创建组件识别数据的方法.pdf - 第167页
167/178 IM-SMT-VSNDI C-0002 -000 <<For special shape> Mark O utsize X (mm) Sets the Bod y Size X. Mark O utsize Y (mm) Sets the Bod y Size Y . Area (mm 2 ) Sets the area. Outline Sets the perim eter Table 114 3.…
166/178
IM-SMT-VSNDIC-0002-000
The recognition results are detected at the sub-pixel level.
The following shapes can be selected as the mark shape:
- Circle
- Square/ Oblong
- True triangle
- Special shape
The surface area and shape constant are checked for each of these shapes.
(Shape constant of the “Special shape” and “Oblong” are not checked. Only the surface area
is checked for the shapes.)
This recognition method is suitable for objects with brightness contrast.
<< Items to be set for center of gravity detection >>
Shape Type Selects from circle/square/triangle/special shape.
Surface type Selects reflecting/non-reflecting
Algorithm Type Normal
Mark Threshold Sets the binarization level. (If this is set to 0, the
binarization level is set automatically.)
Tolerance Sets the shape tolerance.
Search Area Sets the detection range.
Cut Inner Noise Selects from 0 through 9.
Cut Outer Noise Selects from 0 through 9.
Table 112
Figure 135
<< For circle / (square /rectangle)/ true triangle >>
Mark OutSize
Sets the Body Size.
(As shown in Figure 135, set the diameter for circle, the length of the
side for square and triangle, and for oblong, set the length of the
bottom line for X, height for Y.)
Table 113
The value obtained from the formula ( The perimeter
2
/ The surface area) is set as a
constant value, like those shown below. This is referred to as “shape constant”.
Circle: Perimeter
2
divided by surface area = 4π
Square: Perimeter
2
divided by surface area = 16
True triangle: Perimeter
2
divided by surface area = 36 / √ 3
[ What is
“
Shape constant
”
? ]
167/178
IM-SMT-VSNDIC-0002-000
<<For special shape>
Mark Outsize X (mm) Sets the Body Size X.
Mark Outsize Y (mm) Sets the Body Size Y.
Area (mm
2
) Sets the area.
Outline Sets the perimeter
Table 114
3.2. Template matching (Recognition type: pattern)
In this method, a mark that has been registered in advance is used to detect the mark that is
the closest to the registered mark from among those on the image that has been captured.
A template is created from the specified field.
Figure 136
Matching is performed on the captured image by using the registered template.
Figure 137
The position that matches most closely to the registered mark is detected.
The recognition results are detected at the sub-pixel level.
In “PTRN Outline” mode, recognition is performed by registering the contour part of the
fiducial mark as a template.
In “PTRN Graylev” mode, the straight line on each vertical and horizontal direction are set in
order to speed up the detection, and the rough position is determined by using the brightness
Template registration
Matching
168/178
IM-SMT-VSNDIC-0002-000
distribution on the straight lines, then the detailed matching is performed by using those
positions as the center.
As speeding up process is not adopted for “PTRN Whole” mode, it takes time for processing
compare to other modes. However, there is the least possibility for the detection error and it
provides the high detection accuracy.
The matching degree is calculated by normalized correlation. Therefore the overall fluctuation
in brightness does not affect the degree.
The matching degree decreases when unevenness of lighting or reflection fluctuates.
Also it does not correspond to the size variation and rotation (as it may decrease the matching
degree), it is not applicable for the objects that change size or rotate.
<< Items to be set for Matching detection >>
Shape Type --
Surface type --
Algorithm Type Selects from “PTRN Outline” ,”PTRN GrayLev” or “PTRN Whole”
Mark Threshold
--
Tolerance
Specifies the allowable matching level. When the matching level
is more than (1.0 – tolerance /100), it is determined to be
satisfactory.
Search Area Sets the detection range.
Cut Inner Noise
--
Cut Outer
Noise
--
Pattern Size X Template size X
Pattern Size Y Template size Y
Offset X Offset amount X from the center of the template
Offset Y Offset amount Y from the center of the template
Table 115
3.3 Corner detection
With this recognition method, the linear elements are detected from the contour of the object
with the largest surface area, and from those elements, those that are thought to be corners
are detected and recognized.
Caution:
In this mode, recognition should be performed only when there is only one corner in the
detection range.
If there are more than one corner in the detection range, it is impossible to determine which
coordinate of the corner to be detected.
Figure 138