabilities of a CMM and then compare those
capabilities to the computer vision approach that laser scanners offer.
sarting point as a
32. Computer Vision and Reverse Engineering
Computer vision bridges diverse fields from electrical engineering to computer
science to cognitive prrholog computer vision systems seek to develop com-
puter models of the real world Through processing of image data from sensors
such 28 video cameras or-as in our case—3-D range scanners. Because computer
16 RE, we begin this section by first investigating tradi-
tional (noncomputer vision) approaches to RE, and then use these methods as
a backdrop for laser range scanners.
vision is relat
2.2.1 Coordinate Measuring Machines *
This photograph shows a measurement of the disc brake using calipers {Calipers
are a common #d-Aoc approach to RE. These measurement devices allow engi-
neers and machinists to determine accurate diameters, lengths, and other di-
mensions of objects. This approach to RE is a manual process that requires sig-
nificant effort for complicated objects and surfaces. CMM technology is the first
effort 10 automate the RE process. Before CMM and probably still popular for
most simple tasks, engineers and machinists have used measurement calipers]
For the disc brake, we could use calipers to measure the diameters of the variot!
holes and cylinders that comprise the basic shape of the brake, as in Figure 2.2
Then, from these measurements, we could manually lay out a computer model
of the brake using CAD primitive For a simple object, this manual process of
RE is straightforward, but as the Complexity of the object shape increases, the
basic CAD primitives such as planar and quadric surfaces are no longer suitable.
A free-form surface, for example, that is nonplanar and nonquadratic (Campbell
and Flynn 2600) does not lend itself readily to characterization with just caliper:
Free-form surfaces require special consideration and attention. Calipers arene)
practical for capturing their subtleties.
4s an alternative, CMMs first appeared in the early 1960s and are a more prac-
al means for characterizing and inspecting free-form surfaces. A CMM consists
of probe supported on three mutually perpendicular (x, y, and z) axes; each axis,
has a built-in reference standard’ Figure 2.3 provides a conceptual view of
\ rt
probe aoe ate meaworements sboag each axis relative to the
“ MM generates edinate pomts os the probe moves
face Opers te ran a CMM in a manaal mode where they ma
Se cred an abect and callect coordinate measurements. of they
Frgore 22 » brake using a caliper
yas
2 ew
Fagere 2.3 ooceptuad ew of a CMM that dustrates the mayor components of mast systems 2.2 Computer Vision and Reverse Engineering 17
1m the probe tomaneuver automaticaly on its own, Thislater method
ismore relevant tothe CARE definition under discussion) Different CMM many.
facturers oer diferent schemes o help operators pla the path that the probe
willow [The more advanced CMM systems allow operators upload a Ci
model of the object and then the CMM uses this model for the path pI enning
strategyThe CMM will analyze the CAD model to identify critical points and
regions-#ich as discontinuity creases or tight spaces Tight ieee eee
emphasis because the probe must come shen WTF be close to the poil
‘may progra
int due
ent occurs{If the probe cannot reach the point
ioentesion consi (ie or limited space (i, the probe
extension constraints (ie, the point is to far)
to extension constraints (1 P eater
iat location. The
is too large), then the system cannot measure at that locatic
touch or be near the measurement location. This constraint is the motivation for
computer vision approaches to CARE because computer vision offers a more
standoff measurement)
.222 Active Illumination 3-D Stereo
Computer vision attempts to develop solutions through sensors such as video
Tameras or other imaging-based systems. For an introduttion, [see Faugera
(1993), Suk and Bhandarkar (1992), or Klette et al. (1998){To acquire 3-D infor-
mation, the most well-known methodology that probably Comes to mind is ste
reovision, in which two cameras operate in a manner similar to human eyes.
Unfortunately, although stereo-based imaging has become useful in such appli-
cations as navigation planning and obstacle avoidance for Mars Rover missions,
of does not offer the accuracy that industrial RE applications de-
mand. {The major drawback of stereovision is the need to establish correspon.
dengeDetween a point in one camera’s image and the same point in the second
camera's image. If we can establish this correspondence, we can construct an
accurate stereo depth map, or range image, but correspondence is an elusive
problem. The term passive stereo is important because it indicates that the cam-
eras in-thé system do not rely on active forms of illumination, except ambient
light. One way to overcome the correspondence problem is to control illumina
tion in an intelligent manner and thus simplify the correspondence search. Such
methods are known as(active illumination stereo (or more simply active stereo)
‘where we replace one of the stereo cameras with a well-defined light source.
A laser is typically the light source of choic
Active stereo using laser-based illumination allows more accurate. depth meas-
uréments than traditional passive illumination stereo.)(We use the term depth
‘measurement as an alternative to a 3-D point measurement from CMMs, At this
Point, the sublle nuance between these two terms is not important but they are
interchangeable because a depth measurement is a 3-D point measurement.)
Although passive stereo is not viable for a CARE system, active stereo does offer
potential.\To.understand active stereo, we should consider three popular ap-
Proaches (6 illumination control. These major categories are (a) continuous ee cthod
ranu-
>»robe
CAD
ining
sand
nt of
aton
tdue
robe
nust
1 for
rore
deo
“ras
ns,
Ay Yq hdeiogies and Techniques for Reverse Engineering
Ww
eee. b) time-of flight estimation, and (c) structured light triangu
wave systems measure depth by using a coherent light source, Le,
+ ar seasure phase differences of the light wave as it travels to and from
He! ely! Femure 2.4 shows an example, The distance of an object from the cam.
ee nal to the phase difference between the transmitted wave and the
poder uer fi sve
¥ of these systems correlates with the accuracy in measuring the
PVA eases and the wavelength of the laser; Time-of-light methods operate
odar where the system measures the time required for a laser pulse to
return from an object. Figure 2.5 illustrates such a system. Highly
‘ measurements require precision electronics to measure speed-of-
Anfiehec mcrements. As a third option for active stereo, structured-light tech-
“pate depth through geometric triangulation, The camera, the object,
=< ®= ght source form the triangulation geometry. See Figure 2.6. This con.
stmilar to passive stereo geometry, except that we have replaced one
with a light source. The accuracy of these methods is a function of
ers resolution, geometric dimensions, and illumination precision, but the
> essamaeter for controlling and increasing accuracy is the camera resolu
ses Seem geometry and illumination are not as critical. Thus, structured. light
tees oer a solution more practical than passive stereo in achieving the accu-
soy seessary for an RE system. In alater section, we will explore structured-light
scSeeqers more in depth, but first we seek to compare and contrast the benefits
reo to CMM.
ao \
yo ae
Laser light
AY
Detector.)
Sper 24 Ae Srp ot a continous are stem The era a ght wae fo he
ect and back the detector Bsa wave tele btundrges 9 Poe Ne
Sr sneencei ae revea te abe sane 22 Computer Vision and Reverse Engineering 1%
cee ce
Ww
Detector
Figure 2.5 Active stereo example of ate of fight system _K pont laser emis a pub that reflected
fom objet of tee. The lerenca betwta Fe PTaTine whe the ple aa eet ad the
tee ha ety tathe deca coats wae aje tance
Pon et
Tangua
an
A afl!
ut =
et Gomera
Jo Figure 2.6. Active teeoexanle of a struc ah em Th example shows a angulation based
UL apprath where B and care typically Kan trough system cation ted
the
ned
2 Methodologies and Techniques fr Reverse Engineering
2.2.3 Benefits and Drawbacks
1 the previous sections, we discussed CMMs and active stereo as methods
CARE. Now we explore the benefits and drawbacks of these systems in cos
accuracy, speed, and reliability. Because we are not discussing specific manufac-
urers and because active stereo is an emerging technology for CARE, we limit
ourselves to a qualitative comparison at a more theoretical level, rather than
4 quantitative comparison at a marketing level.
To begin, we note that CMMs are generally mechanical systems. Although
they have computer control and use state-of-the-art electronics for measure-
‘nents, they are bound by physical principles where a measurement probe must
Physically maneuver around an object close to its surfaces, Cameras and range
Scanners, on the other hand, are solid-state sensors that take measurements
‘rom a distance. Thus they require less mechanical precision in movement. The
accuracy of a CMM is based directly on the mechanical control of the probe;but
the accuracy of an active stereo system depends more on its solid-state sensors.
‘An insightful analogy would be to consider the difference between analogue
watches with gears and springs and digital watches with transistors and quartz
Crystals. With solid-state systems, this analogy brings to mind other points of
interest such asthe benefits of fewer moving parts, the advantages of speed, and
‘he ease of software reconfigurability,
As noted previously, CMMs must contact the object of interest to measure its
geometry. By contrast, computer vision systems are noncontact in that a camera
makes measurements at some standoff distance. (We should clarify that some
commercial CMM systems claim to be noncontact devices, but they do not nec-
cssarily allow standoff measurements. They still require a measurement probe to
by near the point of measprement, just not touching it) The standoff rature of
computer vision systems offers two advantages, The firs is that they avoid dam-
age to cther the object or the measurement probe. As a probe scans and trav.
erses a surface, it may scratch that surface if the user fails to handle the probe
Bropeny or i the surface ise stoo brite or highly malleable. Additionally i
the temperature of the surface is too hot or too cold, the heat transfer could
‘damage the measurement probe, These scenarios are not a ‘major concern with
Computer vision systems. A typical laser-based range scanner can measure
Points onan object ata distance of 10 to 20 centimeters, ifnot more,
The second advantage to the standoff capability is the simplification of path
Planning, With a touch probe, a CMM (or a user) must carefully select a meas-
urement path that properly covers the surfaces of an object but that avoids
‘wedging the probe into tight spaces. The CMM must use a path that covers the
‘object and yet obeys the physical constraints imposed by interaction between
the object and the probe. Typically, a CAD model of the object is necessary to
aid path planning by serving as a predicior of the object shape. Path planning
{or a computer vision ptem is less restive, A camera or laser range scannet
can hover around the object and collect measurements just a8 a person taking
Pictures of the object might do. Some planning is necessary to ensure that we ) Structured light Range Imaping 1
measure the entite object. but the system does not have to check physical con
rants as with 4 CMM probe Again, an analogy is il
10 one of whom is wearing a blindfold. Both men are asked to describe an
«1 The man with no blindfold will simply pick up the object, turn it aro
times to obtain different views, and then report his description. Because he
Cannct see the second man must be late more careful as he uss his fingers to
probe around the object and explore its general structure by touch. His path
planning 1s obviously more complicated than that of the first man. This analogy
aks wufficient sophistication, as with most generalizing analogies. It uninten-
tvonally implies that CMMs and touch probes are less accurate than computer
1 methods, but the analogy does convey the essence of path planning with
cach approach.
The previous discussion highlights the main advantages of computer vision
systems for CARE, but in practice, their disadvantages hinder their practical use
over CMMs. As a newcomer to RE, computer vision solutions are a young tech-
nology that has not reached the commercial maturity that CMMs have attained.
Currently, CMM technology has tremendous market penetration. The existing
capital investment in CMM filts industry toward upgrading rather than replac-
ng CMM technology The old adage-if it ain't broke, don't fix it-gives credence
{0 this resistance to adopting computer vision technology. Also, commercial
computer vision systems for CARE are just beginning to emerge from the re-
search community and many research questions are stil unresolved. Among
these questions are how to improve accuracy. Beyond market penetration, the
now sipeaficant disadvantage of computer vision systems is accuracy. Ultra-
high accuracy CMs work in the 1-2 micron range, and more moderate (in
terms of cost) CMMs inthe 10-20 micron range. Computer vision methods can.
not compete—as of yet—with these levels where most systems operate in the
‘ubmullimeter range of accuracy. The trend, however, indicates that CMMs have
Pacaued. Only a few muictons difference in accuracy can result in more thea
4 SI0DK increase in cost. On the other hand, computer vision research indicates
that greater accuracy is yet to come with more precise lasers and higher resolu-
Fat aging sensors. With this hope, the next section will delve deeper into the
Freseg a ininon computer vision approach a structured-light system, and later we
‘nvestigate many of the research problems in this approach,
3 Structuredtight Range Imaging
We begin this section by first outlining three categories of structured-light solu-
light ethos tDBS imaging, Then, we selec one-o these categorien he sheet-of-
light method, and discuss it in more tea section seeks to explain the
manner in which a compute collects 3-D point information.
‘Atronder presentation is given in Klett ea. (1998) Integrated Vision Prod.
"18 (2000) gives specifics for a commercially available system. |
4
vision syster eves and Techniques for Reverse Engineering
2.2) Source Illumination Categories
© mentioned several approaches to range imaging in a previous
‘ost promising for CARE ie the structuted-ight spproahes, Re
ore triangulation methods, in which the geometry of the laser,
bbject are exploited to recover the distance from the camera to
distance, or range value, is a function of three parameters: the base~
- she object is farthest, Figure 2.7c, the projection is near the right edge.
Figure 2.7, we have used a point laser for illumination, but other types of
ces are possible, The three most common sources of illumination are
slepoint lasers, sheet-of-light lasers, and coded-light pattergs. AS already