JPH0668257A - Detecting method for edge of object in image and detecting method for distance between edges - Google Patents

Abstract

PURPOSE:To enable the sure detection of an edge by performing a cumulative addition of the density value of a picture element around an edge in a direction being orthogonally crossed with the edge, determining a curve having a cumulative value as a value along a coordinate axis and defining the feature point of this curve as an edge location. CONSTITUTION:In the image of a printed board 12 obtained by a camera 20, when the edge 24A of a land 24 is detected within a cumulative addition area, the location of an edge is hidden behind noise when the density value of each picture element in each picture element column and the location of the picture element column are expressed by a vertical axis and a lateral axis, respectively if the edge is noticed for every picture element column parallel to an X-axis. In an image processor 22, noise cancels with each other because noise generates at random as density differential value is integrated vertically to an X-axis, but the edge 24A of the land 24 exists on the same location on the X-axis and is emphasized, and a curve can be obtained. By determining the feature points of this curve such as a minimum value, a maximum changed point, etc., as an edge, the detection of the edge becomes possible.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of processing an image obtained by an image sensor or the like to detect edges of an object in the image and a distance between the edges.

[0002]

2. Description of the Related Art As a method of detecting an edge based on an obtained digital image in the process of image processing, a method of searching for a position where a characteristic such as density of the image changes rapidly is common. Hideyuki Tamura, "Introduction to Computer Image Processing," March 10, 1960, published by Soken Shuppan, Chapter 5, 1.
Some of them, as described on pages 19 to 122, detect an edge by using the sum of densities in a neighboring pixel region or the difference of average values.

Further, as disclosed in, for example, Japanese Unexamined Patent Publication No. 62-202290, the position where the change in brightness is maximum is determined as an edge by weighted averaging weighted by the difference in image brightness. There is something.

There is also a method of detecting an edge in a target image by using relative position information. For example, JP-A-4-
As disclosed in Japanese Patent No. 47467, the inevitable position of the tip of the intermediate lead is calculated using the tip position coordinates of both ends of the lead in the integrated circuit, and only the vicinity of the inevitable area is searched to determine the tip position of the intermediate lead. There is something to calculate. Further, the one disclosed in Japanese Patent Laid-Open No. 2-116908 uses the relative position information to narrow the search range in order to increase the processing speed.

As a method for detecting the distance between target points in an image
Are disclosed in, for example, JP-A-60-263804.
As shown in the figure, measure the part of known length and obtain the correction coefficient.
Therefore, there is a method to calibrate the area between target points.
It This is a mark of known length, as shown in Figure 7.
Actual length of M L_sAnd measured by image processing _s
The correction coefficient α by α = L_s/ L_sAsking for, eyes
The length l obtained by image processing of the length L between the reference points AB
And α are used to obtain by the formula L = α × l
It In FIG. 7, reference numeral 1 is an object, 2 is an image thereof, and 3 is a camera.
Show.

[0006]

The edge detection method described in the "Introduction to Computer Image Processing" can be applied when the density change due to the edge is sharper than the density change due to the noise in the small area. In the case of the same level, there is a problem that it is buried in noise and cannot be applied.

Further, the method disclosed in Japanese Patent Laid-Open No. 62-202290 does not disclose a method of setting the range when the weighted average is taken. Therefore, when the setting is erroneous, the wrong edge position is detected. There is a problem that it does.

Further, the above-mentioned Japanese Patent Laid-Open Nos. 4-47467 and 2
The one disclosed in Japanese Patent Laid-Open No. 116908 can be applied when other edges having similar features do not exist in the inevitable existence region and the distinction between the edges is clear, but they have similar features. If there are a plurality of edges and it is difficult to distinguish each edge, there is a problem in that it cannot be applied because the method of setting the existence-necessary area is not taken into consideration.

Furthermore, in the method disclosed in the above-mentioned Japanese Patent Laid-Open No. 60-263804, since the correction coefficient obtained from the known mark M is directly applied to the target point, the correction coefficient α
This is applicable only when the distance from the mark used to obtain the distance to the camera (image sensor) is equal to the distance from the target point to the camera, and there is a problem in that it is very difficult to set the mark.

The first aspect of the present invention has been made in view of the above-mentioned problems of the prior art. In an image, an edge can be detected even when the density change due to the edge is poor and it is difficult to distinguish it from noise. An object of the present invention is to provide an edge detection method for an object inside.

The second aspect of the present invention is an object in an image for which a target edge can be surely detected even if there are a plurality of edges having similar characteristics in the area where the edges are necessarily present in the image. It is an object of the present invention to provide an edge detection method of

Further, according to the third aspect of the present invention, when the distance between edges is detected, even if the distance from the reference object to the camera and the distance from the target edge portion to the camera are different, the image can easily detect the edge distance. An object is to provide a method for detecting a distance between edges of an inside object.

[0013]

According to a first aspect of the present invention, the density values of pixels around an edge of an object having a linear edge are cumulatively added in a direction orthogonal to the edge and are orthogonal to the edge in the image. The above object is achieved by a method for detecting an edge of an object in an image, which is characterized in that a curve having the cumulative value as a value is obtained along a coordinate axis of a direction, and a characteristic point of the curve is obtained as an edge position.

It is also possible to obtain an approximate curve passing through the three points of the accumulated value near the characteristic point and set the apex of this curve as the edge position.

According to a second aspect of the invention, the process of detecting the position of a feature point in the image of an object and the relative position information of the edge whose relative position from this feature point is known in advance are used to determine the position of the search range of the edge and Using the process of setting the size so that no edges other than the edge are included, the process of detecting the edge by processing the image data in the search range, and the relative position information from the obtained edge position. And a step of determining the position and size of a search range of another edge, and a step of detecting the other edge by processing image data in this search range. The object of the present invention is achieved by a method of detecting an edge of an object in an image, in which the process of determining the position and size of the above is repeated, and the target edge is finally detected.

According to a third aspect of the present invention, an object having a constant size in the image is used as a reference object, and the reference object size in the image, the target edge distance in pixel units, and the actual size are defined. Create a table in which the conversion factor indicating the ratio is calculated in advance for the sizes of various reference objects in the image, and when measuring the target edge distance, refer to the target edge distance in pixel units. The size of the object is measured, a conversion coefficient is calculated from the size of the reference table and the reference object in pixel units, and the target edge is calculated by multiplying the conversion coefficient by the target edge distance in pixel units. The object is achieved by a method for detecting a distance between edges of an object in an image, which is characterized in that an actual size of the space is obtained.

[0017]

According to the first aspect of the present invention, noise is randomly generated and cancels each other even in the case of an edge in which the density change due to the edge of the image is poor and it is difficult to distinguish from the noise. On the other hand, since the edge is a straight line and always exists at the same position and is emphasized, the value obtained by cumulatively adding the pixel density values around the edge in the direction perpendicular to the edge has a high S / N ratio. , The edge can be reliably detected from the curve based on this value.

According to the second aspect, when the curve is obtained, the apex of the approximate curve passing through the three accumulated values in the vicinity of the characteristic points is used as the edge, so that the accurate edge position can be obtained.

According to the second aspect of the present invention, since the position and size of the search range where other edges do not enter are finally determined, the image data within this search range is processed. Even if there are a plurality of edges having similar characteristics, the target edge can be reliably detected.

According to the third aspect of the present invention, since it is not necessary to set the distance from the reference object to the camera (image sensor) and the distance from the target edge portion to the camera to the same value, the target edge is attached. It is possible to use a part to be edited or a part of the background as a reference object, and to easily set the reference object.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The method according to the present invention will be described in detail below with reference to the drawings.

First, the overall system configuration of the printed circuit board mounting line 10 shown in FIG. 1 which utilizes the method for detecting an edge of an object in an image according to the present invention will be described.

The printed circuit board mounting line 10 automatically measures the position of a pattern on the printed circuit board 12 by image processing, and automatically mounts electronic components on the printed circuit board 12 based on this data.

Printed circuit board 12 which is the object of position measurement
Is sent to the front position of the mounting machine 16 by a carrier (belt conveyor) 14, and the printed board 12 roughly positioned by the positioning mechanism 18 is photographed by a camera 20 from above, and the obtained image is an image. The image is sent to the processing device 22, and the image processing device 22 obtains the position of the land 24 on the printed circuit board 12 by image processing. The mounting head 26 places the land 24 on the mounting machine 16 at the obtained position. The electronic component 17 is automatically mounted.

Here, the positioning mechanism 18 has a "play".
Therefore, the positioning cannot be performed with sufficient accuracy for mounting the electronic component 17 on the printed circuit board 12. Further, there is a positional error between the outer shape of the printed circuit board 12 and the land 24 on which the electronic component 17 is mounted due to a manufacturing error. Therefore, even if the printed circuit board 12 can be accurately positioned, the electronic component 17 is placed on the land 24. Can not be mounted in the correct position.

Therefore, it is necessary to directly determine the position of the land 24, more precisely, the position of the edge thereof, by processing the image photographed by the camera 20 by the image processing device 22 as described above.

Next, the process of detecting the position of the land 24 in the feed direction by the carrier 14 will be described. In addition,
Positioning in the direction orthogonal to this is also performed by the same method.

Although the edge 24A of the land 24 is linear, when the density of the upper surface of the printed circuit board 12 which is the background is close to the density of the land 24, it cannot be clearly seen in the image, and therefore, in the feeding direction of the carrier 14. The edge cannot be obtained from the change in the density of the image.

However, since the edge 24A of the land 24 is in the same direction as the outer edge of the printed circuit board 12 and the edge 18A of the positioning mechanism 18, in the camera input, the horizontal direction on the screen (in FIG. It was set so as to be orthogonal to the X axis which is the feeding direction). The image processing device 22 uses the edge detection method by cumulative addition to detect the land 24.

For example, when the edge 24A of the land 24 is detected in the cumulative addition area 28 indicated by the broken line in the image 13 of the printed circuit board 12 obtained by the camera 20 shown in FIG. 2, one pixel parallel to the X axis is used. Focusing on each column, when the density value of each pixel in each pixel column is represented on the vertical axis and the position (screen coordinates) of the pixel column is represented on the horizontal axis, the edge position indicated by the broken line in FIG. It cannot be detected because it is hidden in.

In the image processing apparatus 22 of this embodiment, since the density differential values are integrated perpendicularly to the X-axis, noises are randomly generated, and therefore these cancel each other out. Since the edge 24A is always present at the same position on the X-axis, it is emphasized and a curve as shown in FIG. 3B can be obtained. The characteristic points of this obtained curve,
For example, the edge can be detected by obtaining the minimum value, the maximum change point, and the like as the edge. In the illustrated embodiment, the maximum value is detected as an edge.

Here, as shown in FIG. 3C, the maximum change point of the curve of the cumulative addition value of raw data may be obtained as an edge.

By the way, in order to properly mount electronic parts on the land 24, the position of the land 24 should be ± 0.1 mm.
It is necessary to calculate with the accuracy of
Since it corresponds to mm, to measure in 0.1 mm units, 1
It is necessary to obtain the position of the edge 24A with an accuracy of less than pixels.

As described above, the image processing apparatus 22 handles a curve whose value is the cumulative addition value of the density change of the pixel with respect to the coordinate in the direction orthogonal to the edge 24A.

The actual curve by the obtained cumulative addition value is
In contrast to the bent line shape shown by the symbol B in FIG. 4, an ideal curve when the resolution in the image is infinite is smooth as shown by the symbol A in FIG.
It is expected to be similar to a quadratic curve.

The minimum point X of the smooth curve indicated by the symbol A
The deviation between a and the minimum point Xb of the bending curve B becomes a sampling error. In the image processing apparatus 22 of the embodiment, the three points B1, B2, B3 near the minimum point Xb of the bending curve B are approximated by a quadratic equation, and the vertex of the curve passing through these three points is determined as the position of the edge 24A. Has been In this case, bending curve B
The local shape near the minimum point Xb of is sufficient by the polynomial, and the approximation calculation time is short, and the vertices can be obtained with certainty.

Although it is ideal if the equation expressing the curve A is obtained, the curve A is formed based on the combination of the illumination light of the printed board 12 and the complicated reflected light from the peripheral equipment. However, it is very difficult to estimate the equation representing the curve A.

Next, when an edge having a similar structure to the edge 24A of the land 24 is present,
A case of detecting a target edge will be described.

Next to the edge 24 of the land 24, as shown in FIG.
As shown in FIG. 5, the edge 25A of the adjacent land 25 exists, and since both the edges 24A and 25A are boundaries between the resin and the metal pattern of the print substrate 12, the density change from the periphery on the image. Are exactly the same,
It is impossible to distinguish between the edges 24A and 25A from the change in density.

Here, the distance between the edges 24A and 25A is
When it is smaller than the sum of the positioning error of the printed circuit board 12 by the positioning mechanism 18 and the error between the outer shape of the printed circuit board 12 and the land 24 due to the manufacturing error of the printed circuit board 12, the discrimination becomes more difficult. For example, when the rattling of the positioning mechanism 18 is ± 4 mm and the positional deviation between the outer shape of the printed circuit board 12 and the pattern due to a manufacturing error of the printed circuit board 12 is ± 3 mm, the land 24 in the screen is changed.
The variation in the position of the edge 24A of the
Since the distance between 5A is larger than 5 mm, it is impossible to set an appropriate search range as described above.

In this embodiment, the image processing apparatus 22 sets a search range such that the edge 25A of the adjacent land 25 does not enter when detecting the edge 24A of the land 24.

First, the edge 18A of the mounting portion of the positioning mechanism 18 at the position fixed to the camera 20 on the carrier 14 is noticed and used.

The edge 18A of the positioning mechanism 18 is
Since the relative position with respect to the camera 20 (X and Y directions) is fixed and always appears at almost the same position in the screen of the image processing device 22 and there is no same or similar edge in the periphery, the edge 18A is searched. The range to be set can be set wide, and it will not be confused with other edges.

The image processing device 22 uses the relative position information from the edge 18A of the positioning mechanism 18 to set the search range of the edge 12A (parallel to the edge 18A) of the outer shape of the printed circuit board 12. Here, the positioning mechanism 1
Since the error of 8 is ± 4 mm of the positioning mechanism 18,
Based on this, the search range of the edge 12A is set to have a width of 8 mm. No other edges such as patterns exist in this range.

In the image processing device 22, the printed circuit board 12
The search range of the edge 24A of the land 24 is set using the relative position information from the outer edge 12A. Since the variation in the position of the land 24 with respect to the edge 12A is the same as the manufacturing error of the printed circuit board ± 3 mm, the search range of the edge 24A is set to be 6 mm wide. Edge 24A
The distance between and 25A is 5 mm, which is within ± 3 mm, so that the edge 25A does not appear in the width of this search range.

As described above, starting from the detection of the edge 18A which can be surely detected even in consideration of the variation range in the screen, the relative position information from the detected edge 18A is used to detect other edges 12A. Set the search range, and
Based on this, the edge detection search range of the edge 24A of the target land 24 is appropriately set and detected.

Next, the actual size distance between the edge 24A of the land 24 and the edge 12A of the printed board 12 obtained by the image processing as described above, and further the edges 24A and 25.
The process of obtaining the actual size distance between A will be described.

The unit of the coordinates obtained for the edge 24A of the land 24 is a pixel, and the actual size can be obtained only when this is converted into millimeters. First, the printed circuit board 12 is fixed off-line by the positioning mechanism 18, and the actual size of the distance between the edge 18A of the mounting portion of the positioning mechanism 18 and the edge 24A of the land 24 is accurately measured.

Next, the camera 20 from the printed circuit board 12
The distance, that is, the camera distance is appropriately changed, and this distance and the diameter D of the capacitor 30 on the printed circuit board 12 are measured in pixel units (the number of pixels) by image processing. from this result,
A look-up table as shown in FIG. 6 is created in advance, which shows the length D in pixel units and the ratio P in pixel units and the actual size of the distance between the edge 18A of the positioning mechanism 18 attachment portion and the edge 24A of the land 24. I'll do it.

Here, the value of D in FIG. 6 changes depending on the distance between the camera 20 and the printed circuit board 12, that is, the camera distance. The shorter the camera distance, the larger the value of D. Further, measurement positions 1 to 3 in FIG. 6 indicate three measurement positions on the printed circuit board 12 which are different in distance from each other when the distance from the camera 20 to the printed circuit board 12 is the same.

By the image processing, the length in D pixel units and the distance in pixel units between the edges 24A and 18A at some or all of the measurement positions 1 to 3 are obtained for each image, and the lookup in FIG. 6 is performed. Calculate the value of the conversion coefficient p referring to the table,
This p is multiplied by the distance between the edges 24A and 18A in pixel units to obtain the actual distance. Here, the electronic component is the land 2
Since the distance between 4 and the camera 20 is different, it is possible to obtain a conversion coefficient each time an image is input, and to accurately measure the actual land position.

Since the position of the edge 18A of the positioning mechanism 18 is immovable, the absolute position of the edge 24A can be obtained from this position, and based on this information, the mounting head 26 can correct the electronic component at the correct position on the land 24. It will be installed. Conventionally, a calibration mark was previously attached on the printed circuit board, and the conversion coefficient p was calculated for each image with reference to this, but in this embodiment, the calibration mark is unnecessary, The size on the board does not change at all times, and the conversion coefficient is obtained by using the capacitor 30 on the already mounted electronic parts, which is the part that can be surely confirmed in the image, and the mounting density is further improved. Can be made.

Although the present invention is applied to the mounting line of the printed circuit board in the above-mentioned embodiment, the present invention is not limited to this, and the components are automatically mounted and the existence is confirmed. And is generally applied when detecting an edge of an object in an image processing process for detecting wear of a component.

[Brief description of drawings]

FIG. 1 is a perspective view showing a mounting line of a printed circuit board to which the present invention is applied.

FIG. 2 is a plan view showing an image of a printed circuit board in the image processing apparatus on the mounting line.

FIG. 3 is a diagram showing the relationship between the screen coordinates near the edge on the printed circuit board, the pixel density value, and the cumulative addition value.

FIG. 4 is a diagram for obtaining an edge of a land by a cumulative addition value using curve approximation by a quadratic equation.

FIG. 5 is a plan view showing an image around an edge including a fixed length portion that serves as a reference of a correction coefficient when detecting an edge using relative position information and when detecting an inter-edge distance by a calibration method.

FIG. 6 is a diagram showing a lookup table for calibration.

FIG. 7 is a schematic perspective view showing a conventional calibration method.

[Explanation of symbols]

 10 ... Printed circuit board mounting line 12 ... Printed circuit board 12A, 18A, 24A, 25A ... Edge 14 ... Conveyor (conveyor) 16 ... Mounting machine 18 ... Positioning mechanism 20 ... Camera 22 ... Image processing device 24, 25 ... Land 26 ... Mounting head 28 ... Cumulative addition area 30 ... Capacitor

Claims (4)

[Claims] 1. A density value of pixels around an edge in an image of an object having a linear edge is cumulatively added in a direction orthogonal to the edge, and the cumulative value is obtained along a coordinate axis in a direction orthogonal to the edge in the image. A method for detecting an edge of an object in an image, characterized in that a curve having a value of is obtained, and a characteristic point of this curve is obtained as an edge position. 2. An edge detection method for an object in an image according to claim 1, wherein an approximate curve passing through three points of said cumulative value in the vicinity of said characteristic point is obtained, and a vertex of this curve is set as an edge position. . 3. The position and size of the search range of the edge are calculated from the process of detecting the position of the feature point in the image of the object and the relative position information of the edge whose relative position from this feature point is known in advance. , A step of setting an edge other than the edge so as not to enter, a step of detecting the edge by processing the image data in the search range, and a relative position information from the obtained edge position, A step of determining the position and size of the edge search range, and a step of detecting the other edge by processing the image data within the search range,
An edge detection method for an object in an image, wherein the steps of determining the position and size of the edge search range and detecting the edge are repeated to finally detect the target edge. 4. An object having a constant size in an image is used as a reference object, and a size of the reference object in a pixel unit and a ratio of a target edge distance between the pixel unit and an actual size are shown. When a table is created in which the conversion factors are calculated in advance for the sizes of various reference objects in the image and the target edge distance is measured, the target edge distance and the reference object size are calculated in pixel units. The conversion factor is obtained from the table created in advance and the size of the reference object in pixel units, and the conversion factor is multiplied by the target edge distance in pixel units to calculate the actual distance between the target edges. A method for detecting a distance between edges of an object in an image, which is characterized by obtaining a dimension.