JP3765888B2 - Part position detection method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a method for detecting the position of a component, and more specifically, a component for detecting a positional deviation at the time of suction when a component such as an electronic component mounted on a chip mounter is sucked and transported to a mounting substrate. The present invention relates to a position detection method.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, a chip mounter has been provided with a suction head having a suction nozzle that sucks an electronic component (hereinafter simply referred to as a component) such as an IC chip component. Adsorbed, transferred and mounted on the circuit board. Since normal parts are not necessarily picked up in the correct posture, the picking posture of the part is picked up by an image pickup device such as a CCD camera, and image recognition of the part is performed. After the deviation from the center position and the amount of inclination are corrected, the component is mounted on the circuit board.
[0003]
In order to obtain or correct this positional deviation amount, a position detection algorithm corresponding to the type of component has been developed, but there is a problem that component data necessary for each algorithm must be described for each component. In order to solve this complexity, for example, template matching as described in Japanese Patent Publication No. 8-12050 is known as a general-purpose algorithm. Template matching is a method in which a template representing a part is created and moved in an image including the part, and a part is recognized at a template position where a large similarity can be obtained between the image and the template.
[0004]
In the above publication, a first template including part or all of a part (reference mark) and a second template that is point-symmetric with the first template are registered, and an image of the part to be inspected and the first and second templates are registered. Each of the matching is taken, and the center deviation is obtained with the midpoint of the two corresponding points of each template when the matching is taken as the center of the part.
[0005]
The technique described in this publication can detect the center position of a component, but cannot detect the inclination of the component. Further, in the above-mentioned publication, the center position is determined by setting a window including a part or all of the parts in advance, creating a point-symmetric template with this window, and matching it with the registered template. An example for obtaining is also shown. In this example, there is an advantage that the memory for storing the template can be omitted. However, a stable result can be obtained if the part is in a nearly constant position in the image, but if it is not, the part will not be in the window and will not contain valid features. There is a problem that it cannot be obtained.
[0006]
This can be dealt with to some extent by increasing the window size, but the template contains a lot of extra information, which not only reduces the matching accuracy but also increases the computation time. When a template is registered and executed in advance, the above does not occur, but if the part has an inclination and there is a rotational deviation from the template, the matching accuracy will decrease, and the center will still be at the expected accuracy. Can't get position.
[0007]
[Problems to be solved by the invention]
Since the template matching method can be implemented as a general-purpose position detection algorithm for atypical parts that fall within the category of the conventional positional deviation detection algorithm, the present invention uses the template matching method to Thus, an object is to provide a component position detection method capable of detecting the position of a component with a highly accurate position detection algorithm.
[0008]
[Means for Solving the Problems]
The invention described in claim 1
Take an image of a point-symmetric part about its center,
Specify a reference window that contains part or all of the part image using coordinate values ,
A window symmetrical to the reference window with respect to the X axis or Y axis passing through the center of the part is set as a template ,
Matching a template of a window symmetrical to the reference window and an image obtained by rotating the image around its center, and detecting the inclination of the component based on the amount of rotation when the matching is achieved Features and
The invention according to claim 2
Take an image of a point-symmetric part about its center ,
Create a template that contains part or all of the part from the image of the part, register it ,
Set the reference window at the template position where the degree of correlation is maximized by matching the registered template and the image of the part ,
A window symmetrical to the reference window with respect to the X axis or Y axis passing through the center of the part is set as a template ,
Matching a template of a window symmetrical to the reference window and an image obtained by rotating the image around its center, and detecting the inclination of the component based on the amount of rotation when the matching is achieved Features .
[0009]
Further, in the inventions described in claims 3 and 4, in the inventions described in claims 1 and 2, respectively, a second template that is point-symmetric with the first template is created using the reference window as the first template. A midpoint of a straight line connecting a point in the region of the first template when matching between the component image and the second template is obtained and a point in the region of the second template corresponding to this point is obtained. The present invention is characterized in that it is obtained as the center of a part in the inventions of claims 1 and 2 .
[0010]
In such a configuration, the center and inclination of the component can be detected with high accuracy by template matching.
[0012]
Further, in the present invention (claims 2 and 4), a template in which a part or all of a part is entered from the part image is created, registered, and matching between the registered template and the part image is performed. Since the reference window is set at the template position where the degree of correlation is the highest, even if the position of the acquired image of the part is unstable, it is guaranteed that part or all of the part will be surely present in the window Is done.
[0013]
Further, the rotation is performed at a predetermined step angle, the difference or derivative of the correlation value of matching is obtained , and the inclination of the component is detected by obtaining the zero cross point of the difference value or derivative value . As a result, the same effect as increasing the resolution of the step angle can be obtained, and the accuracy of inclination detection can be improved.
[0015]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0016]
[Overall configuration]
FIG. 1 is a block diagram of a system for carrying out a component position detection method according to the present invention. In the following, the object to be processed is assumed to be a part that is point-symmetric with respect to its own center and is sucked by the suction nozzle of the chip mounter.
[0017]
This system includes a component imaging stage 1, a position control device 2 for controlling the posture of the component, a man-machine interface 3 for an operator to control the position control device 2 and the image processing device 12, and a TV camera for imaging the component. 4, a display device 5 for displaying the image, and an image processing device 12 for processing the image data to detect the center and inclination of the component. In addition, the image processing apparatus 12 performs template matching on a template memory 6 in which a stored template is expanded, an image memory 7 that stores an image from the TV camera 4, a non-volatile memory 11 that stores a created template, and template matching. The calculation unit 10 includes a calculation unit 10, an interface 8 with an external circuit, and a control unit 9 that controls the flow and processing of each data.
[0018]
FIG. 2 is a flowchart schematically showing the overall work and the flow of processing. The position detection module is driven based on recognition condition data created in advance so as to be able to flexibly cope with various components and operation conditions (calculation speed priority, accuracy priority, etc.). The operator can select a position detection process suitable for the system by creating a recognition condition module (data group) in consideration of the characteristics of the parts and the operation conditions.
[0019]
Hereinafter, these will be described.
[0020]
First, in step S1, it is determined whether or not to use a reference window setting position function based on template matching. This function is a function for ensuring that a part or all of the parts are surely present in the window even when the position of the parts on the image is unstable.
[0021]
If the position of the component is stable, this function is not used and the process proceeds to step S2 to define a reference window. The reference window can be designated as a known coordinate value, and the window position can be made variable depending on the type of part. The operator sets the window area in the recognition condition module.
[0022]
[Create template]
On the other hand, if the position of the component is unstable, the process proceeds to step S3 to create a template. The template creation routine is illustrated in detail in FIG. First, an image including a normal posture component is input in advance (step T1). This imaging is performed by capturing an image including all of the components set on the component imaging stage 1 of FIG. Then, this image is displayed on the display device 5. Since the part does not always have a normal posture on this image, the operator controls the position control device 2 via the man-machine interface 3 to move the part imaging stage 1 to correct the part posture (step). T2, T3). During this operation, the image is taken into the image memory 7 in real time and displayed on the display device 5 as described above. FIG. 7 schematically shows how the component posture is corrected, and the image of the component 20 that was initially in the position indicated by the dotted line is in a normal posture as indicated by the solid line by the above-described correction. .
[0023]
Next, as shown in FIG. 8, the operator designates a reference point R and an area for creating a template 21 in which part or all of the component 20 enters the image processing apparatus 12 via the man-machine interface 3 ( Step T4). When the template is confirmed (step T5), the control unit 9 of the image processing apparatus receives this designation via the interface 8, acquires the image data from the image memory 7, and registers the template in the nonvolatile memory 11. (Step T6). On the other hand, if the template to be registered is not yet determined, the process returns to step T1 and the above-described processing is repeated.
[0024]
When the template is created and registered in this way, the process returns to step S4 in FIG. 2, and the template is set in the recognition condition module. Subsequently, when using the inclination calculation function based on the relative position of the corner of the component described with reference to FIG. 14, the component outer dimension is set in the recognition condition module (step S <b> 6). This step S6 is skipped, a recognition condition data group is transmitted in step S7, and the position of the component is detected in step S8.
[0025]
This position detection routine is illustrated in detail in FIG. First, as shown in step R1, a template is designated according to the recognition condition data set in step S2 or S4, or a window position is designated. When the component position on the image is stable, the window position is designated, and when it is unstable, the reference window setting position is acquired in step R2. This is done by the following process.
[0026]
[Reference window]
As shown in FIG. 9, the image of the component 20 to be inspected is taken into the image memory 7 from the TV camera 4, and the control unit 9 expands the previously registered template 21 into the template memory 6 from the nonvolatile memory 11, A matching calculation process is entrusted to the calculation unit 10. In the matching process, for example, the template is moved one pixel at a time in the X direction and the Y direction, and each time the template is correlated with the captured image of the corresponding region, the position with the highest degree of correlation is set as the position where the matching is obtained. End with. The control unit 9 can know the position of the template 21 on the component image to be inspected after confirming the completion of the matching calculation with the calculation unit 11. A rectangular area having this position as the upper left and the template width and height as the width and height, respectively, is used as an area 22 for obtaining the center position of the component, that is, a reference window.
[0027]
[Center position detection]
Subsequently, the center position of the component is detected based on the reference window position acquired in step R2 or the designated window position (step R3). The details are shown in the subroutine of FIG. In essence, the template matching is based on two point-symmetric templates shown in step R32. This is performed as shown in FIGS. A window 22 (region A) in which a part or all of the part enters on the image including the part is set by template matching using the template 21 described above. The actual image of this window (or the reference window itself) is used as the first template, and a second template 23 that is point-symmetric (rotated 180 degrees) is created with respect to this template (FIG. 10). As shown in FIG. 11, the region B, that is, the second template region 23 is detected when the part image and the second template 23 are matched. A midpoint M of a straight line connecting an arbitrary point a in the area of the first template 22 and a point b in the area of the second template corresponding to the point is detected as the center point of the part.
[0028]
If the actual image of the window is the first template (branch to “Yes” in step R30 in FIG. 5), the above-described matching with the point-symmetric template is performed (step R35) to detect the center. However, if the reference window itself is used as the first template, the part may have an inclination and a rotational deviation from the template created in step S3 may occur. If the rotational deviation is large, the matching accuracy is lowered, and therefore the optimum reference window area cannot always be set in step R2. Therefore, the process of step R32 is repeated several times near the position obtained in the reference window (step R31), and the reference window is corrected in step R33. Among these, the combination having the highest correlation value is adopted to obtain the center position (step R34).
[0029]
[Tilt detection]
Subsequently, returning to FIG. 4, in step R4, the inclination of the component is detected. This is performed according to the routine shown in FIG. When the part is a figure that is point-symmetric with respect to its own center, in the XY coordinate system with the center of the part shown in FIG. 12 as the origin, it is Y-axis symmetric with respect to the region 22 acquired in step R2, ie, window A. Region 24 (window C) is set. This region is used as a template for inclination detection (step R41). Assuming that the inclination of the part is θ, when the figure of window C is rotated by −2 * θ around the center of the part, a figure symmetrical with respect to the figure of window A appears.
[0030]
Using this relationship, the image is continuously changed in the window C with a predetermined angle of rotation. Matching is performed with the rotation image obtained at that time (step R44), the step angle is updated at step R45 to rotate the figure (step R43), and this is repeated (step R42).
[0031]
The correlation value of matching when the rotation step angle is updated becomes a distribution curve C as shown in FIG. 13, and the component inclination θ can be obtained by obtaining the step angle i that maximizes the correlation value. (Step R46). Further, in order to increase accuracy, the difference between the position i at which the highest correlation value Y1 is taken and the correlation values Y2 and Y3 at the adjacent step angles (i-1) and (i + 1) is taken, and the zero cross of the difference value is taken. By obtaining the point Z, the inclination θ can be obtained with a resolution greater than the step angle (R47). Further, as shown by the dotted line C ′ in FIG. 13, the distribution curve C may be differentiated to obtain the zero cross point.
[0032]
In the example shown in the figure, the Y-axis symmetric template 24 is created. However, it is also possible to obtain the inclination of the part by creating an X-axis symmetric template and performing template matching in the same manner.
[0033]
In order to improve the tilt accuracy, it is only necessary to make the step angle finer, but that takes time. Further, when the rotational deviation falls within one pixel, the deviation amount cannot be detected. Therefore, if you want to further improve the accuracy (branch to “Yes” in step R47 ′), the center of the part and a certain degree of inclination are known in the processing so far, so by acquiring the external dimensions of the part (Step R48), it is possible to set the best template window with the center of the farthest point P of the part having the largest deviation amount (Steps R49, R50). This is illustrated in FIG.
[0034]
For example, since the external dimensions of the part are known, the window (D) 30 having the upper left corner P of the part as the center can be set. When the image of the part is rotated by −2 * θ with the center of this window as the center of rotation to create a template 31 that is symmetric with respect to the Y axis, the template is a template that shows the characteristics near the upper right corner Q of the part. By moving this template in the X and Y directions and taking the matching as shown in the lower part of FIG. 14, the position of the upper right corner Q of the part on the image can be obtained, and the inclination θ is determined from the positional relationship of PQ. (Steps R51 and R52).
[0035]
If the window 30 used here is located farther from the center than the window 22 shown in FIG. 12, it is possible to expect accuracy higher than that obtained in FIGS. At the time of this matching, the position can be obtained with subpixel accuracy from the position having the highest correlation value and nine neighboring correlation values by subpixel calculation, and PQ is the combination having the largest difference in X components. Therefore, the influence of the matching error is reduced, and better accuracy can be expected. In this way, it is possible to verify the inclination obtained based on the rotation amount and correct it to obtain a highly accurate inclination.
[0036]
As described above, when the position detection of the component is completed, the flow returns to the flow of FIG. 2, and the detection result is received in step S9 following step S8, and the center position and / or the inclination is corrected based on the detection result in step S10. Correct the suction posture of the parts.
[0037]
【The invention's effect】
As described above, according to the present invention, the center and inclination of a component can be detected with high accuracy using the template matching method. Therefore, it can be mounted as a more general-purpose algorithm in a chip mounter or the like, and electronic components can be mounted more easily and accurately.
[0038]
In addition, when the image is rotated and the inclination is obtained, the matching area and the template size are equal. Therefore, only one correlation operation is required for each inspection angle, and the template is created from the image, so that there is no need for normalization. For this reason, the difference can be used as a matching evaluation function, and the calculation time is shortened. In addition, since the matching area has already been determined, it is only necessary to perform rotation conversion calculation for only the matching area when acquiring a rotated image, and in this respect, the calculation time can be shortened.
[Brief description of the drawings]
FIG. 1 is a block diagram showing the overall configuration of a system in which the present invention is used.
FIG. 2 is a flowchart showing the overall flow.
FIG. 3 is a flowchart showing a routine for creating and registering a template.
FIG. 4 is a flowchart showing a routine for performing position detection.
FIG. 5 is a flowchart showing a routine for detecting the center position of a component.
FIG. 6 is a flowchart showing a routine for detecting the inclination of a component.
FIG. 7 is an explanatory diagram illustrating a state in which a component posture is corrected when a template is created.
FIG. 8 is an explanatory diagram showing creation of a template.
FIG. 9 is an explanatory diagram illustrating a state in which a reference window position is acquired.
FIG. 10 is an explanatory diagram showing a state in which a point-symmetric template is formed from a reference window.
FIG. 11 is an explanatory diagram showing matching with a point-symmetric template.
FIG. 12 is an explanatory diagram showing a state in which the inclination of a component is detected.
FIG. 13 is a diagram showing a distribution of correlation values by matching at the time of inclination detection.
FIG. 14 is an explanatory diagram showing another example of tilt detection.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 Component imaging stage 2 Position control apparatus 3 Man-machine interface 4 TV camera 5 Display apparatus 12 Image processing apparatus 20 Components 21, 23, 31 Templates 22, 24, 30 Window

Claims (5)

Take an image of a point-symmetric part about its center,
Specify a reference window that contains part or all of the part image using coordinate values ,
A window symmetrical to the reference window with respect to the X axis or Y axis passing through the center of the part is set as a template,
Matching a template of a window symmetrical to the reference window and an image obtained by rotating the image around its center, and detecting the inclination of the component based on the amount of rotation when the matching is achieved A feature position detection method. Take an image of a point-symmetric part about its center,
Create a template that contains part or all of the part from the image of the part, register it ,
Set the reference window at the template position where the degree of correlation is maximized by matching the registered template and the image of the part ,
A window symmetrical to the reference window with respect to the X axis or Y axis passing through the center of the part is set as a template ,
Matching a template of a window symmetrical to the reference window and an image obtained by rotating the image around its center, and detecting the inclination of the component based on the amount of rotation when the matching is achieved A feature position detection method . Take an image of a point-symmetric part about its center,
Specify a reference window that contains part or all of the part image using coordinate values ,
Using the reference window as a first template, creating a second template that is point-symmetric with the first template,
A midpoint of a straight line connecting a point in the first template region when matching between the image and the second template is obtained, and a point in the second template region corresponding to this point As the center of parts,
A window symmetrical to the reference window with respect to the X axis or Y axis passing through the center of the part is set as a template ,
Matching a template of a window symmetrical to the reference window and an image obtained by rotating the image around its center, and detecting the inclination of the component based on the amount of rotation when the matching is achieved A feature position detection method. Take an image of a point-symmetric part about its center ,
Create a template that contains part or all of the part from the image of the part, register it ,
Set the reference window at the template position where the degree of correlation is maximized by matching the registered template and the image of the part ,
Using the reference window as a first template, creating a second template that is point-symmetric with the first template ,
A midpoint of a straight line connecting a point in the first template region when matching between the image and the second template is obtained, and a point in the second template region corresponding to this point As the center of parts ,
A window symmetrical to the reference window with respect to the X axis or Y axis passing through the center of the part is set as a template ,
Matching a template of a window symmetrical to the reference window and an image obtained by rotating the image around its center, and detecting the inclination of the component based on the amount of rotation when the matching is achieved A feature position detection method . The rotation is performed at a predetermined step angle, the difference or derivative of the matching correlation value is obtained , and the inclination of the component is detected by obtaining the zero cross point of the difference value or the derivative value. 5. The method according to any one of 4 above.

Images