Reverse Engineering

 Reverse Engineering is a process of obtaining a geometric CAD model

from 3-D points acquired by scanning/ digitizing existing parts/ products.
 Optimize the resources available and reduce production time cycle –
Importance of RE.
Reverse Engineering
Applications for Product design:
 New product design
 Existing product redesign
 Custom product design
 Physical model based shape modification
Reverse Engineering Process
Steps in R.E for product design
1. Data capturing
 Acquires the geometrical shape of the physical models using some
measuring devices.
 Captured Information is usually represented as a set of three-dimensional
(3D) data points, 2D slices or images.
 A variety of equipment is used for data capturing.
Steps in R.E for product design
2. Pre-processing
 The captured data are first pre-processed to facilitate subsequent surface
modeling.
 Data pre-processing steps include registration, data cleaning, data reduction
and segmentation.
 Registration transforms the measured results obtained from different setups
into a single coordinate system.
 Data cleaning wipes out all points not on the part surface.
 Data reduction reduces the number of points if they are too many to handle.
 Segmentation subdivides the combined data set into individual regions, each
representing a single geometric feature that can be mathematically represented
by a simple surface.
Steps in R.E for product design
3. CAD modelling/ model updating
 The segmented data is transformed into individual surfaces.
 Several mathematical schemes, interpolation or fitting techniques for
identifying the parameters of the geometrical shape.
 Fitting individual surfaces to measured points, the surfaces must be further
processed to form the complete CAD model.
CAD model updating:
 If a product design is directly modified on the shop floor through adjustments
to the physical prototypes.
 The CAD model must be further updated according to the changes made to the
physical part.
Data capturing
 Scanning Point Cloud data Polygon mesh/NURBS/CAD
model
 3D Scanners play a major role in data acquisition.
Contact Techniques
Non-Contact Techniques
Measuring devices
 Probing unit: provides the action(s) necessary between the measuring
equipment and an area of the part surface being digitized.
 a laser diode, projecting lens, and a laser beam.
 Detecting unit: generates related signals corresponding to the range/depth
information from the measuring probe to the probing area.
 receiving lens, a position sensitive device (PSD), a charge coupled device (CCD) and related
circuitry.
 Scanning unit: enables one to change the probing areas in the required directions
so that the entire surface of the object could be measured.
 the x, y, z guiding mechanism and their scaling units, the mechanisms of the two
rotational axes of the probe head, the servomotors, etc.
 Processing/control unit: takes care of the overall control of the measuring system
and transforms the acquired range information to 3D data.
 interfacing hardware, control box, processing computer and software.
Contact Non-destructive methods
 These methods use sensing devices with mechanical arms.
 Coordinate Measuring Machines(CMM), Computer Numerical
Control(CNC) machines are used to digitize a surface.
 Point to point sensing with touch trigger probes
 Analogue sensing with scanning probes
Coordinate Measuring Machine
Contact Destructive methods
 This method is applicable for small and complex objects.
 Internal and external features are scanned.
 A CNC machine exposes images, which are then gathered by a CCD
(Charge Coupled Device) camera.
 The scanning software automatically converts the digital bitmap image to
edge detected points as the part is scanned.
Contact Methods
Advantages
 High accuracy
 Low costs
 Ability to measure deep slots and pockets
 Insensitivity to color or transparency
Limitations
 Slow data collection
 Distortion of soft objects by the probe
Non- Contact Methods- Reflective
 In this method, 2D cross sectional images of objects are obtained by
projecting energy sources (light, sound or magnetic fields) on to it.
 The reflected energy source is observed.
 The geometric data for an object is calculated by using triangulation, time
of flight, wave-interference information and image processing algorithms.
Non- Contact Methods- Reflective
a) Triangulation
 In this method, a high energy source is focused and projected at a pre-
specified angle(θ) onto the surface of an object.
 A photosensitive device senses the reflection from the illuminated point
on the surface.
b) Structured light
 A light pattern is projected at a known angle onto the surface of interest
and an image of the resulting pattern, reflected back is captured.
 The image is then analyzed to calculate the coordinates of the data point
on the surface.
Non- Contact Methods- Reflective
Non- Contact Methods- Reflective
 Interferometry (Moire Effect)
 When two sets of multi-slit beams are superimposed spatially at a certain
angle, a beat pattern or Moire pattern is created and a set of optical
parallel planes called Moire planes are generated.
 If a 3D surface is placed in the field of view, the Moire planes will cut it
into slices and produce the Moire contour lines.
 A 2D detector detects the Moire contour lines and an image of the
contours can be recorded. The 3D information is then analyzed to
produce surface points in 3D coordinates.
 Time of Flight(TOF)
Non- Contact Methods- Transmissive
 CT- computerized tomography
 MRI - Magnetic resonance imaging
CT- computerized tomography- Transmissive Method
 A CT scan uses X-rays to create detailed pictures of organs, bones, and other
tissues. The data collected can be assembled to form three-dimensional
images. The images reveal abnormalities in both bone and soft tissues, such as
pneumonia in the lungs, tumors in different organs, or bone fractures.
MRI - Magnetic resonance imaging
 MRI also creates detailed pictures of areas inside the body, but it uses radio
waves and a powerful magnet to generate the pictures. The person also lies on
a table that moves into a doughnut-shaped device, but the doughnut is much
thicker. Similarly, these pictures can show the difference between normal and
diseased tissue.
Non-Contact Methods
Advantages
 No physical contact.
 Ability to detect colours.
 Ability to scan highly detailed objects, where mechanical touch probes may be
too large to accomplish the task.
 Fast digitizing of substantial volumes.
Limitations
 Possible limitations for colored, transparent or reflective surfaces.
Data Pre-processing for Surface Reconstruction
 The measured points are usually divided into several subsets for surface
modeling through a process called data segmentation.
 The major segmentation techniques can be classified as :
 Edge based:- Edge-based techniques identify the boundaries or edges
between surfaces from a dense set of points.
 Face based:- Face-based techniques connect regions of points with similar
properties.
Edge based segmentation
Face based segmentation
Fig: General workflow for the conversion of
data sets to 3D-printed data
physicalizations.
 For a given composition of data sets (A), a hull
is generated first (B). Here, the composition of
data sets contains a volumetric (1), point cloud
(2), graph (3), and image stack (4) data set. (C)
The enclosure, together with the available
printer resolution, thus determines the
dimension and number of the generated layers.
The data set is then processed for each layer
(D), according to “Volumes,” “Point clouds,”
“Curves and graphs,” and “Image-based”
sections, respectively (E), to generate, to
generate per-pixel material information. Here,
every layer’s pixel contains an associated
position and is given the actual data set and
additional information governing the desired
appearance of the final physical visualization.
The material information of each data set is
then composited (F) and converted to
material-mixing ratios (G). Finally, the
material-mixing ratios are dithered to binary
bitmap layers (H), one for each material given
in the printer.
Fig : Point cloud data processing
workflow and representative 3D-
printed models from point cloud
data sets.
 (A) Initial point cloud data containing point-
specific attributes. (B) Determination of
containment for the point cloud. (C) The
containment, combined with the available
printer resolution, determines the
dimension and number of the generated
layers. (D) The point cloud is processed for
each layer. (E) For each pixel within a single
layer, the point cloud is queried for nearby
points, which are interpolated and rasterized
to generate the final material data. (F)
Material information is dithered into binary
material deposition descriptions.
 (G) and (H) show representative 3D-printed models
from point cloud data sets. (G) The point cloud
representing a statue from the Tampak Siring Temple
in Bali consists of 3.6 million points and was generated
through an automated, cloud-based, photogrammetric
processing service (38). The digital elevation model of
the moon shown in (H) is represented through a point
cloud of 21 million points. The data were captured by
NASA’s Lunar Reconnaissance Orbiter, which was
launched in 2009 and has since orbited the moon (39).
Segmentation
 Segmentation partitions the entire set of points, often called the point
cloud, into individual surface areas or individual point sets, surface
reconstruction transforms each of the segmented areas/point sets into a
compact mathematical surface model.
 Depending on the type of surfaces/features to be identified and the
number and distribution of the acquired points, one may use either
interpolation or fitting techniques to transform the points into a surface
model.
MODEL REPRESENTATION
 Needs to determines a mathematical model for representing the
geometrical feature
 Types of mathematical surfaces
 Algebraic surfaces – plane, cylinder, cone or sphere – 80% engg. features
 Parametric surfaces – Bezier, B-Spline, Non-uniform rational B-spline
(NURBS)
Algebraic surfaces
 These are defined by a polynomial equation of the type f (x, y, z) = 0.
 A quadric surface defined by the following equation:

Where are free-coefficients. The above equation can

be used effectively to represent shapes such as planes, spheres, cones, cylinders
and ellipsoids.

Parametric Surfaces
 NURBS surface is a generalization of Bezier and B-spline surfaces.
 NURBS is currently the de-facto standard in the CAD industry, so most of
today's CAD systems support NURBS
 Bezier curve
1. Bezier curve always passes through the first & last control points.
2. The degree of polynomial defining the curve segment is one less than the
number of defining polygon points. Therefore for 4, control points the degree
of polynomial is three.
3. The curve generally follows the shape of the defining polygon.
4. The curve lies entirely within the convex hull formed by control points.
5. The curve exhibits the variation diminishing property. This means that the
curve does not oscillate about any straight- line move often than the defining
polygon
6. The curve is invariant under an affine transformation.
Disadvantages
 The number of control points is directly related to the degree. Therefore,
to increase the complexity of the shape of the curve by adding control
points requires increasing the degree of the curve or satisfying the
continuity conditions between consecutive segments of a composite
curve.
 Changing any control points affects the entire curve or surface, making
design of specific sections very difficult.
 B-spline
 "B-spline" stands for "Basis-spline," where "basis" refers to the set of control points that determine the
shape of the curve or surface, and "spline" implies a smooth, continuous curve or surface.
 B-spline curve is defined by a set of control points and a set of basis functions. The control points
influence the shape of the curve, while the basis functions control how much each control point
contributes to the curve's shape.
 B-splines are typically piecewise defined, meaning they are composed of multiple polynomial
segments joined together. Each segment of the curve is determined by a subset of the control points,
and these subsets can overlap.
 Control points are points in space that influence the shape of the B-spline curve or surface.
 The position of control points directly affects the shape of the resulting curve or surface. Moving
control points can deform the shape accordingly.
 “Knot" refers to a point or location in a spline curve or surface where two or more curve segments meet
 Knots play a crucial role in defining how the basis functions are distributed along the curve or surface
and, consequently, how the curve or surface is generated.
 B-spline curves and surfaces
 B-spline curve is defined by an order k (k-1 being the so-called degree), a set of
knots t and a set of n control points
 B-spline surface is defined by its orders ku and kv, two set of knots, and a set of
n =nu x nv control points
 A B-spline curve is defined by the following equation

where t is the location parameter of the curve that uniquely locates a point on
the curve,
are normalized B-splines that can be uniquely defined by their
order k and the knots t
B-spline curves and surfaces are piecewise polynomial curves and surfaces

Open knots Closed knots

 NURBS curves and surfaces
 NURBS stands for Non-Uniform Rational B-Spline.
 A NURBS curve is defined by an order k, a set of knots t, a set of n control
points and the corresponding weights.
 "weight" refers to a value associated with each control point of a NURBS curve
or surface.
 A NURBS surface is defined by its orders ku and kv , two sets of knots, a set of n
=nu x nv control points and the corresponding weights.
 NURBS curves and surfaces are further generalizations of Beziers and B-
splines.
 When all the weights of a NURBS curve are equal and constant, a NURBS curve
becomes a B-spline curve.
 When all the weights of a NURBS surface are equal and constant, a NURBS
surface becomes a B-spline surface.
Subdivision Surfaces
 This is used for modeling objects having arbitrary topology.
 The basic idea of subdivision is to define a smooth shape from an initial
control polygon (or polygonal mesh) by repeatedly and infinitely adding
new vertices and edges according to certain subdivision rules.
 A subdivision surface can be defined by a control mesh of arbitrary
topology, but not spline surfaces.
 Subdivision surfaces are becoming popular in CAD modeling due to their
capability of modeling objects with arbitrary topology.
Model Reconstruction
B-Spline based
 One needs to determine the order(s), the required set(s) of knots and the
corresponding control points.
 The entire fitting process can be divided into three major steps:
 Parametrization of the measured points for determining approximate
location parameter(s) for each of the measured points.
 Knots-selection for determining the order, the number of control points
and one or two sets of knots for curve or surface fitting, respectively.
 Least squares fitting for computing the control points.
Parametrization of Measured Points:
1. Uniform parametrization:
2. Cumulative chord length parametrization:
3. Centripetal model

4. Exponential model
5. Base surface parametrization
 This method provides a stable solution and is a convenient approach for the
parametrization of randomly distributed points.
 A first approximation of the final fitted surface and is used to allocate
parameters to the captured points for surface fitting.
 It can be constructed solely from the boundary information or boundary
curves plus some other internal cross-section curves.
 Each of the measured points is first projected onto the base surface.
 The parameters of the projected point are then used as the parameters of the
measured point.
 The projection is realized by minimizing the distance between the
corresponding measured point and a surface point.
 parametrize both regularly distributed and randomly distributed points.
Knots Allocation
 For surface fitting, one needs to determine an order, the number of control
points and a set of knots for each of the u- and v-directions, respectively.
 The number of control points should be selected such that 2 ≤ n ≤ m .
 The smaller the number of control points, the smoother the final fitted curve
or surface.
 The larger the numberof control points, the better the approximation of the
final fitted curve or surface to the measured points.
 If the number of control points equals the number of measured points with
n=m, one obtains an interpolation B-spline curve.
 The order for curve fitting should be selected such that 2 ≤ k ≤ n and the
orders for surface fitting should be selected such that 2 ≤ ku ≤ nu and 2 ≤ kv ≤
nv
 The higher the order, the smoother the final fitted curve or surface. However
the lower the order, the better the controllability for shape local
modification.
 In case k=2, the fitted curve will be a piecewise linear curve and, in case
of ku = 2 and kv = 2 , the fitted surface will be a piecewise bilinear
surface.
 When the order equals the number of control points, i.e., with k=n for
curve fitting and with ku = nu and kv = nv for surface fitting, the final
fitted curve or surface will be a Bezier curve or surface.
 For freeform shapes, one usually uses fourth order B-splines and the
final fitted B-spline curve or surface is a cubic B-spline curve or surface.
 For uniform parametrization, the knots for curve fitting
are defined by

 In case of surface fitting, the knots are defined as

 For average knots parametrization with curve interpolation n=m, the
knots for curve fitting are defined by
 Least Squares Fitting
 Surface local updating
 Computing updated control points
Surface model updating
Related Work and General Strategies
The whole procedure for local updating (Ma and He 1998) is subdivided into
the following steps:
 Initial surface identification and data segmentation.
 Identification of surface patches for each of the identified surfaces for
local updating.
 Knot insertion for region isolation and identification of control points for
local updating.
 Computing the updated control points through a fitting procedure.
Pre-Processing Steps for Surface Local Updating
 All measured points are segmented and all individual surfaces needing
updating are identified.
 all measured points are segmented and all individual surfaces needing
updating are identified.
 For each surface, the affected surface patches are further identified.
 At the same time, the measured points are also parametrized based on the
projection results.
 all identified regions are isolated through knot insertion and control points
used for local updating are identified.
 If one changes the position of a control point, the shape of the surface will
only change locally and there will be in total k uxkv patches affected by
modifying a single control point. To
EXAMPLES ON MODEL RECONSTRUCTION
Example 1: Parametrization for a B- Spline
Surface
Example 2:
B-Spline surfaces
 Initial preparation
 Manual boundary digitizing
 Rough CAD modeling
 Automatic digitizing
 Base surfaces creation
 Least-squares fitting of different features
Case study on Reverse Engineering of car wind shield