PROINSPECT

User Manual
Version 2.9
1. Preface ......................................................................................................1
2. Starting and halting the system ..................................................................1
3. Using ProInspect........................................................................................2
4. Recipes .....................................................................................................9
4.1 Recipe settings ................................................................................... 10
4.2 Training .............................................................................................. 11
4.3 Calibration .......................................................................................... 12
4.4 Saving and retrieving recipes .............................................................. 13
5. Creating a recipe (tutorial)........................................................................ 14
6. Vision tool guide....................................................................................... 18
6.1 Search................................................................................................ 19
6.2 Edge tool ............................................................................................ 22
6.3 Caliper................................................................................................ 26
6.4 PatMax............................................................................................... 28
6.5 PatInspect .......................................................................................... 40
6.6 Blob tool ............................................................................................. 45
7. Other tools ............................................................................................... 51
7.1 Fitting tools ......................................................................................... 51
7.2 Distance tools..................................................................................... 52
8. Software Installation................................................................................. 53

Copyright ? 2006 Univision

All rights reserved.
PROInspect 1

1. Preface
ProInspect is a configurable inspection and robot guidance system. It can be
easily tailored for a specific inspection or guidance task through its intuitive
user interface.

ProInspect is based on standard PC hardware with the proven vision

technology from Cognex Corporation.

This manual contains information about using the ProInspect Vision System. It
includes information on the following topics:

- Vision Tools
- Creating recipes
- Installation
- Interfacing ProInspect

The manual assumes the reader to have a basic knowledge of the Microsoft
Windows ? operating system.

2. Starting and halting the system

ProInspect in normally set up to start automatically when the system is switched
on. The last inspection recipe that was used is activated at start time.

In order to switch off the system, the normal shutdown sequence of Microsoft
Windows must be carried out. The sequence can be started by closing the
ProInspect application and confirming that you want to switch off the system at
the exit dialog.

Important Note: powering off the system without the correct shut down
sequence could result in a damaged file storage system.

Your unit may be equipped with a uninterruptible power system (UPS). In this
case the shutdown operation is executed automatically and the system can be
power off directly.
PROInspect 2

3. Using ProInspect

Fig. 3-1 ProInspect display

Inspection On/Off button

Main functions buttons

Recipe selection

Recipe control

Pass/Fail status

Result table User level selection

Access levels
In order to simplify and avoid unwanted access to setup operations ProInspect
provides three levels of access as depicted in the following table.

Level Description
Operator Operator level only allow to alter the
display mode, Start and Stop the
Inspection and change the active recipe
Engineer Engineer level allow the user to train and
change settings such as sensitivity to
defects.

Installer Full access to all operations including

editing inspection recipes.

You can modify the access level by clicking the key icon in the bottom bar or by
selecting Access Level from the File menu.

Fig. 3-2 User level control

The system shows a dialog with the three options. You should select the
desired access level, enter a password if required and confirm with the OK
button.
PROInspect 3

Note: a password can be required in your particular installation for the Installer
and Engineer level.

This manual describes the complete set of operations available at the installer
level.

Inspection operations
In normal operations ProInspect is triggered by an external interface to carry out
the measurements and inspection tasks defined by the active inspection recipe.

One of several possible interfaces is used to start the inspection sequence:

? PC parallel port
? Serial port
? Ethernet

Other interfaces might be supported for your particular installation. Please refer
to the appendices of this manual for details of your specific interface.

The inspection sequence might involve several “views” taken by different

cameras and many inspection tasks for each view.

After the inspection is carried out the results are given back through the
selected interface and shown on the system display.

The Inspection can be disabled and the system put in a special setup mode.
This mode is provided as some configuration and setup tasks cannot be carried
out while the inspection is running.

Examples of operations that can only be carried out in setup mode include:

? Changing the active recipe

? Training models
? Creating or editing the recipe

Start / Stop buttons

The Stop and Start buttons allow to stop and restart the automatic inspection
mode.

Fig. 3-3 Start / Stop buttons

PROInspect 4

Exit button
The exit button allows to close ProInspect or shut down the system. Upon
clicking the exit button a dialog is shown to confirm the desired operation.

Fig. 3-4 Exit button and dialog

Display Mode
ProInspect provides several indications of the inspection results. The results are
displayed as a color overlay on the inspected image and individual
measurements are shown in the result table. In addition the Pass/Fail decision
is indicated in the status area.

The display mode for the graphics overlay and the result table id selected by
pressing one of the buttons of the display control as depicted by the following
table.

Fig. 3-5 Display mode control

Still The display on hold

All Display is updated at any new inspection
for the view selected with the view
control (see further in this chapter)

Defects The display is frozen to the image and

results of the last reject/failed inspection.
The system automatically switches to the
view that generated the last defect.

1st Defect Searches and display the first defective

result and switches to the Still mode.
PROInspect 5

Pass/Fail display
The Pass/Fail decision is shown in the status area. This indication is always
updated with the last inspection regardless of the setting of the display mode.

The green/red square boxes provide an hint on which view caused the reject.

First Second Third

View View View
Pass Pass Fail

Recipe selection
The recipe selection lets you select the active recipe from the list of recipes that
were saved on the disk.

You select two items:

? Recipe: the actual recipe (associated to a file with the same name in
the PC disk).
? Type: an optional type name that is associated to a Folder in the
PC file system.

The Type is merely a grouping of recipes that helps organizing recipes and
maintaining an acceptable length for the recipe list.

Fig. 3-6 Recipe selection

Recipe control
A recipe is the definition of the guidance or inspection task. A recipe consists in
a collection of Views and Objects.

A view is associated to an image taken by one of the cameras connected to the

system.

Objects are the definitions of “what to do”for each view.

Recipes, views and objects are described more thoroughly in chapter 4.

The recipe control shows a list of the views in the recipe. The view can be
opened to display the hierarchy of objects associated to that particular view.
PROInspect 6

Fig. 3.7 Recipe control

Selecting objects
One of the purpose of the recipe control is to select an item of the recipe. You
select an object by clicking its icon. A selection has several effects:

? The image associated with the view that contains the selected item is
shown in the image display area.

? The results of the objects associated with the current selection are
shown in the result table.

? Certain operations such as training, displaying live images, testing,

sampling, calibrating, showing intermediate images are carried out
exclusively for view or objects associated with the current selection.
Accessing object’s properties
Double clicking an item has the effect of opening a dialog that lets you configure
the properties of that particular object. Properties for specific items are
described further in this manual.

Result table

Fig. 3-8 Result table

Results for the selected objects are displayed in the result table. For each
object the name, the fail/pass status and specific measurement for the type of
object are shown as depicted by the above illustration.

If the recipe contains a calibration x,y measurements are shown in a double

pixel / units format:

Square brackets such as [127.52, 239,2] are used for pixels.

Round brackets such as (1.31, 2.32) are used for units..
PROInspect 7

Statistics
ProInspect can provide the following inspection statistics:

? count and percentage of rejects for each view

? count and percentage of rejects for selected objects
? bar graph of rejects per view
? live chart of reject percentages per view.

Fig. 3-9 Statistics page

Acquiring images
You can manually acquire an image for test purposes by pressing the acquire
button. If the selected view is associated to a camera that requires a “fast
trigger”the acquisiton will not take place until the trigger is received. During this
period the button will remain pressed. This operation can be aborted by
pressing the button a second time.

You can even display a live image by selecting the Live button. Live will carry on
until the button is pressed a second time.

Note: several operations cannot be carried out while live display is on. Their
controls, buttons or menus, will be disabled until live video is switched off.

Training
Training has two purposes:

? It defines the regions within the image were the analysis takes place.
? It builds reference models that will be used at run time
Training involves selecting the relevant regions by dragging a set of cursors
over the image of the part to be inspected.

You start training by clicking the Train button. Once all the regions are correctly
positioned you end training by clicking it another time
The Golden Image
Training associates positions, sizes and shapes of a set of regions with a
reference image, the “Golden Image”. This image is saved with the models in
the recipe file and retrieved each time a training session is started.
At times you might need to update the reference image for adjustments of the
camera, lighting or for part changes. In order to accomplish that you simply
need to click the Acquire button while training.
PROInspect 8

Testing

You can test objects of the recipe by clicking the Test button. You test either an
entire view or individual objects depending upon the current selection of the
recipe control. The test is carried out with the current image (shown in the
image area) . Prior testing you have the option of acquiring a new image or
retrieving an image from the disk ( Load Image from the Image menu).

The test results are displayed as graphics overlay and shown in the result
table.

Statistical Sampling

Some inspection techniques take advantage of statistical sampling. Statistical

sampling is a process where a number of images are measured in order to
extract useful statistical data. Such data are used to build better models and to
automatically define inspection thresholds.

You start sampling by clicking the Sample button. Sampling is carried out for the
selected view only. Sampling is completed when the programmed number of
images are measured. You can abort sampling by clicking another time the
sample button.

When statistical sampling ends successfully models and inspection thresholds

are updated. In order to make them non volatile you must save them with the
recipe.
PROInspect 9

4. Recipes
A ProInspect recipe is the definition of the guidance or inspection task. A recipe
is made of the entities described in the following section.
Views
Each camera connected to the system provides a view of a part to be located or
inspected. They can be multiple views of the same part or views of different
parts to be inspected simultaneously such as for a system with multiple
inspection stations.

ProInspect currently supports up to 4 cameras or views.

Sites
A Site is a container for vision tools. All the tools in a site may depend upon the
position of the site. Their search areas or region of interest are relocated
according to the result (x, y, angle) of the locating tool that provides the position
of the site. There can be sites without any registration. In this case the site is
merely a grouping of tools.
Vision Tools
Vision tools provide a pass/fail decision or some kind of result to other tools.
Tools can be combined in a recipe to obtain powerful inspection capabilities.
For example the Search tool could find accurately the position of a part before
the Caliper tool gauges a feature.

Fig. 4-1 Design / recipe edit dialog

PROInspect 10

4.1 Recipe settings

Object settings
You can change objects properties through their configuration dialogs
accessible by double clicking their icon in the recipe control.

Vision tool properties are described in detail for each tool in chapters 4, 5, 6.
Other object properties

Id’s
Each object in the recipe possess an unique Id that is shown in its dialog.
Enable/Disable status
The execution of each object can be enabled or disabled through the Enable
checkbox of its dialog.
Camera port (for views)
The view object allows to select the camera port. The current ProInspect
version supports up to 4 cameras. More camera properties can be set with the
camera properties dialog.

Fig. 4-2 View settings dialog

PROInspect 11

4.2 Training
Training establishes the region of interest of inspection tools. Furthermore it
acquires image templates for tools that require that. Training is needed after a
tool has been inserted in the recipe.

Fig. 4-3 Training a rectangular region

Start Training
Press the Train button or select the Inspect – Train menu to start training.
Select a site
Training is carried out for all the tools of the site selected with the recipe control.
If a particular site is selected, you can manipulate only the tools located in that
site. Tools that belongs to a space can be manipulated by selecting their space.
Golden Image
Train always recalls the “golden image”of the inspected part. The golden image
is stored the first time a recipe is trained. It can be changed simply by acquiring
or loading a disk image in a train session.
Tool Windows
During training you will relocate and resize the windows that appear as color
overlay on the image with the mouse pointer. The active window is shown in
yellow with pink handles. In order to resize a window you should drag an
handle. Drag the yellow lines to relocate.

Once you are satisfied with the windows positions, press again the train button
to conclude training.
PROInspect 12

4.3 Calibration
Calibration is the process with which the system establishes the relationship
between the camera space and a known space. Calibrating involves using
known data such as the coordinates of a set of points and measuring them in
the image space. Because ProInspect uses error minimizing resolution
techniques, the RMS error (root mean squared error) is provided as estimate of
the calibration quality.
In designing the calibration experiment, care should be taken to avoid
redundancy in the data that may lead to mathematical problems. A typical
situation of redundancy is when the calibration points are aligned. Good point
configurations are shown in the following figure.

Point calibration
Point calibration works with a small number of points. The acual number of
points is established at design time. The World coordinates have to be entered
in a table. The Vision coordinates are measured by dragging a set of cursors
over the calibration targets.

To start the calibration select a tool in the space you want to calibrate in the
recipe tree and activate Vision - Calibrate menu command.

The system shows a set of cursors, one for each calibration point. You should
drag the cursors over the targets as accurately as possible. The system shows
the camera coordinates in the Pixel X , Pixel Y Columns of the table in the
right side of the screen. Enter the corresponding X,Y coordinates and press
Done to finish.

Robot calibration
Robot calibration works with a single target that is placed in a set of known
position by a manipulator. Prior calibrating a locating tool must be established
and trained to locate the target. Robot calibration is done by moving the target
to each position. The Robot will issue a specific “Add Calibration Point”
command providing the robot coordinates for that point. ProInspect measures
the pixel coordinates of the target and adds them to a calibration table. After
having completed the process for the desired number of points, the robot issues
the “Calibration Solve” command.
This type of calibration offers several significant advantages. The calibration
operation is carried out automatically without the need of manual
measurements. This calibration maps pixel directly into the robot or machine
reference system.

Details of the calibration commands are depicted in appendix A.

PROInspect 13

Grid calibration
Grid calibration is done by showing a known grid of dots to the system. The
system measures the position of each dot and computes the calibration map. A
significant advantage of this type of calibration is that the large number of dots
enables a better error minimizing mathematics. With this solution it is possible
to significantly reduce errors such as the geometrical optical aberration, a
particularly significan error with some type of lenses.

4.4 Saving and retrieving recipes

Once you are satisfied with your settings, the recipe should be saved onto the
disk. To save a recipe use the File – Save menu command. When you save a
recipe the system prompts you to assign a name and a numeric code to the
recipe. The numeric code is number from 1 to 255 used for remote selection of
the active recipe. To retrieve a recipe from the disk you can use the File –
Load Recipe menu command or alternatively select the active recipe directly
with the recipe control.
PROInspect 14

5. Creating a recipe (tutorial)

This chapter describes the process of creating a new recipe. We will use a
sample project as a tutorial. We encourage you to try with your system while
you read the steps described in this chapter.

Our sample project consists in locating the position and orientation of a pipe
fitting on a conveyor and measure a the size one end to check that the part is
the correct one.
Creating a blank recipe
We start by creating a blank recipe by selecting New . from the File menu.

Editing the recipe

We open the design dialog by selecting Design from the Vision menu. The
dialog has two panes. The left pane show the object types that can be added to
the recipe. The right pane shows a tree the represents the current structure of
the recipe.
Creating a View
We create a view by dragging the camera icon from the left pane to the right
pane. Dragging an object from the left to the right pane is a used in general to
add an object in the recipe.

After dropping an object one has the option of giving it a name. We called our
view “ Conveyor View”. Names can be changed at any time by selecting the
object, clicking its name and typing in a new name.

Fig. 5-1 Design dialog

PROInspect 15

Setting view properties

To set some view properties we open the view dialog by double clicking its icon
in the recipe pane.

Object properties can be accessed also outside the design dialog by double
clocking their icons in the recipe control.

Fig. 5-2 View settings dialog

We set the trigger property to “Remote” as the inspection sequence will be

started by a command from the robot controlled that is linked ProInspect over a
serial line.

We must now test image acquisition. We close the design dialog and clock the
Live button. We can now adjust the camera and lighting.

Adding a PatMax Tool

We decided to use PatMax for locating the part as we expect them to have a
random orientation. Let’s get back to the design dialog and drag in a PatMax
object. We drop the PatMax icon in the view.

Objects in the recipe have a certain hierarchy. Vision tools such as PatMax are
always contained in views.

We must set some PatMax parameters.

? Search area is the entire image (default)

? Accept quality is 0.7
? Angle range is –180, + 180.

We do this opening the PatMax dialog by double clicking its icon.

Training the PatMax model

We will now prepare for training. We close the design dialog and press the Train
button.

A rectangular cursor is shown in the image area. Once we click near the edges
of the rectangle, small squares at each corner appear. Dragging the corners
has the effect of resizing the rectangle while dragging the edges moves it.

The small cross in the center of the rectangle is a special point called “ origin”
that will be returned as search result. We must rather accurately position the
origin.
PROInspect 16

In order to adjust its position we zoom in by clocking the right mouse button and
selecting “Zoom +”.

Once we are satisfied by the position of the model area we press the Train
button again to end training.

Testing PatMax
We can now test the PatMax search by moving the part under the camera,
acquiring an image and executing the test.

ProInspect finds the part quite accurately. Let move on to the gauging task. We
must measure one end of the fitting .

The caliper is a perfect tool for this task. The caliper must be applied with a
consistent position and orientation relative to the position of the part.
Adding a site
Sites are made for this purpose. Sites get a displacement from the associated
set of locating tools. This displacement is applied to the execution of the tools
contained in the site.

A site is added to the recipe just like any other object by dragging an icon from
the left pane to the recipe view.

Establishing the site position

Let’s look at the result in the recipe view:

The newly created site has a sub-tree with a Pre-alignment object.

The Pre-alignment is a container for the tool that provides the position to the
site.

In order to establish PatMax as the position reference for the site we just drop
the PatMax icon into the Pre-alignment.

Note that PatMax was dragged from the recipe view and not from the left pane.
The instance of PatMax here is merely a reference to an object that exists
already in the recipe.
PROInspect 17

Adding and training a Caliper

We now add the caliper. In order to be correctly positioned the caliper must be
contained in the site. We simply drop it into the site.

The caliper like most tools require training. We close the design dialog and start
training. We place the caliper region perpendicularly to fitting end that we must
measure. The caliper detects the size by locating a pair of edges and
measuring their distance,.
We finally have to set some caliper parameters. We do that with the Caliper
dialog:
? nominal size = expected size
? size tolerances
? edge polarities light to dark, dark to light.

After testing it we can now save the recipe. We select Save as from the File
menu.
PROInspect 18

6. Vision tool guide

This chapter the introduces ProInspect Vision Tools. Vision Tools analyze the
image to extract useful information such as the position or the integrity of an
object. Vision tools provide a pass/fail result as well as other measurements
such as the position of a part or the size of an object. These data can be
communicated to another system or taken into account by other operators. For
example a distance tool can consider the position of two points and calculate
their distance.

Vision tools are the basis of the powerful inspection capabilities of ProInspect.

This chapter provide a basic description of each tool and depicts the available
settings. It includes information on the following tools:

? Search: a tool to locate a pattern based on its distribution of gray

levels.

? Edge tool: a tool to locate an edge within a well-defined area of an

image.

? Caliper: a tool to locate an edge pair and measure the distance

between edges on the edge tool technique.

? PatMax: a tool to locate a pattern based on its geometrical

properties with outstanding capabilities in locating patterns
that are rotated, scaled or incomplete.

? Light a tool that chechs the distribution of gray levels against a

Meter: trained model.

? Blob tool: a tool that finds and counts objects or defects with a gray
levels that differ from the background.

? PatInspect: A tool that uses PatMax technology to detect and report

pattern defects.

For a more through description of the underlying technologies this manual

includes a “Vision Tool programmer guide” in appendix A.
PROInspect 19

6.1 Search
The purpose of Search is to locate and measure the quality of one or more
previously trained features in an image.

The search operation measures the extent to which a feature in an image

matches a previously trained model of that feature.

A feature is any specific pattern in an image. In most cases, you train a

representative model from one image and use it to search for similar patterns in
that image or in other similar images.

Search locates features by finding the area of the image to which the model is
most similar.

Figure 6-1 shows a model and an image, and the areas of the image that are
most similar to the model. An image and model similar to those shown in Figure
1 might be used to search for a single instance of a feature such as a fiducial
mark on a printed circuit board.

Fig. 6-2 Search candidates

Model
Candidates

Best Match

Search Score
Search finds the location of a pattern in a search image based on a model
image of that pattern. In addition to returning the location of the pattern in the
search image, Search also indicates how closely the pattern in the search
image matches the pattern in the model image by returning a score. The score
indicates how close a match exists between the trained image and the image
whose location was returned. Scores range from 0.0, indicating no similarity
between the model and the feature, to 1.0, indicating a perfect match.
Origin point
The location of the pattern is reported for a special point called “origin” . The
origin is normally an arbitrary point of the pattern chosen at train time.
Synthetic models
In alternative to models trained from actual images synthetic models are
generated by assuming a certain shape of the pattern to be located such as a
circle or an annulus. Synthetic models eliminate the need of manual training
and provide the additional advantage of having a non arbitrary origin point. For
example the origin of a circle is established in the center.
PROInspect 20

Search tools supported by ProInspect

ProInspect supports the following types of Search tools:

? Search rect Standard search tool with rectangular model area and
(standard): rectangular search area.

? Search circle Search tool with circular model area and rectangular
search area.

? Search annulus Search tool with annular model area and rectangular
search area.

? Search rect Search tool with rectangular synthetic model and

synth. rectangular search area.

? Search circle Search tool with circular synthetic model and rectangular
synth. search area.

Search Settings

Programmable origin option

When this option is enabled, the model origin can be set graphically at train
time. Otherwise the origin is set to the middle point of the model region.

Note: the model origin is an arbitrary point of the model region. The result of the
search is the position of this special point in the run time image.

Pattern match limits

Accept - The acceptance level for the search score. Results with
Shape Index scores below this limit are not accepted.

Contrast A tolerance for the contrast of the pattern found relative to

the model. Results with a contrast change higher than this
tolerance are not accepted.

Search range
X, Y The X, Y, in pixels, defines the area around the nominal
Search range position of the template in which the system performs the
search.

Tolerances
X, Y The tolerances for the search result. The tool fails if the
Tolerances position is found outside these tolerances.

Skip results on An option to avoid considering results that are found along
edges the edges of the search areas as these results are often
unreliable.
PROInspect 21

Number of results
The maximum number of results to search and report..
Search results
Score Match quality index. A value from 0 to 1 indicating the level
of conformity between the model and the pattern found.

Contrast A measure of the relative contrast of the pattern relative to

the model. 1.0 indicates a contrast equal to the trained
pattern. Higher or lower values indicate higher or lower
contrast.

X, Y Position of the pattern found.

PROInspect 22

6.2 Edge tool

The Edge tool offers extremely rapid and precise edge detection and location
within a well-defined area of an image.

The Edge tool differs from other vision tools in that it requires that you know the
approximate location and characteristics of the feature or edges you want to
measure or locate.

The Edge tool is typically used to develop detailed information about the precise
locations of features within an object. It is not appropriate for developing
information about the shape of objects or features.
Projections
The first step in using the Edge tool is to specify a projection region within the
image to which you want to apply the tool. The Edge tool depends on a carefully
defined projection region to isolate just the edge information from a small
section of the image.

The projection operation sums all the information in the projection region,
accentuating the strength of edges that lie parallel to the projection rays and
reducing the effects of noise.

Figure 6-3 shows how a projection can accentuate the edge information in a
two-dimensional image. Notice how the strength of the edge is stronger in the
projected image than in the initial image.

Fig. 6-4 Accentuating edges strength with projection

Projection
region

Projection
direction
Edges of
interest

Projected image

How the Edge tool detects edges

The Edge tool uses the projection region to produce a one-dimensional
representation of the portion of the image that contains the edges of interest.
This one-dimensional projection image will contain not only the edges of
interest, but also other edges caused by noise and unwanted information in the
original image. Applying a filter to the one-dimensional projection image
increases the strength of the edges of interest while at the same time
decreasing image noise.
The filtered image no longer visually resembles the input image. However it
does have an important new characteristic: a graph of the filtered pixel values
reveals that the peaks in the values, both positive and negative, correspond to
the location of the edges within the initial image. The Edge tool uses these
PROInspect 23

peaks in the filtered image to determine the location of edges in the original
image. In addition to producing an image with peaks that correspond to edges
in the input image, image filtering also removes noise and spurious edges from
the input image.

Figure 6-5 illustrates an example where the image contains two true edges
along with spurious edges caused by variations in pixel values around the edge.
When a filter with a size of 1 is applied to the image, both the edges of interest
and spurious edges appear as peaks in the filtered projection image.

Fig. 6-6 Filtering with filter size of 1

Filtered image
(filter size = 1)

Figure 6-7 shows the effect of applying a filter with a width of 2. The peaks that
correspond to the edges of interest are broader, and most of the spurious peaks
from Figure 4 are no longer present in the filtered projection image.

Fig. 6-8 Filtering with filter size of 2

Projected image

Filtered image
(filter size = 2)
PROInspect 24

Edge selection
Once it has located the edges in the original image, the Edge tool computes a
score for each potential edge pattern in the image. The score for each edge
pattern in the image is computed based on a set of selected criteria.

You control the method that the Edge tool uses to score edge pattern
candidates by performing the following steps:

? Select the active criteria’s

? Adjust the weight (importance) of each criterion.

The tool computes a separate value for each attribute for each edge candidate
based on the scoring method that you supply, then computes an overall score
for each edge pattern candidate by combining the individual scores. It finally
selects the edges with the highest score.

Edge tool settings

Tolerances
Position offset The tolerances offset of the edge position found relative to
the middle point of the search region established at train
time.

The tool fails if the offset is outside these tolerances

Accept score The score acceptance level. The score indicates the
conformity of the edge with the specified criteria’s. Refer to
the “Edge selection” section of this chapter.

Edge detection
Edge polarity The expected edge polarity. Polarity can be set to one of the
following values:

? Light to dark.
? Dark to light
? Don’t care

Only edges with the specified polarity are evaluated. All

edges are evaluated in “Don’t care” mode.

The scan direction is relevant in defining the edge polarity.

See note et the end of this section.

Contrast The contrast, in gray levels, above which a transition is

threshold considered an edge.

Filter size The filter width for edge extraction.

PROInspect 25

Note: The edge tool scan the projection from left to right.The left side is defined
at angle 0. The left side can be inverted by rotating the edge tool region. The
right is indicated by the rotating handle of the cursor.

Fig. 6-9 Scan direction

Scan direction

Right side

Edge selection- Scoring

The checkboxes in this section of the Edge tool dialog let you select which
criteria to take into account for calculating the score used for selecting the edge.

Note: The absolute value of the score for a particular criterion depends on the
its expected value.

? contrast = 100, exp. contrast = 100, weight = 1 score is 1.0 (perfect edge)
? contrast = 100, expected contrast = 200, weight = 1 score is 0.5

Expected values for contrast are set explicitly in the dialog. The expected value
for the position depends on the search region length and is set according to the
type of position criterion.

Contrast Stronger Stronger edges get higher scores.

Weaker Weaker edges get higher scores.

Custom Freely programmable scoring function.

Please refer to appendix A for details.

Position Centered Edges in the center of the search region

get higher scores. The score decreases
with the distance from the center.
Closer Edges closer to the left side of the search
region get higher scores. The score
decreases with the distance from the left
side.

Refer to the previous section for a

definition of the left side.
Further Edges further to the left side of the search
region get higher scores. The score
increases with the distance from the left
side.

Custom Freely programmable scoring function.

Please refer to appendix A for details.
PROInspect 26

Edge tool results

Score Match quality index. A value from 0 to 1 indicating the level
of conformity of the edge with the specified criteria’s

Contrast Contrast in gray levels

Position Position (x, y) of the edge found.

6.3 Caliper
The Caliper tool uses the same techniques described for the edge tool applied
to edge pairs. The caliper measures the location and spacing of pairs of edges
in images.

The caliper tool differs from the edge tool in that it locates a pair of edges
instead of a single edge. By locating two edges it reports as additional result the
distance between them.

The following tables depict the settings and outputs of the caliper tool. Most
items are identical to the edge tool.
Edge selection - Scoring
Size Stronger Stronger edges get higher scores.

Weaker Weaker edges get higher scores.

Custom Freely programmable scoring function.

Please refer to appendix A for details.

Position Centered Edges in the center of the search region

get higher scores. The score decreases
with the distance from the center.
Closer Edges closer to the left side of the search
region get higher scores. The score
decreases with the distance from the left
side.

Refer to the previous section for a

definition of the left side.
Further Edges further to the left side of the search
region get higher scores. The score
increases with the distance from the left
side.

Custom Freely programmable scoring function.

Refer to appendix A for details.
PROInspect 27

Size – Right Highest scores are given to edge pairs

(Distance whose size is closer to the specified “
between the pair expected value”
of edges)
Larger Highest scores are given to edge pairs
with larger size

Smaller Highest scores are given to edge pairs

with smaller size

Custom Freely programmable scoring function.

Refer to appendix A for details.

Caliper results
Score Match quality index. A value from 0 to 1 indicating the level
of conformity of the edge with the specified criteria’s

Contrast Contrast in gray levels

Position Position (x, y) of the edge found.

Size Distance between the pair of edges

PROInspect 28

6.4 PatMax
Like other pattern-location technologies, PatMax trains a pattern, then locates
one or more instances of that pattern in one or more search images.
PatMax offers three key features that distinguish it from other pattern-location
technologies available in machine vision:

? High-speed location of objects whose appearance is rotated, scaled,

and/or stretched
? Location technology that is based on object shape, not on grey-scale
values
? Very high accuracy

PatMax differs from other pattern-location technologies in that it is not based on

pixel grid representations that cannot be efficiently and accurately rotated or
scaled. Instead, PatMax uses a feature-based representation that can be
transformed quickly and accurately for pattern matching.

Figure 6-10 shows an image, the corresponding pattern, and the features that
make up the pattern.

Fig. 6-10 Image, pattern, and features

Note The pattern is made up of both the features and the spatial relationship
among the features. Image Pattern Features

Understanding Pattern Transformation

The appearance of an object in an image can vary in several different ways.
PatMax can
find objects whose appearance varies in any or all of the following ways:

? Size (overall size change or individual x- and y-axis size change)

? Rotation
? Location

PatMax finds trained patterns in search images no matter what combination of

transformations the pattern has undergone. You can limit PatMax to only
consider certain degrees of freedom, and within a degree of freedom, only a
specified range. By doing this you can ensure that PatMax finds all of the
variations that your application encounters in the smallest amount of time.
For each instance of the pattern that PatMax finds in a search image, PatMax
returns the location of the instance as well as values for each of the degrees of
freedom of the transformation the pattern has undergone.
PROInspect 29

PatMax also computes a score between 0.0 and 1.0 that provides an indication
of how closely the pattern in the search image matches the pattern from the
training image after accounting for the transformation that the pattern has
undergone.

How PatMax Works

This section provides an overview of how PatMax works. The information in this
section will help you understand how to get the most out of PatMax.

PatMax Patterns
When you train a PatMax pattern from an image, PatMax constructs an internal
representation of the object. The pattern is a geometric representation of the
shapes in the training image. The individual shapes, called features, are made
up of continuous boundaries between regions of dissimilar pixels in the image.

Pattern Features
A PatMax pattern is a collection of features. When you train a pattern from a
training image, PatMax isolates all of the features in the image and uses them
to train the pattern.

An individual feature is defined to be a continuous boundary between regions of

dissimilar pixels. The regions can have different intensity, contrast, or texture. A
feature can be open or closed.

Figure. 6-11 shows examples of open and closed features in an image.

Fig. 6-11 Open and closed features

PatMax pattern features are represented by an ordered list of feature boundary

points.

A feature boundary point has a location and an angle along with links to its
neighboring boundary points. The location of a feature boundary point is a point
through which the feature boundary passes. The angle of a feature boundary
point is the angle between the image coordinate system x-axis and a line drawn
through the feature boundary point perpendicular to the feature boundary and in
the dark-to-light direction.
PROInspect 30

shows the feature boundary points that define a pair of features.

Fig. 6-112 Feature boundary points

Feature Size and Pattern Granularity

The features that make up a pattern can be of different sizes, from features a
few pixels in size to features up to 50 or 100 pixels in size. Most images contain
features with a range of sizes.

PatMax uses features of different sizes to locate patterns in images. In general,

PatMax uses large features to locate a pattern in a search image quickly and
small features to determine a pattern’s location precisely.

Image Features

Fig. 6-12 Large features used for coarse location and small features for
fine location
PROInspect 31

The features that PatMax detects in an image are controlled by the granularity
used by the PatMax when it analyzes the image. To detect only the large
features in an image, PatMax uses a larger granularity setting. To detect the
small features in an image, PatMax uses a smaller granularity.
Granularity is expressed as the radius of interest, in pixels, within which
features are detected. Figure 6-13 illustrates two important characteristics of
pattern granularity.

? Large features such as the outline of the diskette are detected at both
small and large granularity settings.
? Smaller features are present or absent from the image depending on the
granularity setting.

In some cases, however, a feature might be present at a fine granularity and at

a coarse granularity, but not at an intermediate granularity.

Fig. 6-13 Pattern granularity

At the smallest pattern granularity, the trained pattern includes one or two
features for each letter on the diskette label. As the pattern granularity
increases, the number of features decreases.
In addition to affecting the features that are trained as part of the pattern,
pattern granularity also affects the spacing of boundary points along a feature
boundary. In general, the spacing of feature boundary points is approximately
equal to the pattern granularity.

PatMax uses a range of pattern granularities when it trains a pattern from an

image;

PatMax automatically determines the optimum granularity settings when it trains

a pattern. The smallest granularity used to detect features in the pattern training
image is called the fine granularity limit. The largest granularity used to detect
features in the pattern training image is called the coarse granularity limit.
You can obtain a diagnostic display that shows the actual features and feature
boundary points trained using the coarse and fine granularity limits.
PROInspect 32

Note PatMax trains the pattern using a range of granularities, not just the
coarse and fine granularity limits. The coarse and fine limits are the largest and
smallest granularities that PatMax uses.

Pattern Polarity
Each of the boundary points that describes a pattern feature has a polarity. The
polarity of a boundary point indicates whether the boundary can be
characterized as light-to-dark or dark-to-light. You can configure PatMax to find
only objects in which every boundary point has the same polarity as the trained
pattern, or you can configure PatMax to find objects with mismatched polarity.
Ignoring pattern polarity increases the variety of patterns that PatMax finds.

How PatMax Finds Patterns in an Image

This section describes how PatMax finds and reports pattern instances in the
search image.

Search Space
When you search for a PatMax pattern in the search image, you define the
search space that PatMax uses. The search space is determined by the
degrees of freedom you enable, and the range of values to consider within each
degree of freedom.

When you run PatMax, it returns a transformation that describes how the
trained pattern maps into the found instance. You can use the information in this
transformation in two ways:

• As a transformation object that you can use to convert any location from the
trained pattern to the corresponding location in the search image

• As individual values for the ordinary degrees of freedom (the location of the
pattern origin) and individual values for each of the generalized degrees of
freedom that you have enabled

When you run PatMax, you can obtain individual results for each enabled
degree of freedom, or you can obtain a single transformation object that you
can use to transform points between the client coordinate system of the training
image (translated by any nonzero pattern origin) and the client coordinate
system of the search image.
PROInspect 33

Figure 6-15 shows an example of a pattern being translated, scaled in the y-

axis, and rotated. While you could use the individual values for translation, y-
scale, and rotation change to compute where the point of interest is in the
search image, simply applying the returned transformation object to the point of
interest returns the new point of interest directly.

Fig. 6-14 Using the transformation to locate points of interest

Score
For each instance of the trained pattern that PatMax finds in the search image,
it computes a score value between 0.0 and 1.0. The score an instance receives
indicates how closely it matches the trained pattern. A score of 1.0 indicates a
perfect match; a score of 0.0 indicates that the pattern does not match at all.
When you specify the Patmax algorithm, PatMax scores instances on both the
closeness of pattern fit (the degree to which the shape of the features in the
search image conforms to the shape of the features in the trained pattern) and
the presence of clutter (extraneous features). When you specify the Patquick
algorithm, PatMax scores instances on pattern fit only.
In considering the fit, PatMax considers the shape of the pattern. Differences in
brightness or contrast (as long as the polarity is the same) are ignored. (You
can specify that PatMax ignore polarity changes in addition to brightness and
contrast changes.)

Contrast
In addition to the overall score, PatMax also returns the image contrast of each
instance of the pattern it finds in a search image. The contrast is the average
difference in grey-level values for all of the boundary points that PatMax
matched between the trained pattern and the pattern instance in the search
image.
Since PatMax computes the score for a pattern based on the shape of the
pattern, the contrast value and score value are generally independent. You can
use the contrast value to get additional information about the object.
You can specify a contrast threshold for PatMax searches. If you specify a
contrast threshold, only pattern instances where the average difference in grey-
level values for all of the boundary points exceeds the contrast threshold are
considered by PatMax.
PROInspect 34

Fit Error, Coverage, and Clutter

PatMax returns three additional score values for each pattern instance it finds in
the search image: the fit error, coverage score, and clutter score.

Fit Error
The fit error is a measure of the variance between the shape of the trained
pattern and the shape of the pattern instance found in the search image.
The fit error is computed by taking the square root of the sum of the weighted
average distances between each boundary point in the pattern and the
corresponding boundary point in the pattern instance in the search image. If the
pattern instance in the search image is a perfect fit for the trained pattern, the fit
error is 0.0. You can use the fit error to assess the degree to which the shape of
a pattern instance matches the shape of the trained pattern.

Coverage Score
The coverage score is a measure of the extent to which all parts of the trained
pattern are also present in the search image.
The coverage score is computed by determining the proportion of the trained
pattern that is found in the search image. If all of the trained pattern is also
present in the search image, the coverage score is 1.0. Lower coverage scores
indicate that less of the pattern is present.
You can use the coverage score to detect missing or occluded features.

Clutter Score
The clutter score is a measure of the extent to which the found object contains
features that are not present in the trained pattern.
The clutter score is the proportion of extraneous features present in the found
object relative to the number of features in the trained pattern. A clutter score of
0.0 indicates that the found instance contains no extraneous features. A clutter
score of 1.0 indicates that for every feature in the trained pattern there is an
additional extraneous feature in the found pattern instance. The clutter score
can exceed 1.0.

When PatMax computes the clutter score, it considers all features within the
area in the search image that corresponds to the entire image area used to train
the pattern.

Controlling the PatMax Search

This section describes how you control the operation of a PatMax search.

Degrees of Freedom
When you perform a pattern search using PatMax, for each generalized degree
of freedom (a degree of freedom other than x-translation or y-translation), you
must specify either
• That the degree of freedom is enabled or disabled
• If the degree of freedom is enabled you must specify the permitted range of
values for that degree of freedom. PatMax will find instances of the pattern that
have values for the degree of freedom within the specified zone, and PatMax
will compute and report a value for the degree of freedom.

Finally, PatMax might return some model instances that are slightly outside the
zone you specify. For example, if you specify a scale between 0.95 and 1.05,
PatMax might return results with scale values of 1.09 or 0.91.
.
PROInspect 35

Elasticity
By default, PatMax requires that each boundary point in the instance of a
pattern found in a search image closely correspond to a boundary point in the
trained pattern. PatMax can match and identify any change that can be
described by a linear geometric transformation (assuming you specify the
appropriate degrees of freedom and zones).

When patterns experience nonlinear geometric changes, PatMax can fail to find
them, or it can return a low score or inaccurate location information.

You can specify the degree to which you will allow PatMax to tolerate these
kinds of elasticity by specifying an elasticity value. You specify the elasticity
value in pixels.

In general, you should specify a nonzero elasticity value if you expect

inconsistent variation in patterns in search images.

You should keep the following points in mind when specifying a nonzero
elasticity value:

• Specifying a nonzero elasticity value does not affect PatMax’s execution

speed.
• Increasing the elasticity value does not decrease PatMax’s accuracy.
However, location information returned about additional object instances that
are found as a result of increasing the elasticity value can be less accurate.
• If the elasticity value is too low, you will see low scores and your application
may fail to find patterns in the search image and/or the positions will be
incorrect or unstable.
• If the elasticity value is too high, PatMax may match false instances and may
return inaccurate or unstable results.

In general, you should start with an elasticity value of 0; if necessary, increase

the value slowly until you obtain satisfactory results.

Ignoring Clutter When Scoring

When you specify the PatMax algorithm, you can tell PatMax to ignore clutter
when computing the score of an instance of the pattern in the search image. If
you tell PatMax to ignore clutter, then pattern instances receive the same score
regardless of the presence of extraneous features.

Using PatMax
This section describes how to use PatMax.

PatMax Training Guidelines

This section provides an overview of the parameters you specify when you train
a PatMax model.
PROInspect 36

Training Image
When you train a PatMax pattern, PatMax uses all of the information in the
pattern training image you supply. You should avoid including features in the
training image which will not be present in the search image.

In order to train a pattern from a pattern training image, the image must contain
distinguishable features. For best results, you should observe the following
guidelines when selecting a training image:

• The pattern should include both coarse and fine features.

• The pattern should include information that will let PatMax distinguish
instances that vary in all enabled degrees of freedom.

PatMax can return diagnostic information at training time that indicates whether
a trained pattern is potentially degenerate in a particular degree of freedom.

Which Features Are Trained

PatMax uses pixels from the root image upon which the pattern training image
is based for pattern training. This means that PatMax will train all features found
inside the pattern training image window, but it can cause PatMax to detect
features outside the pattern training image. The two effects of this change are

• If the pattern training window is close to the edge of the root image, feature
boundary points that are closer to the edge of the root image than the coarse
granularity limit may not be detected.
If the pattern training window is not close to the edge of the root image then it is
guaranteed that every feature inside the window is detected, regardless of the
coarse granularity setting.

• PatMax can detect features that are outside of the pattern training image if
those features are within the coarse granularity limit of the edge of the pattern
training window and if they are not within the coarse granularity limit of the edge
of the root image.
PROInspect 37

PatMax settings – PatMax page

The PatMax page contains general search and tolerances settings.
Programmable origin option
When this option is enabled, the model origin can be set graphically at train
time. Otherwise the origin is set to the middle point of the model region.

Note: the model origin is an arbitrary point of the model region. The result of the
search is the position of this special point in the run time image.
Search range
Search range Specifies one of the possible settings for defining the
mode search area.

? Entire image
? Centered Search area centered
around the position of the
model in the trained image.
The size of the search area

? Free Freely programmable

search area. The area is
graphically defined at train
time.

X, Y The X, Y, in pixels, defines the area around the nominal

Search range position of the template in which the system performs the
search.

Tolerances
Position Offset The tolerances offset of the edge position found relative to
the model origin established at train time.

The tool fails if the offset is outside these tolerances

Rotation Rotation relative to the model orientation at train time.

Number of results
The maximum number of results to search and report..
PROInspect 38

PatMax settings – Alignment page

The alignment page contains settings that influences the search algorithm.
Conformity settings
Elasticity Degree of tolerance to deformations of the pattern features
(contours).
Score The acceptance level for the global search score. Results
with scores below this limit are not accepted.

Contrast A tolerance for the contrast of the pattern found relative to

the model. Results with a contrast change higher than this
tolerance are not accepted.

Coverage Acceptance level for the coverage index. A coverage

below this limit is not accepted.
Clutter Acceptance level for the clutter index. A clutter above this
limit is not accepted.
Fit error Fit error level for the clutter index. A fit error below this limit
is not accepted.
Score using clutter When this option is enabled, the system takes the clutter
into account in computing the global score. Else, when the
option is disabled, the clutter is non relevant in determining
the score.

Tolerances – rotation and scale ranges

Angle Angle search range

Scale Scale search range

Pattern granularity limits

Auto-select option When this option is enable the system perform, at train
time, an estimate of the optimal settings for Fine and
coarse granularity.

In some cases it is interesting avoid autoselecting the

granularity values.

Fine Fine granularity limit

Coarse Coarse granularity limit

Alignment – Algorithm
Alignment Specifies what algorithm will be used for searching.

? PatQuick A quicker and less expensive variant

of the PatMax algorithm. PatQuick
provides the basic PatMax results but
lacks separate measurements of
clutter, coverage and fit error.
PatQuick may yield a lower accuracy’s

? PatMax Full PatMax

PROInspect 39

PatMax results
Score Match quality index. A value from 0 to 1 indicating the
global level of conformity between the model and the
pattern found.

Contrast A measure of the relative contrast of the pattern relative to

the model. 1.0 indicates a contrast equal to the trained
pattern. Higher or lower values indicate higher or lower
contrast.

Fit (error) The fit error is a measure of the variance between the
shape of the trained pattern and the shape of the pattern
instance found in the search image.
Coverage The coverage score is a measure of the extent to which all
parts of the trained pattern are also present in the search
image

Clutter The clutter score is a measure of the extent to which the

found object contains features that are not present in the
trained pattern.

X, Y Position of the pattern found.

Angle Angle of the pattern found relative to the model orientation.

PROInspect 40

6.5 PatInspect
The purpose of PatInspect is to detect defects. PatInspect creates the template
image of an object from a sample of training images and then compares run-
time images of the object against the template image. For PatInspect, a defect
is any change in the run-time image that is beyond the expected variation in the
training sample. A defect can be an erroneous or unwanted mark on an object,
an incorrectly shaped feature, or the absence of a feature.

PatInspect detects defects according to two distinct modes:

• Intensity difference inspection

• Feature difference inspection
Intensity Difference Inspection
Intensity difference detects differences in grey-scale pixel value between the
trained image and a run-time image. Intensity difference includes support for
image normalization so that grey-scale pixel value differences caused by
lighting variations are not marked as differences.
Feature Difference Inspection
Feature difference detects defects based on difference in feature position,
shape, or size between the trained image and the run-time image. Feature
difference mode is immune from both uniform and non-uniform intensity
differences. Feature difference is also capable of detecting sub-pixel-sized
defects.

Figure 5.1 shows an example of the intensity differences and feature

differences between two images. The folded corner of the diskette label creates
two regions within the image that have different intensity levels. These regions
are apparent in the intensity difference image.

Training
During this phase you supply PatInspect with a sample of acceptable instances
of the image that you want to inspect at run-time. For intensity difference and
blank scene inspection the output of this phase consists of a template image, a
standard deviation image and a threshold image. For feature difference
inspection the output is just a template image (see Figure 6-16).

Fig. 6-15 Overview of training

PROInspect 41

Run-time Inspection
During run-time inspection PatInspect uses the information acquired during
training to inspect a run-time image (see Figure 6-17).

Fig. 6-16 Overview of run time inspection

Intensity Difference Inspection

PatInspect intensity difference inspection follows these steps:

1. PatMax locates the alignment region in the run-time image unless you
provide an external pose.

2. The run-time region is transformed to be aligned with the corresponding

template region.

3. The transformed run-time region is normalized to the template region

using a method that you specify.

4. The normalized run-time region is subtracted from the template region

to produce a difference region

5. The threshold region is applied to the difference region.

6. The resulting “error image”is segmented into connected defects

PROInspect 42

Figure 6-18 provides a summary of the steps involved in PatInspect intensity

difference inspection. In the example the template region contains the string
“COGNEX” while the corresponding run-time inspection region contains the
string “COGNET”.

Fig. 6-17 PatInspect Intensity difference inspection

Inspection results and decision

The final result of PatInspect is the count of defects and the area for each
defect.

Defects are distinguished in “major” and “minor” defects based on their size.
Major defects have a size above a major limit. Minor defects have a size below
the major limit and above a minor limit. Defects below the minor limit are
neglected.

The pass/fail decision is than taken as described by the following table:

Major defect count > 0 Fail

No major defect, Fail

Minor defect count > limit
No major defect, minor Pass
defect count < limit
PROInspect 43

PatInspect settings
PatInspect has two type of settings.

? alignment parameters that are identical to the PatInspect tool.

? Inspection parameters\

Refer to the PatInspect chapter for a description of the alignment parameters.

Inspection parameters are described in the following table.

Intensity Inspection settings

Threshold – Sensitivity to Sensitivity index for the intensity of defects. An higher
defect intensity index corresponds to higher sensitivity. The intensity of
a defective pixel is a difference in gray levels between
the model and the inspected pattern.

Dimension Sensitivity index for the size of defects. An higher index

corresponds to higher sensitivity.
Max number of (defect) Limit for the count of minor defects.
areas
Major defect area Limit for the size of defect. Defect larger than this limit
are deemed as “major defect”

Boundaries inspection settings

Not enabled in this software version
Normalization
Normlization ? None / Identity No normalization is carried out
mode
? Histogram Derive the pixel map from the histogram
calculated during training.

A histogram of the run-time region is

generated and compared to the histogram
calculated during training. The mapping is
constructed so that the histograms
match. (That is, at any bin in the histogram
of the run-time region, the total number
of pixels with a value less than or equal to
that bin number equals the total number
of pixels in the training histogram with such
a value, within a round-off tolerance).

? Mean and std Derive the pixel map from the mean and
deviation standard deviation of the pixel grey-scale
values calculated during training.

The mapping is constructed so that the

mean and standard deviation of the
normalized run-time region will match the
mean and standard deviation from
training.
? Tails Tails matching: Derive the pixel map from
the left and right tails calculated during
training.

The tails are outlying points (representing

grey-scale values) on the left and right
ends of a histogram beyond which grey-
scale values are considered unreliable;
PROInspect 44

grey-scale values outside the tails are

discarded before the mapping is done. The
range of grey-scale values between the
tails of the test image’s histogram is
linearly mapped to the range between the
tails of the training histogram.

PatInspect results
PatInspects reports alignment results identical to PatMax.

In addition the system reports the count and the area of the larger defect found.
PROInspect 45

6.6 Blob tool

This chapter describes Blob, a vision tool that provides advanced blob analysis
of images. This chapter contains the following sections.

The blob tool is used to detect and measure two-dimensional shapes (blobs)
within images.

Blob analysis can provide your application with information about the number,
location, shape, and orientation of blobs within an image.

Image Segmentation
Since blob analysis is fundamentally a process of analyzing the shape of a
closed object, before blob analysis can be performed on an image, the image
must be segmented into those pixels that make up the blob being analyzed, and
those pixels that are part of the background.

In general, all images that are used for blob analysis start out as grey-scale
images of scenes. While it might be easy for a human observer to identify blobs
or objects within the scene, before blob analysis can analyze the image, each
pixel in the image must be assigned as an object pixel or a background pixel.
Typically object pixels are assigned a value of 1 while background pixels are
assigned a value of 0.

Several techniques can be used to segment images into object pixels and
background pixels. The following sections describe these techniques.
Binary Thresholding
The simplest technique for segmenting an image is to pick a threshold pixel
value. All pixels with grey-scale values below the threshold are assigned as
object pixels, while all pixels with values above the threshold are assigned as
background pixels. This technique is called binary thresholding or hard
thresholding.

Figure 6-19 shows an idealized scene, the application of a hard threshold, and
the resulting segmentation of the image into object and background pixels. In
the example below, a threshold value of 150 was used. All pixels with values
greater than or equal to 150 are treated as background; all pixels with values
less than 150 are treated as object.

Fig. 6-18 Segmenting an image with a binary threshold

PROInspect 46

Regional thresholding
Binary thresholding, although a simple and fast method has several drawbacks.

Binary threshold works well with high contrast blobs and images with a
homogeneous background level.

In addition it cannot cope with variations of light or reflectivity of the parts.

An alternative method of thresholding with better performances in case of non

homogeneous background and light variations is called regional thresholding.

With the regional thresholding the level of a pixel is checked against a

neighborhood of pixels instead of a fixed threshold.

There are two variants of regional thresholding.

Unsigned difference
With unsigned difference the absolute value of the difference between the pixel
level and a weighted average of its neighbors is taken into account. This
method provides slightly higher speed but cannot distinguish the color of the
blobs. The threshold value has the dimension of a grey level (from 0 to 255).
Signed difference
Signed difference take into account the sign of the difference. Thus it can
distinguish blobs having levels above or below the background. The thresholds
in this case are two signed values (up to +127 for white blobs, down to –127 for
black blobs).
Model difference
Model difference is different from other segmentation techniques in that it uses
a model image of the region of interest stored at train time. The difference
between the model and the run time image is compared to a fixed threshold.
Model difference is the segmentation method more suited to images having a
background pattern. However a fine alignment is required to inspect scenes
with strong patterns. Consider also the PatInspect tool for this purpose.
Connectivity Analysis
Once an image has been segmented into object pixels and background pixels,
connectivity analysis must be performed to assemble object pixels into
connected groups of object pixels or blobs.

Connectivity analysis uses connectivity criteria to assemble the object pixels

within the image into discrete, connected blobs. In general, connectivity analysis
is performed by joining all contiguous object pixels together to form blobs.
Object pixels that are not contiguous are not considered to be part of the same
blob.
Properties of Blobs
Once an image has been segmented, and the blob or blobs have been located
and identified, an application can begin to consider information about the blob
or blobs.

A blob is an arbitrary two-dimensional shape. The shape of a blob can be

described using a number of different measures. The measures that the Blob
tool returns are depicted in the following sections.
PROInspect 47

Figure 6.20 illustrates the geometric properties of a blob.

Fig. 6-19 Blob geometric properties

Area
The area of a blob is defined as the sum of the weighted pixel values for every
nonzero pixel in the blob. If a hard binary threshold was applied to the image to
segment it, then the area is simply the number of nonzero pixels in the blob. If
grey-scale analysis is being performed, then the area is the sum of the pixel
weights.
Perimeter
The perimeter of a blob is the length of the outside edge of the blob. A blob’s
perimeter is computed by counting the number of pixel edges between pixels
with nonzero values and pixels with values of 0.
Because computing the perimeter in this way tends to produce a perimeter that
is lager than the actual blob perimeter, a correction factor is applied to the
perimeter.
Center of Mass
The center of mass of a blob represents the blob’s balance point. If a sheet of a
uniform material were cut out in the shape of the blob, the point upon which the
blob would balance is the center of mass.

The center of mass is computed by determining the average weighted pixel

location for each pixel in the blob in the x-axis and y-axis.
Principal Axes
The major axis of a blob is the axis about which it would be the easiest
to spin the blob. The major axis of a blob always passes through the blob’s
center of mass. The minor axis of a blob is defined as the axis through the
center of mass about which the second moment of inertia is the largest. The
minor axis of a blob is the axis about which it would be the hardest to spin the
blob. The minor axis of a blob is always a line that is at 90 ? ?to the major axis
passing through the center of mass. Together, the major and minor axes are
referred to as the principal axes.
PROInspect 48

Geometric Extents
The geometric extents of a blob are the distances between the blob’s center of
mass and the four sides of a rectangle that is oriented to the blob’s principal
axes and that completely encloses all the blob’s pixels.

Inspect or measure mode

The blob tool can be used for two separate purposes.
Inspect mode
Inspect mode verifies the absence of defects. Any segmented pixel is
considered a defect. The blob tool allows to specify certain tolerance on the size
and number of defects.

Defects are distinguished in “major” and “minor” defects based on their size.
Major defects have a size above a major limit. Minor defects have a size below
the major limit and above a minor limit. Defects below the minor limit are
neglected.

The pass/fail decision is than taken as described by the following table:

Major defect count > 0 Fail

No major defect, Fail

Minor defect count > limit
No major defect, minor Pass
defect count < limit
Measure mode
Measure mode verifies the properties of the blob against a set of tolerances.
Note that the properties of a single blob are taken into account. In case of
multiple blobs found in the region of interest the larger blob is selected.

Blob settings

Blobs property page

The Blob page contains the parameters and specification relevant for the reject
decision.
Inspect mode
Defect area limit – Tolerable A limit for the total defect area.
area of all defects
Tolerable number of minor Limit for the count of minor defects.
defects
Minor defect area limit Limit for the size of defect. Defect larger than this limit
are deemed as “major defect”
PROInspect 49

Measure mode
Position offset Limit for the position offset of the blob fount relative to the train
time situation.

Rotation Limit for the rotation of the blob relative to the train time situation.

Perimeter Limit for the perimeter size relative to a specified nominal value.

Area Limit for the blob area relative to a specified nominal value.

Area/Perimeter Limit for the area/perimeter ratio relative to a specified nominal

ratio value. The ratio provides a basic indication of the blob shape.

Axis calculation ? Inertia axis Principal axes of inertia

mode
? Bounding box The axes of the bounding rectangle
(un-rotated) parallel to the x-y axis

? Bounding box The axes of the bounding rectangle

(rotated) parallel to the principal inertia axes.

Axis tolerances Tolerances for horizontal (H) and vertical (V) axis.
(H, V)

Threshold page
The threshold page contains settings that influence the segmentation process.
The threshold page includes an image panel showing the result of the
segmentation. The graphics is updated interactively at any parameter change.

The image panel shows a graphics that change with the setting of an option
mode with the following possible choices:
? Defects – bounding boxes
? Blobs – segmented pixels in red
? Threshold – mapped image – red for lighter pixels – blue for darker
pixels
PROInspect 50

Mode (of segmentation)

Mode Regional thresholding with Regional thresholding with unsigned
unsigned difference difference
? Unsigned difference Regional thresholding with unsigned
difference
? Signed difference Regional thresholding with unsigned
difference
Filter size For regional thresholding mode specifies the size of the neighborhood
of pixels taken into account to compute the difference

Sensitivity
Sensitivity Black Only “black” pixels, darker than the
mode background are segmented

White Only “white” pixels, lighter than the

background are segmented

Black and white Black and white pixels are segmented

with separate thresholds
Black Threshold for black pixels

White Threshold for white pixels

Min area Blob size limit. Blobs smaller than this limit are neglected

Blob results

Inspect mode
? Number of blobs
? Max blob size
Measure mode
The following measurement are reported for the largest blob

? Perimeter
? Area
? Shape factor (Area/perimeter ratio)
? H and V axis size
? Position (x, y of the center of mass)
? Offset (x, y relative to the model position)
? Angle (of major inertia axes)
PROInspect 51

7. Other tools

7.1 Fitting tools

The Fitting tools find the line or circle that best fits a set of points that you
supply.

You use the Fitting tool to locate lines or circles based on individual point
locations that you determine using other vision tools.

How The Fitting Tool Works

Fitting a Line
The Line tool considers the best fit line by computing the sum of the squares of
the minimum distance from each point to the line. The best fit line is the single
line that has the lowest value for this computation.

The result of the Line tool is the center position and the orientation of the line.
Fitting a Circle
The Circle tool lets you fit a set of points to a circle by minimizing the absolute
value of the signed square distance error. This method computes the radius of
the fitted circle.

The result of the circle tool is the center and the size (diameter) of the circle.

RMS Error
The Fitting tools provide with the RMS (root mean square) error an indication
about the quality of the fitting.

The RMS error is the square sum of the distances between the actual positions
of the points and their predicted values, divided by the number of points.
PROInspect 52

7.2 Distance tools

Distance tools calculates the distance between points or between a point and a
line.

The Point Distance tool takes two locating tools as input and returns the
distance between them.

The Line distance tool takes a line and a point locating tool as input and returns
the orthogonal distance.
PROInspect 53

8. Software Installation
The ProInspect software is likely to be already installed in your system. This
section provides instructions in case you should install the ProInspect software
from scratch using the installation CD provided with your system.

The PC must be prepared with the following hardware and software before
installing ProInspect:

? Microsoft Windows NT 4.0 + Service Pack 5 or higher

? Cognex 8100 PCI image acquisition module
? Microsoft Windows Installer

If your system does not already have Widows Installer the Instwmsi.exe install
program is provided in the principal CD folder.

ProInspect components are provided as “Microsoft Windows Installer” (.msi)

packages. The installer is started automatically by double clicking the install file.
Windows Installer will then guide you through the installation process.

A typical ProInspect installation requires the following components:

? Cognex 8100 driver

? Parallel Port Driver
? ProInspect core program
Cognex 8100 driver installation
A suited driver for your 8100 hardware must be installed. The installation
package is \Driver\WINNT4\Cognex 8100M.msi
Parallel Port Driver (Tinyport)
Your system might require a special parallel port driver if the parallel port is
used to interface the system. The parallel port driver installation package is
\Driver\WINNT4\Tinyport.msi
ProInspect Installation
The core ProInspect installation is the ProInspect 2.4.msi file in the principal CD
folder.