创建组件识别数据的方法.pdf - 第173页
173/178 IM-SMT-VSNDI C-0002 -000 Append ix A ppendix 1. Noise processing 1) “Cut Inner N oise” Noise is elim inated b y performing com pression proces sing after expansion proce ssing has been carrie d out (See Figure 14…
172/178
IM-SMT-VSNDIC-0002-000
those components that are thought to be rectangles or triangles are detected and
recognized. This approach is different from contour detection on the point that all of the edge
positions in the detection range are targeted for detection.
Therefore, this recognition method is effective for the objects which are difficult to be detected
when performing the center of gravity detection and the contour detection due to the effects
such as uneven reflection.
There is a possibility that recognition will still be possible even in situations when a large
portion of the pattern is missing. In terms of accuracy, errors are more likely to occur.
Figure 143
<< Items to be set for Contour rectangle detection / contour triangle detection >>
Shape Type Selects from (Square/Oblong) / Triangle
Surface type Selects from Reflect / Non reflect
Algorithm Type Special 2
Mark Threshold Sets the threshold for edge detection.
Tolerance
This is used when checking the surface area and the
perimeter (of the outer edge of the line detection results).
Search Area Sets the detection range.
Cut Inner Noise Selects from 0 through 9.
Cut Outer Noise Selects from 0 through 9.
Mark Outsize X (mm) Sets the Body Size of the mark (side length
).
Mark Outsize Y (mm) Sets the Body Size of the mark (side length
).
Table 121
173/178
IM-SMT-VSNDIC-0002-000
Appendix
Appendix 1. Noise processing
1) “Cut Inner Noise”
Noise is eliminated by performing compression processing after expansion processing
has been carried out (See
Figure 144
).
2) “Cut Outer Noise”
Noise is eliminated by executing expansion processing after compression processing
has been carried out (See
Figure 144
).
Noises of inside and outside of the component can be eliminated by using both “Cut
Inner Noise” and “Cut Outer Noise“ functions.
The processing results vary depending on the order in which the noise elimination
functions are executed. The execution sequence of the functions should be selected
according to need.
Figure 144
<Expansion processing>
The bottom figure in the middle in
Figure 144
shows the result after expanding the image on
the left by 1 pixel. As shown in this example, if the gray area is expanded by 1 pixel, a section
consists of two or less pixels in the component is embedded.
Multiple pixels can be specified for expansion, so the section containing the multiple pixels
(noise) is embedded. Performing compression processing following the expansion processing
produces an image without the noise. (See the bottom figure on the right in
Figure 144
).
Dele
ted
Compressed
Expand
Deleted
Deleted
Deleted
Expand
Compressed
Connected
Connected
Noise
elimination
outside the component
Noise
elimination
inside the component
174/178
IM-SMT-VSNDIC-0002-000
<Compression processing>
The top figure in the middle in
Figure 144
shows the result after compressing the image on the
left by 1 pixel. As shown in this example, if the gray area is compressed by 1 pixel, a section
consists of two or less pixels in the component is deleted. As multiple pixels can be
specified for compression, the section containing the multiple pixels (noise) is deleted.
Performing expansion processing following the compression processing produces an image
without the noise (See the top figure on the right in
Figure 144
).
Appendix 2. Detection level differences used to detect leads (threshold value)
This parameter is used when leads are being detected. For lead detection, scanning is
performed within the search range in order to separate the leads from the background.
During the scanning process, if there are any pixels that are brighter than the value specified
as the difference in detection level from the background density, those pixels are recognized
as leads. Consequently, setting a lower value for this parameter makes it possible to detected
dark leads. If the value is too low, however, the results are more easily affected by the noise.
Moreover, setting a larger value is more effective in terms of noise, but there is a greater
possibility of error in detecting the leads themselves.
The recognition conditions should be set in such a way that the illumination produces a
strong contrast between the background and the leads. In the figure below, the illumination
has been set to maximize the difference in levels between the lead section and the
background section. If the lighting is too dark, the brightness of the lead itself decreases,
and there will be no contrast. Conversely, if the lighting is too bright, the brightness of the
background increases, eliminating contrast between the background and the lead.
Figure 145
Brightness
Brightness
Brightness
With dark lighting
With appropriate
lighting
With bright lighting