Objectives

At the end of this Section Participants will be able to:

Pattern & Logic Tools
Section 3
- Apply the Property Sheet parameters and auto-inserted
information for FindPatterns to a sample image
- Identify uses for the PatMax technology
- Configure the PatMax Pattern tools
- Create basic mathematical formulas involving If, And, InRange
and Not functions

Section 3 | Slide 2

The third section of the In-Sight Spreadsheets Standard training will focus on the Pattern and Logic At the end of this section Participants will be able to:
Tools.
- Apply the Property Sheet parameters and auto-inserted information for FindPatterns to a
sample image
- Identify uses for the PatMax technology
- Configure the PatMax Pattern tools
- Create basic mathematical formulas involving If, And, InRange and Not functions

Section 3 | Slide 1 Section 3 | Slide 2

 Steps for Application Creation Block Inspection

1. Analyze Problem
1
2. Create Prototype Job 2
3
3. Design Operator Interface
e

4. Complete & Deploy

4 4 4

Section 3 | Slide 3 Section 3 | Slide 4

1. Analyze Problem To pass, the block must meet these tests:

- Determine what needs to be inspected
- Understand what is considered Good and Bad 1. Block must be present
- Tool used: FindPatterns or PatMax
2. Create Prototype Job 2. Block must not have gouge
- Use vision tools and logic to inspect part - Tool used: ExtractHistogram

3. Design Operator Interface 3. Block must have correct length (within tolerance)
- Decide how the results will be given to the user (visual, Ethernet, discrete, etc.) - Tool used: FindSegment

4. Complete & Deploy 4. Block must have correctly sized holes (within tolerance)
- Final details for longevity and ease of use - Tool used: ExtractBlobs
- Maintenance functions like back-up and restore

Section 3 | Slide 3 Section 3 | Slide 4

 FindPatterns Applications

FindPatterns

Section 3 | Slide 5 Section 3 | Slide 6

In this section we will find a model using the FindPattern Tool. The goal of a vision application is to determine whether a part is good or bad, based on image
analysis of the part. You need to define as precisely as possible what distinguishes a good part from
a bad part in order to design an inspection system. Try listing all the criteria that must be met for the
part to pass, keeping in mind all the variations of “good” and “bad” parts.

Considerations:

- Which vision tools might you use for analysis? (Sometimes only one tool will work,
sometimes there will be a choice.)
- Do you require real world units of measure rather than pixels?
- How many cameras and which model(s) will you need to capture all of the required detail?
- Variations in color can affect how the part appears to the vision system.
- If the parts move, will you require strobed lighting?
- For parts whose position varies, how will you locate and fixture the part?
- How will you trigger your system to acquire an image?
- What outputs will your system need to send results to external hardware, for example, to
reject a part?
- How will the operator interact with the system?

Section 3 | Slide 5 Section 3 | Slide 6

 Using FindPatterns to Locate Part FindPatterns

Find Region

Model Region

(x, y)
Angle

Section 3 | Slide 7 Section 3 | Slide 8

When the part moves in the Field of View (FOV) the tools that inspect a certain area on the part must Pattern Match - Train a model part, followed by the model part searching at run time FindPatterns &
also move, this is done by “Fixturing” with a Locating tool to: PatMax®
- Use FindPatterns to Train and Search for a specific model or pattern in the image
- Determine (x, y) part location
- FindPatterns performs pattern-matching
- Determine rotation (angle)
- FindPatterns is available on all In-Sight vision systems at no additional cost
- Determine scale variations within ±10%

The x, y, and angle results will be used to locate the part for other tools.

Section 3 | Slide 7 Section 3 | Slide 8

 Using a FindPatterns Function Adding Vision Tools to the Spreadsheet

Section 3 | Slide 9 Section 3 | Slide 10

To use a FindPatterns Function: To use a vision function, drag it from the Tool Palette and drop it into the cell in the spreadsheet.

- Leave blank rows at the top for building custom operator interface later
- Always insert descriptive comments using an apostrophe

Section 3 | Slide 9 Section 3 | Slide 10

 FindPatterns Property Sheet Setting a Region

Model Region
Graphics

Find Region
Graphics

Section 3 | Slide 11 Section 3 | Slide 12

A Property Sheet is used to configure In-Sight Explorer functions whenever a function returns a Model Region and Find Region graphics (red lines) can: move, resize, and rotate a region. The
structure or the function has many input parameters. mode depends on where you place the curser.

Every Property Sheet has a table that lists the default parameters, values and expressions of a The Regions define the area of the image where the Location or Inspection Tool will perform its
function. Each row of a Property Sheet consists of a parameter name and value, or a group of operation, and are also referred to as the Region of Interest (ROI).
expanded/collapsed parameters and values.
NOTE: The entire Model Region must be within the Find Region to be found.
The FindPatterns Property Sheet includes the following:
- Image – Reference to target image cell
- Fixture – Where tool should fixture itself
- Model Region – Region specifying model features
- Model Settings – Advanced settings (more…)
- Find Region – Region specifying searching area
- Number to Find – How many instances to look for
- Angle Range – ± rotation tolerance
- Scale Tolerance – Enable +/-10% size variations
- Thresh: Accept – Minimum score needed for match
- Thresh: Confuse – Minimum score needed for speed match
- Timeout – Max amount of time spent
- Show – Graphic options for display

TIP: Property Sheet parameters support drag-and-drop placement of their controls into the
spreadsheet. By selecting a parameter from the Property Sheet and then dragging that parameter to
the spreadsheet, the parameter’s label and edit control will be automatically created and referenced
in the spreadsheet.
Section 3 | Slide 11 Section 3 | Slide 12
 How to Set Model or Find Region FindPatterns: Model Settings
1.

2.

3.

4.

Section 3 | Slide 13 Section 3 | Slide 14

To set the Model or Find Region: The Model Settings specifies the model training parameters.

1. Select Model Region from the Property Sheet

- Model Type – Specifies area model or edge model training.
2. Click edit graphics
3. Use graphics (Red Lines) to select features for the model on the image - Coarseness – Specifies the size of the smallest features in the trained model.
4. Click green check to OK selection - Accuracy – Specifies the tradeoff between accuracy/reliability and execution speed.
- Offset Row – Specifies the row offset (-479 to 479; default = 0) from the model’s center to
the response point, as measured in the model’s local coordinate system.
- Offset Column – Specifies the column offset (-639 to 639; default = 0) from the model’s
center to the response point, as measured in the model’s local coordinate system.
- ForceTrain – Enables/disables automatic model training on spreadsheet updates. When
ForceTrain is OFF, the model is trained only when the Model Region or Model Settings are
changed. When ON, the model will be re-trained every time the spreadsheet executes.
- Patterns – Specifies whether the FindPatterns function will search using its own model, or
a model contained in another FindPatterns function. The default value is 0, which means
that current FindPatterns function will use its own model.
NOTE: This behavior does not chain; it is invalid to reference a PatFind model that, in turn,
references another PatFind model.

Section 3 | Slide 13 Section 3 | Slide 14

 FindPatterns: Model Settings: Area Model FindPatterns: Model Settings: Edge Model

• Uses pixel values in region

• Creates a geometric model of just
• Normalized correlation search the edges
• Use when all pixels values in region • Use when edges are what is
should match model important, not pixel grayscale value
in region

Look for green outline of boundary

Section 3 | Slide 15 Section 3 | Slide 16

The Area Model training creates patterns based on a uniform sampling of greyscale pixel values The Edge Model training creates patterns based on a sampling of points biased to the immediate
from the Model Region. The area model similarity metric is based on the normalized greyscale vicinity of the greyscale discontinuities typically found on object’s boundaries. The edge model
correlation coefficient. similarity metric is based on the normalized comparison of greyscale derivatives.

The Area Model may be used when: The Edge Model may be used when:
- Model Region is small - Back lighting
- No well-defined edges - Non-linear lighting changes, such as shiny metal parts
- Speed is critical - Significant changes in focus or background

NOTE: The maximum number of points that may be used to define a model is 4096. NOTE: The maximum number of points that may be used to define a model is 4096.

Section 3 | Slide 15 Section 3 | Slide 16

 FindPatterns: Coarseness FindPatterns: Accuracy

Section 3 | Slide 17 Section 3 | Slide 18

Coarseness specifies size of the smallest features in trained model. Accuracy specifies the tradeoff between accuracy/reliability and execution speed.

- 0 = Fine – Smallest features are approximately 4 pixels in size - 0 = Accurate – Specifies higher accuracy/reliability and slower execution speed
- 1 = Medium (default) – Smallest features are approximately 4 to 8 pixels in size - 1 = Medium (default) – Moderate accuracy and speed
- 2 = Coarse – Smallest features are larger than 8 pixels in size - 2 = Fast – Specifies lower accuracy/reliability and faster execution speed

Section 3 | Slide 17 Section 3 | Slide 18

 Coarseness & Accuracy Coarseness & Accuracy Examples

Coarseness Accuracy Min Feature Edge Tolerance Coarseness: Coarse

Fine Accurate 1-2 0.5-1.5 Accuracy: Fast

Fine Medium 2-4 0.5-2

Fine Fast 4-6 1-2.5

Medium Accurate 4-6 0.5-2.5 Coarseness: Medium

Medium Medium 6-8 1-3 Accuracy: Medium
Medium Fast 8-10 1.5-4

Coarse Accurate 10-12 0.5-3

Coarse Medium 12-14 1-4 Coarseness: Fine

Coarse Fast 14-16 2-6 Accuracy: Accurate

Section 3 | Slide 19 Section 3 | Slide 20 3-

20
For Edge Models, Coarseness and Accuracy together determine the size of features to look for and The green outlines around the Cognex name shows the differences in the coarseness and accuracy
the tolerance for the edge locations in training. The following table shows the translation between – you will notice in the top example that the outline is not recognizing any of the letters, while in the
these two parameters and the minimum feature size and tolerance for edge movement in the trained bottom example all edges of the letter are found.
pattern.
Notice the position of the red reticule – to the human eye, there is no difference between the
Min Feature – the minimum size (in pixels) of the dark and light region on each side of an edge for Coarse/Fast and the Fine/Accurate, but to the vision tool it would be off by a few pixels. A pixel could
that edge to be included in the model. If the feature size is small, all of the really fine features in the be a mm or more.
pattern will be included in the model, but too much noise might be included for a very coarse model. If
the feature size is coarse, the noise will be ignored, but so will the very fine features if they exist.

Edge Tolerance – the maximum amount that an edge can move (± in pixels) from the trained
location and still give a strong response. If this tolerance is small, the pattern can be located more
accurately, but small steps must be taken during run time to ensure that the pattern is found, resulting
in a slower search. If this tolerance is large, the pattern can't be located as accurately, but larger
steps can be taken during run time, resulting in a faster search.

For Area Models, the Edge Tolerance numbers derived from Coarseness and Accuracy are stored
during training so that coarser or finer searching can be done, resulting in a small difference in search
speed.

For both types of models, an Accuracy setting of Fast causes the run-time to evaluate fewer similar
looking possible results than an Accuracy setting of Accurate. This also results in a small speed
difference.

Section 3 | Slide 19 Section 3 | Slide 20

 Accept & Confusion Thresholds Accept & Confusion Thresholds

Phase 1 – Lower Resolution Phase 2 – Full Resolution

Accept (50) Confusion (70)

Accept (50) Confusion (70) 0 100

0 100
No Yes

Maybe Re-examine Maybe’s

Section 3 | Slide 21 Section 3 | Slide 22

Phase 1: Phase 2:

- Uses lower resolution - Only runs if Phase 1 returns an insufficient number of valid results
- Fast - Uses full resolution
- Ignores candidates below Accept threshold - Evaluates only those candidates that scored between the Accept and Confusion thresholds
- Considers candidates above Confusion threshold to be matches

Section 3 | Slide 21 Section 3 | Slide 22

 Accept & Confusion Thresholds

Accept (70) Confusion (70)

0 100

Fastest

NO YES Using PatMax to Find a Model

No MAYBE Zone

Accept (50) Confusion (100)

0 100
Most
Accurate
NO Large MAYBE Zone

Section 3 | Slide 23 Section 3 | Slide 24

This slide depicts how to make the tool use only Phase One or Phase Two. Results would only deal In this section we will find a model using the PatMax Tool.
with the accuracy of the ‘found’ pattern.

Section 3 | Slide 23 Section 3 | Slide 24

 What is PatMax? How Does it Work?

PatMax uses two tools:

FindPatMaxPatterns
- Finds features in an image based
on the trained pattern

TrainPatMaxPattern
PatMax: Feature-based FindPatterns: Pixel grid-based
- Extracts and trains a pattern: a
trained geometric description of an
object you wish to find
- PatMax is a pattern-based search technology.
- The models in PatMax use geometric (feature based) representation;
FindPatterns is pixel grid-based. Feature based is more accurate.

Section 3 | Slide 25 Section 3 | Slide 26

A PatMax pattern is a collection of geometric features where each feature is a point on the boundary PatMax offers three key features that distinguish it from other pattern-location technologies in
between two regions of dissimilar pixel values. PatMax calculates geometric representations of machine vision:
curves. This makes it more accurate than FindPatterns, which approximates curves with line - High-speed location of objects whose appearance is rotated, scaled and/or stretched
segments that follow the pixel grid. PatMax can be more accurately rotated or scaled. - Location technology that is based on object shape, not greyscale values
- Very high accuracy

PatMax uses two tools:

- FindPatMaxPatterns
- Finds features in an image based on the trained pattern
- TrainPatMaxPattern
- Extracts and trains a pattern; a trained geometric description of an object you wish to
find

Section 3 | Slide 25 Section 3 | Slide 26

 Advantages of PatMax TrainPatMaxPattern Parameters

• PatMax location results have a very

high level of accuracy

• PatMax can deal with difficult situations

such as
- shiny parts with reflecting light
- desired feature is similar to other features in
background

• PatMax is an option available on In-

Sight vision systems at an additional
cost

Section 3 | Slide 27 Section 3 | Slide 28

Situations that require the accuracy and reliability of the PatMax Patterns Tool: The TrainPatMaxPattern Property Sheet includes the following:

- When variations in lighting and reflections are difficult to control. For example, metal parts - Image – Reference to target image cell
can reflect light in random directions. - Fixture – Where tool should fixture itself
- When the pattern being inspected is similarly shaped or shaded in comparison to something - Pattern Region – Region specifying features to train
in the background, or the pattern is being overlapped or partially hidden by other objects in - External Region – Allows use of custom region
the image. - Pattern Origin – Location within pattern to report
- When you want to accurately recognize one type of pattern from other, similar, patterns. - Pattern Settings – Specialized settings
- When the conditions of your deployment environment are too demanding for the Patterns - Algorithm – Choose from PatQuick or PatMax
Location Tool to consistently and reliably. - Elasticity – Specifies allowed perimeter deviation
- Ignore Polarity – Toggles to check for opposite polarity
- Sensitivity – Disabled
- Coarse Granularity – Granularity used to find large features
- Fine Granularity – Granularity used to find small features
- Reuse Training Image – Model image is saved for later use (retrain)
- Timeout – Milliseconds before tool gives up
- Show – Graphic options to display

NOTE: Any time you make region changes to the TrainPatMaxPattern tool (directly or through an
EditRegion), the model is retrained on whatever is inside the training region at that time – even with
the “Reuse Training Image” selected. If that is not selected, any changes to TrainPatMaxPattern will
retrain the model.

The tool is disabled by default.

GetTrained indicates if a pattern was trained or not.
Section 3 | Slide 27 Section 3 | Slide 28
 Algorithm Elasticity

Choose from two algorithms:

PatQuick PatMax

Trained Pattern Elastic Change on Part

Section 3 | Slide 29 Section 3 | Slide 30

Choose from two algorithms: Elasticity specifies the amount of non-linear perimeter deviation (0 to 10; 0 = default).
- 0 = No Tolerance
PatQuick - 1 or Greater = Flexible Boundary
- Optimized for speed
- Uses coarse granularity (details later) A linear change in a pattern is where the whole pattern changes in the same way. For example, if a
cross were twice as large all over, as the model, that would be a linear change, called Scale. But in
PatMax the cross at the right, only part of the cross has changed. For example, it could be made of rubber,
- Optimized for accuracy and the top section was pulled out at the corners. This is a non-linear change.
- Uses fine granularity (details later)
Increasing the elasticity parameter makes it more likely to find a part such as this. But if you increase
NOTE: The color of a trained feature (Green, Yellow and Red) represents the quality of the feature the elasticity value too much, you run the risk of a wrong feature being selected.
candidate for matching. Green represents high quality; red represents low. Yellow is intermediate.
The Granularity parameters should be adjusted until all trained features are green or you should
choose a better image to train. Distinct features (edges) and good image contrast will yield the best
results.

Section 3 | Slide 29 Section 3 | Slide 30

 Ignore Polarity FindPatMaxPatterns Parameters

Section 3 | Slide 31 Section 3 | Slide 32

The Ignore Polarity defines whether or not found patterns can contain color-inverted match features, The FindPatMaxPatterns Property Sheet includes the following:
in respect to the Model pattern (default = Off, unchecked). When Ignore Polarity is applied (On,
Checked), detected patterns with color inverted features, such as black/white vs. white/black in the - Image –Reference to target image cell
Model pattern, for example, will be classified as matching the Model pattern. - Fixture – Where tool should fixture itself
- Find Region – Region specifying searching area
Ignore Polarity instructs PatMax to check for original polarity identified at training and opposite - External Region –Allows use of custom region
polarity as well. This may increase the execution time for PatMax. - Pattern – Reference to a TrainPatMaxPattern
- Number to Find –Number of match to search for
- Accept –Minimum score necessary
- Contrast – Specifies lowest contrast necessary
- Clutter in Score – Toggles consideration of extra features
- Outside Region –Percentage allowed out of region
- Find Tolerances – Sets rotation, scale & aspect ratio range
- Find Overlapping –Sets allowable overlapping
- Timeout – Milliseconds before tool gives up
- Show – Graphic options to display

Section 3 | Slide 31 Section 3 | Slide 32

 Accept Threshold Find Tolerances

Allowable rotation of
match relative to model

Allowable scaling range

0 80 100 relative to pattern
(100% means no scaling)

Not Valid Accept Valid

Matches Threshold Matches

Section 3 | Slide 33 Section 3 | Slide 34

Accept Threshold defines the degree of similarity that must exist between the model pattern and Find Tolerances – Specifies the settings for finding patterns that are rotated or scaled with respect
the found pattern. The minimum acceptable score will vary depending upon the Location Tool to the trained pattern.
selected. Increasing the Accept value reduces the time required for the search.
- Angle Start – Specifies the angle at which to begin searching for matches, in degrees
PatMax uses this to determine if the match represents a valid instance of the model within the search counter-clockwise (-180 to 180, default = -15).
region. Increasing the Accept value reduces the time required for search. - Angle End – Specifies the angle at which to stop searching for matches, in degrees
counter-clockwise (-180 to 180; default = 15).
- Scale Start – Specifies the scale at which to begin searching for matches (1 to 10,000;
default = 100).
- Scale End – Specifies the scale at which to stop searching for matches (1 to 10,000;
default = 100).
- Aspect Ratio – Limits match-finding to uniform scale only or allows aspect ratio in X or Y, in
addition to a uniform scale.
NOTE: The Aspect Start and Aspect End parameters are disabled if Uniform Scale
Change is selected.
- 0 = Uniform Scale Change
- 1 = Uniform and X Change
- 2 = Uniform and Y Change
- Aspect Start – Specifies the minimum scale change, at which to begin searching for
matches, as a percentage of the trained pattern (1 to 10,000; default = 100).
- Aspect End – Specifies the maximum scale change, at which to stop searching for
matches, as a percentage of the trained pattern (1 to 10,000; default = 100).

Section 3 | Slide 33 Section 3 | Slide 34

 FindPatterns & FindPatMaxPatterns Results FindPatterns & FindPatMaxPatterns

A B

Structure
GetRow($A$2,B2)

GetCol($A$2,B2)
Cross GetAngle($A$2,B2)
Tool Structure Auto-inserted Functions
GetScale($A$2,B2)
GetScore($A$2,B2)

Section 3 | Slide 35 Section 3 | Slide 36

Results: This slide shows the formula that is entered into each cell when the tool is added to the spreadsheet.

- Tool Structure – Holds all the result information returned by the tool.
- Cross – Indicates the center of the model found.
- Auto-inserted Functions – Are the most commonly needed information, they pull the data
out of the Structure.

Section 3 | Slide 35 Section 3 | Slide 36

 What is PatMax RedLine?

Using PatMax RedLine to Find a

Model

A complete reinvention of pattern matching!

Section 3 | Slide 37 Section 3 | Slide 38

In this section we will find a model using the PatMax RedLine Tool. PatMax Redline has been designed from the ground up to be optimized for speed on the newer In-
Sight 5 megapixel models. It is not based on PatMax. PatMax Redline is typically 3 – 7X faster than
PatMax, and is sometimes even faster. It allows for both speed in pattern matching and high
resolution (accuracy).

Section 3 | Slide 37 Section 3 | Slide 38

 Requirements for PatMax RedLine How Fast is PatMax RedLine?

PatMax RedLine 150 ms (3.2x faster) PatMax RedLine 111 ms (4.9x faster) PatMax RedLine 31 ms (6.8x faster)
Newer In-Sight models, such as:

PatMax = 486 ms PatMax = 551 ms PatMax = 210 ms

In-Sight 76xx, 78xx, etc.

PatMax RedLine 208 ms (16x faster) PatMax RedLine 54 ms (4.5x faster) PatMax RedLine 810 ms (10.8x faster)

In-Sight 5705 PatMax = 3330 ms PatMax = 242 ms PatMax = 8795 ms*

PatMax Redline is generally 4-7 times faster than PatMax

In-Sight Micro 8405
Section 3 | Slide 39 Section 3 | Slide 40

Requirements for PatMax RedLine: PatMax Redline is generally 4-7 times faster than PatMax, but may be even faster. The speeds
- Newer In-Sight models , such as 57xx, 78xx, and 8xxx series depend on the In-Sight camera model, the nature of the part itself, and the settings in the Property
- Spreadsheet Mode: In-Sight Explorer and firmware version 5.1.0 or higher Sheet, including Model Region, Find Region, and Angle of Rotation.
- EasyBuilder Mode: In-Sight Explorer and firmware version 5.2.0 or higher
NOTE: These timings were obtained on an In-Sight 5705.
PatMax RedLine is not supported by the following:
- In-Sight 51xx, 54xx, 56xx models
- In-Sight 74xx models
- In-Sight Micro 1xxx models

Section 3 | Slide 39 Section 3 | Slide 40

 How Fast is PatMax RedLine? PatMax RedLine consists of two functions

Demo TrainPatMaxRedLine

FindPatMaxRedLine

Section 3 | Slide 41 Section 3 | Slide 42

The demo In-Sight job (PatMaxRedLine_demo) allows you to compare timings between PatMax and PatMax RedLine consists of two functions: TrainPatMaxRedLine and FindPatMaxRedLine.
PatMax RedLine, using an In-Sight camera that has both tools. (The comparison of timings is not NOTE: These functions are only available on In-Sight vision systems running In-Sight firmware 5.1.0
useful on the Emulator, since relative behavior between the two tools might not represent and later.
comparisons on actual cameras.)
- TrainPatMaxRedLine – Extracts and trains a pattern from an image for use with the
To run the demo: FindPatMaxRedLine function.
1. Use a camera that has PatMax RedLine. Position a part under the camera and adjust NOTE: A trained pattern consumes approximately 1Mb when using the default pattern Region
exposure and focus to get a good image. Open the demo job and go into its Custom View. (320 X 440), although the pattern size varies greatly depending on the size of the trained
2. Set up the Pattern Region. This will be used for both PatMax and PatMax RedLine. pattern.
3. Click on Train Pattern. This will create separate models for PatMax and PatMax RedLine,
since they use different algorithms for their models. - FindPatMaxRedLine – Finds objects in an image based on a trained pattern.
4. Set up the Find Region. This will be used for both PatMax and PatMax RedLine. FindPatMaxRedLine finds patterns within an image based on a trained pattern from a Patterns
5. Run the PatMax portion of the spreadsheet by clicking on the PATMAX logo in the Custom structure.
View, then click on the Acquire button. This will display the job time for PatMax.
6. Click on the PATMAX REDLINE logo in the Custom View, then click on the Acquire
button. This will display the job time for PatMax RedLine. Compare with PatMax.
7. Try changing other parameters in the Custom View and repeating steps 5 and 6 to
compare times. For example, increasing the Angle should increase the ratio of speed
between the two tools.
8. Note that if you change a model’s parameter for PATMAX REDLINE, its first timing after
that will be much higher due to an optimization that is done at that point.

Section 3 | Slide 41 Section 3 | Slide 42

 Moving from PatMax to PatMax RedLine Key Differences Between PatMax & PatMax RedLine

• Setup and runtime differences between PatMax and

PatMax RedLine include:

- Accept Thresholds and scoring methods are different

• Existing job files using PatMax can be converted to - Contrast Threshold has different meaning
PatMax RedLine - Automatic search optimization behaves differently
- Angle and Scale Find Tolerance behavior is different

• However, there are some differences in the Property

Sheet parameters and how they work

Section 3 | Slide 43 Section 3 | Slide 44

Moving from PatMax to PatMax RedLine : PatMax RedLine matches tend to score higher than PatMax. Therefore, it may be necessary to use
a higher Accept Threshold for PatMax RedLine compared to PatMax.
- Existing job files using PatMax can be converted to PatMax RedLine
- Load job on supported 5 MP system (5705, 5705C or Micro 8405) PatMax RedLine contrast is a relative measure of contrast change between features in the trained
- Replace TrainPatMaxPattern function with TrainPatMaxRedLine pattern versus target features in the search image. PatMax contrast is an absolute measure of
- Retrain the pattern using PatMax RedLine (PatMax RedLine can not use PatMax contrast for target features in the search image.
patterns)
- Replace FindPatMaxPatterns with FindPatMaxRedLine In PatMax RedLine, an automatic (re)optimization occurs following any change to Angle/Scale Find
Tolerances or search ROI. This includes optimization at every trigger when PatMax RedLine is
- However, there are some differences in the Property Sheet parameters and how they work. fixtured. This can make the tool run slow if it is fixtured. However, PatMax RedLine is so much
- Some parameters are not present in both tools faster than PatMax for a 5MP image that a fixture is no longer needed in most cases.
- Some parameters are in both tools, but behave differently
PatMax RedLine Find Tolerances are strict, meaning the tool will fail if a limit is exceeded. PatMax
Find Tolerances are permissive, meaning there is some allowance beyond the limits. Therefore, if
PatMax RedLine does not find matches that PatMax finds, it may be necessary to widen the Angle
and Scale Find Tolerances in PatMax RedLine.

Section 3 | Slide 43 Section 3 | Slide 44

 PatMax RedLine is NOT PatMax! Comparison of Pattern Matching Tools

PatMax
FIndPatterns PatMax
RedLine

Applications where PatMax may still be needed: ~4 times more

~4 times more accurate
Location Accuracy accurate accurate than
than FindPatterns
FindPatterns
- Fixtured search region (tool can run slow in 5.4 and earlier)
4x - 7x faster than
- Variable aspect ratio, perspective, or skew Relative Speed fastest moderately fast
PatMax
- Non-linear deviations along perimeter
Excellent for Excellent for challenging
- Filtering match results at approximately same X/Y location by Excellent for many
Versatility challenging apps requiring both
applications
Angle and Scale Overlap applications speed & accuracy

Scale Range 90 - 110% 1 - 10,000% 1 - 10,000%

~$750
Additional Price included ~$500
(includes PatMax)

included on all option on all

Which Models? option on newer models
standard models standard models

Section 3 | Slide 45 Section 3 | Slide 46

Differences: Take a moment to review the differences between the Pattern Matching tools that are available in the
various In-Sight vision systems. Comparisons of speed are general, and may not apply to every
- Fixtured search region – In versions 5.4 and earlier, PatMax RedLine’s automatic search situation. The best way to compare speeds is to measure them for your setup.
optimization may cause it to run slower than PatMax if the search region is fixtured,
especially when searching for multiple targets on a confusing image background.
- Variable aspect ratio, perspective or skew – PatMax RedLine does not yet support X/Y
aspect ratio changes, perspective changes, or skew in target compared to trained pattern
- Non-linear deviations along perimeter – PatMax RedLine does not yet have an Elasticity
parameter, and may be less tolerant of deviations along the target perimeter compared to
PatMax
- Filtering match results at approximately the same X/Y location by Angle and Scale
Overlap – PatMax RedLine cannot filter results based on angle and/or scale differences
between two or more matches found at the same X/Y location (as allowed by XY Overlap
setting).

Section 3 | Slide 45 Section 3 | Slide 46

 Mathematical Functions

Logic Tools

Section 3 | Slide 47 Section 3 | Slide 48

In this section we will review many of the Mathematical Logic Tools. In-Sight has an extensive set of Vision, Mathematical, and other types of functions. The ultimate goal
is to make a decision about an object being inspected using the information returned by the vision
tools.
- Logic
- Lookup
- Math
- Statistics
- Trigonometry

In this section we will focus on Logic functions.

Logic functions provide Boolean algebra and conditional testing capabilities in the In-Sight
spreadsheet. All bitwise functions perform the corresponding logical operation on each bit position in
the binary representation of the arguments and return the resulting binary number in decimal form.

For all Logic functions:

- False = 0
- True = any non-zero result

Section 3 | Slide 47 Section 3 | Slide 48

 Logic: If Logic: And

If (condition, true_value, false_value)

And (condition0, condition1, condition2, …)
• If condition is True, cell gets true_value
• If condition is False, cell gets false_value • Returns TRUE (1) if all conditions are TRUE
• Otherwise returns FALSE (0)
Example:
A1 = 200
Example:
A2 = If (A1<128, 1, 0)
A1 = 23, B1 = 10, C1 = 22
Ø A2 contains 0
D1 = If (And(A1>B1, A1<C1), 1, -1)
Ø D1 contains -1
A2 = If (A1<128, “Good”, “Bad”)
Ø A2 contains string of characters “Bad”

Section 3 | Slide 49 Section 3 | Slide 50

When functions evaluate numbers: Example:

- False = 0 A1 = 23, B1 = 10, C1 = 22
- True = any other number (for example -5, -345, 1, 34 are all true) D1 = If (And(A1>B1, A1<C1), 1, -1)
- D1 contains -1
Example:
A1 = 200
A2= If (A1<128, 1, 0)
A2 = If (A1<128, “Good”, “Bad”)
- A2 contains string of characters “Bad”

Section 3 | Slide 49 Section 3 | Slide 50

 Logic: InRange Logic: Not

InRange (value, min_value, max_value) Not(condition)

• Returns TRUE (1) if value is greater than or equal to
min_value AND less than or equal to max_value, • Returns opposite of condition
• Otherwise returns FALSE (0)

Example: Example:
A1 = 23, B1 = 10, C1 = 22 B7 = 92
F1 = InRange (C1, B1, A1) D1 = If(Not(B7>95), 1, 0)
Ø F1 contains 1 Ø D1 contains 1

Section 3 | Slide 51 Section 3 | Slide 52

Example: Example:
A1 = 23, B1 = 10, C1 = 22 B7 = 92
F1 = InRange (C1, B1, A1) D1 = If(Not(B7>95), 1, 0)
- F1 contains 1 - D1 contains 1

Section 3 | Slide 51 Section 3 | Slide 52

 Logic: You Try It How to Enter Functions

A2 = 6
B2 = 244.5
What is in these cells?

A6 = If(A2>7, “High”, “Low”)

F2 = If(Not(B2>128), “Pass”, “Fail”)

D1 = If(And(A2>5, A2<B2), -1, 1)

F1 = Not(InRange(A2, 0, B2))

Section 3 | Slide 53 Section 3 | Slide 54

What is in these cells? To enter functions, follow these steps:

1. Find the function in the Palette pane.
A6 = If(A2>7, “High”, “Low”) 2. Drag function to desired location in the spreadsheet.
F2 = If(Not(B2>128), “Pass”, “Fail”) 3. Make references to the desired cells to be evaluated by the function.
D1 = If(And(A2>5, A2<B2), -1, 1) 4. Finally, click the green check to save the changes made.
F1 = Not(InRange(A2, 0, B2))

Answers: “Low”, “Fail”, -1, 0

Section 3 | Slide 53 Section 3 | Slide 54
 Summary Lab Exercise

• Property Sheets make it easy to specify parameters for

a function.
• Auto-inserted functions are the most commonly needed
results from a tool.
• FindPatterns is a powerful tool for locating features on
every part in 2 phases: Training and Searching.
• PatMax can be used to locate trained models when
appearance of the models is adversely affected.
• PatMax RedLine provides both high speed and high
resolution.
• In-Sight has an extensive set of Math functions used to
help make decisions about an inspection.

Section 3 | Slide 55 Section 3 | Slide 56

In this section we covered the following topics: Complete:

Lab Exercise 3.1 – Logic
- Property Sheets make it easy to specify parameters for a function. Lab Exercise 3.2 – FindPatterns
- Auto-inserted functions are the most commonly needed functions for getting information out Lab Exercise 3.3 – PatMax (if time allows)
of Structure.
- FindPatterns is a powerful tool for locating features in every part in 2 phases: Training and
Searching.
- PatMax can be used to locate trained models when appearance of the models is adversely
affected.
- PatMax RedLine provides both high speed and high resolution.
- In-Sight has an extensive set of Math functions used to help make decisions about the
inspection.

Section 3 | Slide 55 Section 3 | Slide 56