
Chapter
4
Reference Manual
IVC-2D
124
©
SICK AG • Advanced Industrial Sensors • www.sick.com • All rights reserved
Matching
Referencing to a Shape Locator Pose
When inspecting several small patterns on a large object, for example keys on a keypad,
use the Shape Locator first and refer to it by the “Shape Locator step” parameter. This
way, small defects can be found, according to the rule of thumb: “Search big, inspect
small”.
In this mode:
•
The Inspect Pattern search rectangle will automatically move with the Shape Loca-
tor pose and resize according to its scale.
•
The position and angle of the found pattern relative to the Shape Locator pose will
be reported as “Delta X”, “Delta Y” and “Delta angle”. These results are useful for
example to determine that a pattern has the correct position on an object, rather
than in the image.
Technical Comparison with Shape Locator
The Pattern tools are based on a normalized cross-correlation (NCC) method, which is very
robust to variable contrast and ambient light conditions. The method is suitable for finding
any type of pattern and it allows inspection of small defects. It performs best at small
rotation tolerances, although full rotational search is supported at the cost of processing
time.
The Shape Locator tools are based on a geometric pattern matching method, which is the
best choice for locating large patterns (objects) in any rotation. This method is more toler-
ant to occlusion, noise and clutter than the normalized cross-correlation.
The following table provides a selection guide by describing several key aspects of the
different algorithms:
Pattern Shape
Locator
Information used in search
Intensity in all pixels
Edges
Soft edges tolerance
Very good
Limited
Poor contrast tolerance
Very good
Limited
Ambient light robustness
Very good
Good
Occlusion tolerance
Limited Very
good
Noise and clutter tolerance
Good
Very good
Distortion sensitivity
Locate pattern: Medium
Difference image: High
Medium
Scale tolerance
• 20%
with affect on the
score
• 20% without affecting
the score when referring
to a Shape Locator
• 20%
Processing time
Some variation
Constant
Accuracy* (mean error)
Sub-pixel (~0.05 pix)
Sub-degree (~0.1 deg)
Sub-pixel (~0.1 pix)
Sub-degree (~0.1 deg)
Multiple patterns/objects
No Yes
Large search rectangle
Slow Fast
Full rotation search
Slow Fast
Difference image
Yes No
* Accuracy testing procedure: Simple (Pac-Man) shape of radius 50 pixels as reference pattern. Search rectangle
size 400x400 pixels. Pose measurement on 100 images, where the reference pattern has been synthetically
moved and rotated to known poses. 3x3 smooth filter as pre-processing. Robustness parameter set to robust
(Pattern). Accuracy parameter set to high (Shape Locator). No image distortion or noise added.