JP4566686B2 - Method and apparatus for determining shape of object - Google Patents

Description

  The present invention relates to a method and apparatus for determining the shape of an object, and more specifically, a method for determining the shape of an object such as a ball, land, or mark for component polarity determination provided on an electronic component mounted on a substrate. And an apparatus.

  2. Description of the Related Art Conventionally, in a component mounter that mounts an electronic component (hereinafter simply referred to as a component) on a substrate, an image of an adsorbed electronic chip component (hereinafter simply referred to as a component) is captured from an imaging means such as a CCD camera, and the image of this component is captured. The position of the component is detected by recognizing the component, and the component is positioned by correcting the posture of the component by the suction nozzle as necessary, and the component is mounted at a predetermined position on the board.

  In addition, if the part has polarity, it is necessary to mount the part in consideration of the direction of the part.For this purpose, a notch indicating the polarity of the part is formed at the corner of the part, and at the time of part recognition, The position of the corner notch is detected to determine the polarity of the component, and the component is placed on the board in the correct orientation, that is, the corner with the notch is oriented in a predetermined direction (such as the upper right direction on the board). It is carried out (Patent Document 1).

Further, in order to indicate the polarity of a component, a ball, a land, or a mark capable of discriminating the polarity is formed at the corner of the component. For example, as shown in FIGS. 8A and 8B, Rectangular or circular marks 30a, 31a are formed at the corners of the parts 30, 31 having a large number of ball electrodes 30b, 31b arranged in a matrix. The shape, size, position from the center of the part (relative coordinates), etc. are stored in advance as part data, and model template image data is created in the device based on these part data when the part position is detected. Then, the normalized correlation matching is performed, and the mark at the corner of the component is detected and the shape is determined based on the calculated correlation value. In this case, when the mark is circular, for example, a discrimination method is proposed in which the lengths of the diameters in a plurality of directions are compared to determine whether or not the mark is a circle (Patent Document 2).
JP-A-8-288694 Japanese Patent No. 2959017

  When a mark is formed on a conventional substrate and the polarity of the component is discriminated by it, in normalized correlation matching, for example, if the mark is circular or rectangular, the difference in shape is difficult to recognize. Returns a high correlation value. The difference between the two is a level that is buried in the fluctuation due to the size error, so that the presence / absence of position detection can be determined by determining the correlation value, but it is difficult to determine the shape.

  Further, in the image data picked up for positioning, the object (ball, land, or mark) to be used for polarity discrimination is about 5 to 8 pixels and is not a sufficient size for shape discrimination. As shown in FIG. 7A, each rectangular mark 32 (six on the left side) and each circular mark 33 (four on the right side) are digitized as image data. Is not clear, and even if it is a circle, it is difficult to consider its boundary as a curve. Further, when these images are converted into binarized images, as shown in FIG. 7B, the contours become marks 32 ′ and 33 ′ of images different from the original contours, making recognition more difficult. .

  In such image data, shape feature values such as the diameter length, perimeter length, and area cannot be obtained with high accuracy, and it is difficult to determine the shape using these as evaluation values.

  Accordingly, the present invention has been made to solve such problems, and an object shape discrimination method and apparatus capable of accurately discriminating the shape of a circular or rectangular (square or rectangular) object. The issue is to provide.

The present invention (Claim 1)
A method for determining the shape of an object provided on an electronic component mounted on a substrate,
When the object is imaged, and the image data is enlarged by digital interpolation according to the size of the object, and the enlargement process is performed, a scanning line is set radially from the center of the image of the object and the scanning is performed. When sampling line image data and separating the sampled section into two, a point having the highest degree of separation is extracted as a contour point, and the perimeter of the contour line obtained by connecting each contour point Find the area of the part surrounded by the outline,
A first evaluation value indicating circularity is calculated from the obtained perimeter and area,
A model of the object is set, and a second evaluation value indicating the degree of coincidence between the gradient vector at the contour point of the object and the gradient vector at the contour point of the model corresponding to the contour point is set as the object. Calculated by a weighted average obtained by increasing the weight of the gradient vector of the portion where the difference between the shape of the model and the shape of the model is large ,
In accordance with the shape size of the object and the shape determination condition, the weighting coefficient is set to vary for each evaluation value, and the first and second evaluation values are respectively weighted with the set weighting coefficient,
It is characterized by determining whether a target object is circular.

The present invention (Claim 2 )
An apparatus for determining the shape of an object provided on an electronic component mounted on a substrate,
An imaging means for imaging an object;
Image enlarging means for enlarging the image data by digital interpolation according to the size of the object;
When the enlargement process is performed, the scanning line is set radially from the center of the image of the object, the image data of the scanning line is sampled, and the degree of separation is the highest when the sampled section is separated into two. Contour point extraction means for extracting a point that becomes higher as a contour point;
Calculating means for calculating the perimeter of the contour line obtained by connecting each contour point and the area of the portion surrounded by the contour line;
Calculating means for calculating a first evaluation value indicating circularity from the obtained perimeter and area;
A model of the object is set, and a second evaluation value indicating the degree of coincidence between the gradient vector at the contour point of the object and the gradient vector at the contour point of the model corresponding to the contour point is set as the object. Calculating means for calculating a weighted average obtained by increasing the weight of the gradient vector of a portion where the difference between the shape of the model and the shape of the model is large ;
Weighting means for setting a weighting coefficient for each evaluation value in accordance with the shape size and shape discrimination condition of the object, and weighting each of the first and second evaluation values with the set weighting coefficient ; Have
It is characterized by determining whether a target object is circular.

  In the present invention, in addition to the evaluation value indicating the circularity, two evaluation values indicating the degree of coincidence between the gradient vector at the contour point of the object and the gradient vector at the contour point of the model corresponding to the contour point are used. When the evaluation value is used to determine whether or not the object is circular, the accuracy of the determination is improved, and circular or rectangular marks are provided at the corners of the component to determine the polarity of the electronic component In addition, erroneous recognition of the mark is eliminated, and erroneous mounting can be prevented.

  In addition, in the case of a component in which a circular electrode and a rectangular electrode are mixed, since the component can be recognized by discriminating between the circular electrode and the rectangular electrode, the recognition accuracy is improved.

  The present invention can recognize an object such as a ball, a land, or a mark used for determining the polarity of a component with high accuracy, and can discriminate between a circular electrode and a rectangular electrode with high accuracy. The invention will be described in detail on the basis of the examples shown.

  FIG. 1 is a block diagram of a system used to implement the shape discrimination method of the present invention. This system picks up a component 2 and picks up a nozzle 2 for setting the image pickup position, and a machine for controlling the mounting operation of the component on a substrate (not shown) such as moving the suction nozzle 1 and driving the illumination device 3. The control device 13, the standard camera 4 and the high-resolution camera 5 for imaging the component 2, and processing the captured image to recognize the component and detect the object provided at the component existing at a certain position in the image. And an image processing device 12 for determining the shape.

  The image processing apparatus 12 includes an A / D converter 6 that converts an image of a part captured by the cameras 4 and 5 into a digital image, an image memory 7 that stores the image converted into a digital image, a work memory 8, and an image memory. 8 includes an arithmetic unit 9 that performs arithmetic processing on data, a component data storage memory 10, an interface 11, and a control unit 4 that controls the flow of data between each unit or memory and calculates data necessary for component mounting. The arithmetic device 9 and the control device 4 are constituted by a CPU or the like.

  The machine control device 13 transmits in advance information on the electrode of the component, element information for polarity determination, a determination threshold value and the like (hereinafter referred to as component data) to the image processing device 12 via the interface 11. The image processing device 12 stores the received component data in the component data storage memory 10.

  The machine control device 13 normally selects the standard camera 4 or the high-resolution camera 5 depending on the electrode size of the component, sucks the component 2 with the suction nozzle 1, and sets it at the imaging position of the selected camera. At this time, the illumination device 3 is moved and lit so that it can be imaged by the selected camera, and the image processing device 12 is instructed to execute processing together with the selected camera channel information via the interface 11.

  The image processing device 12 controls the designated camera 4 or 5, digitizes the captured image of the component 2 by the A / D converter 6, and stores it in the image memory 7 as multivalued image data. Then, the arithmetic unit 9 processes the data in the image memory 7, detects the position of the component 2, calculates the deviation between the component center and the adsorption center, and the component adsorption inclination, and the component supply direction is incorrect. An object such as a ball, land, or mark for polarity discrimination is detected and the shape discrimination inspection is performed. If the inspection result is OK, the position detection result is returned to the machine control device 13 via the interface 11. If the inspection result is NG, it returns a response that the polarity determination is an error.

  When the machine control device 13 receives the positioning result normally, the machine control device 13 moves the suction nozzle 1 to the mounting position and corrects the component suction displacement (deviation between the component center and the suction center and the suction angle displacement) according to the component recognition result. The component 2 is mounted on the board with the correct polarity. On the other hand, when a polarity determination error is received, the machine control device 13 corrects the orientation of the component and performs a retry, or discards the component.

  Next, the flow of processing in the image processing apparatus 12 will be described in detail with reference to FIG. Here, the arithmetic unit 9 of the image processing apparatus 12 includes calculation means for calculating various data necessary for shape discrimination such as the perimeter of the contour line, the area of the portion surrounded by the contour line, and the evaluation value for shape discrimination. It has become.

  In step S <b> 1, the image processing apparatus 12 controls the designated camera (imaging unit) 4 or 5 to image the component and the object provided in the component, and the captured image is converted into the A / D converter 6. Is digitized and stored in the image memory 7 as multivalued image data.

  Subsequently, the arithmetic unit 9 of the image processing device 12 processes the data in the image memory 7 based on the electrode information stored in the component data storage memory 10, detects the electrode center position, and determines the center and inclination of the component. Calculate and position the component (step S2). When the positioning fails (N in step S3), the image processing apparatus 12 notifies the machine control apparatus 13 of that, interrupts the subsequent processing, and enters a state of waiting for the next command. On the other hand, if the positioning is successful (Y in step S3), the process proceeds to the polarity determination process in step S4 and subsequent steps.

  In step S4, the polarity discriminating element is determined from the relative coordinates from the center of the part of the polarity discriminating element (object such as a ball, land, or mark) stored in the part data storage memory 10 and the positioning result of the part obtained in step S2. Is predicted, the data in the image memory 7 is processed, and the element center and its diameter are detected. If this detection fails, the image processing device 12 notifies the machine control device 13 of the failure, interrupts the subsequent processing, and waits for the next command. If the detection is successful, the process proceeds to the following shape discrimination process.

  The size of the element whose shape can be discriminated should be at least 8 pixels × 8 pixels, preferably 32 pixels × 32 pixels. Therefore, if the element diameter obtained in step S4 is less than 32 pixels × 32 pixels, the image A partial image is cut out from the memory 7 and an enlarged image of the element is generated in the work memory 8 by digital interpolation processing. At this time, since the component inclination has been acquired, if the normalization to the posture with the inclination of 0 ° is performed at the same time, the processing becomes easier. The above process is the process of step S5 (correction means).

  Subsequently, in step S6, contour data (contour points, gradient vectors) of the shape determination object (element) is extracted. In the present invention, since the enlargement process is performed by slightly high-magnification digital interpolation on the captured image, the image is blurred. In order to extract a contour from such an image, a method described in Japanese Patent Laid-Open No. 2000-205838 is used.

For example, in FIGS. 4A and 4B, reference numeral 40 denotes a circular object for polarity discrimination such as a ball, a land, or a mark formed at a corner of a component (corresponding to the mark 31a in FIG. 8). image, reference numeral 41 is rectangular (square or rectangular) shape of an object and the image (corresponding to the mark 30a in FIG. 8), sets the detection window W, the scanning line center 40a of the image, from 41a to radiation form m1 is set, and the image data of the scanning line m1 is sampled. Sampling interval shall be monotonously increasing or monotonically decreasing. Points that separate this section into two are sequentially set from the start point to the end point, and the point having the highest degree of separation is defined as a contour point p1. Then, the scanning lines m2, m3,. . . . . And contour points p2, p3,. . . . . Are connected to form contour lines 40b and 41b. As long as the edge is clear (if the contrast is good), this contour line may be extracted using an edge detection filter or contour extraction by binarized boundary tracking.

  Next, each contour point p1, p2,. . . . . A 3 × 3 Sobel filter is applied to obtain vector components in the x and y directions to obtain gradient vectors. Then, the contour data of the contour points and gradient vectors obtained in this way are stored in the work memory 8.

  Next, the contour data obtained in step S6 is taken out from the work memory 8, the sum of the distances between the contour points is obtained as the perimeter, and the number of pixels in the portion surrounded by the contour lines 40b and 41b is counted. (Step S7).

  In addition, when there is a concave portion with respect to the scanning line, a shadowed pixel is generated. Therefore, scanning in two directions of xy is performed, the pixels outside the object in the rectangle are counted, and subtracted from the rectangular area. To determine the area of the object. The process is performed as follows.

  First, the contour data (contour point coordinate sequence) obtained in step S6 is read from the work memory 8, digital line segment interpolation is performed between the contour points, line drawing data is generated on the work memory 8, and a rectangle that wraps the contour line. Ask for. Subsequently, raster scanning is performed from the left and right ends of the rectangle in the X direction to detect left and right edge pixels of the work (object). At that time, a mark is added to the pixels counted during the x-direction scanning so that the counting can be skipped during the next y-direction scanning. Next, sub-pixel calculation is performed using multi-valued image data to obtain real-valued coordinates, and then raster scanning is performed from the upper and lower ends of the rectangle in the y direction to detect upper and lower edge pixels of the work, Pixels leaked in the x-direction scan outside the object inside the rectangle are counted, and a pixel counter outside the object is subtracted from the envelope rectangular area to obtain an area.

Thus, after calculating | requiring the area of a target object, the circularity is calculated in step S8. In that case, as the evaluation value e (first evaluation value) of the circularity,
e = 4πS / (L * L) (0 <e ≦ 1)
S: area, L: perimeter are used. In the case of a perfect circle, e = 1, and in the case of a square, e = 0.785. As the contour becomes more complicated, e becomes smaller.

As shown in FIG. 5, whether or not the object is a circle is determined by the contour points p1, p2,. . . . . Gradient vectors v1, v2,. . . . . And the gradient vector v1 ′, v2 ′,. . . . . Since the degree of coincidence (cos θ) in the direction of (radiation for detecting contour points) can also be determined, a vector correlation calculation is performed in step S9. This vector correlation calculation is performed as follows. First, a model 42 (model circle) is set for the object 41, and gradient vectors v1, v2,... At each contour point of the object 41 and the corresponding contour points on the model circle 42 are set. . . . . (Solid line); v1 ′, v2 ′,. . . . . (Dotted line) is obtained. For example, the vector component of the gradient vector v2, v2 ′ on the contour point A (= p2) of the object 41 and the contour point B of the model circle 42 corresponding thereto is
v2 (A) = (x0, y0)
v2 ′ (B) = (x1, y1)
Then, the angle θ between the two vectors can be obtained by dividing the inner product of the two vectors by the product of the lengths of the respective vectors.

cos θ = (v2 · v2 ′) / (| v2 || v2 ′ |)
= (X0 * x1 + y0 * y1) / (sqrt (x0 * x0 + y0 * y0) * sqrt (x1 * x1 + y1 * y1))
(-1 ≦ cos θ ≦ 1)
, The degree of coincidence c (second evaluation value),
c = | cos θ | (0 ≦ c ≦ 1) can be obtained. In the above formula, sqrt means the square root of the square. The degree of coincidence is acquired at each contour point, and the average value d = Σci / n is used as the evaluation value (this may be the second evaluation value).

  Further, when determining whether the shape of the object is a circle or a rectangle, the sensitivity can be increased by using a method such as making a sampling point a location where the difference is large or changing the weight distribution. For example, in the case of a rectangle, as shown in FIG. 5, if the sampling points are provided at 22.5 ° ≦ θ1 <45 ° and 45 ° <θ2 ≦ 67.5 °, the sensitivity is improved.

  Subsequently, from the two evaluation values obtained in steps S8 and S9, the weighting is varied according to the state of the workpiece (object) with respect to each evaluation value, and a weighted average is obtained to obtain a final evaluation value ( Step S10). For example, according to the state of the object, the weighting coefficient as shown in FIG. 6 is multiplied by the evaluation value obtained from the circularity and the evaluation value obtained from the vector correlation calculation, respectively, and the weighted average is obtained. In FIG. 6, “work size” indicates how large the object (work) is in pixel units, for example, 32 * 32 is the size of 32 pixels * 32 pixels. It is shown that. “Enlargement” and “Inclination” indicate whether enlargement processing or rotation correction has been performed in step S5, respectively. “Comparison shape” merely determines whether or not the shape is a circle, or Whether to determine whether the shape is a circle or a rectangle is shown. Then, according to these conditions, the weighting factor as shown in “Circularity” and the weighting factor as shown in “Vector correlation” were obtained from the evaluation value (e) indicating the circularity and the vector correlation calculation, respectively. The evaluation value is weighted, and the weighted average is obtained as the final evaluation value.

  Next, the evaluation value obtained in step S10 is compared with a preset threshold value to perform shape discrimination (step S11). If it is equal to or greater than the threshold value, it is determined that it is circular, and OK is returned (steps S11 and S12). If it is less than the threshold value, it is determined that it is not circular (rectangular), and NG is returned ( Steps S11 and S13).

  Here, the threshold value for evaluation value determination is measured in advance by teaching, and if set, the determination accuracy is improved. Therefore, the threshold value is obtained along the flow shown in FIG. This process is also a process for examining in advance whether or not the shape to be used can be determined.

  First, evaluation values are calculated from images of several workpieces having a good shape (OK), and the minimum value (a) is obtained (steps S20 and S21). Subsequently, an evaluation value is calculated from images of several workpieces having a defective shape (NG), and the maximum value (b) is obtained (steps S22 and S23). Next, a and b are compared (step S24). If a> b, the average value of a and b is set as a threshold value to notify that shape discrimination is possible (step S25). When ≦ b, it is notified that the shape is indistinguishable (step S26).

It is the block diagram which showed the structure of the image processing apparatus used for this invention. It is the flowchart which showed the flow of shape discrimination | determination. It is the flowchart which showed the flow which sets the threshold value for shape discrimination | determination. It is explanatory drawing which showed the state which calculates | requires the outline point of the image of a target object. It is explanatory drawing which showed the state which calculates | requires the correlation calculation of the gradient vector in an outline point. It is a table | surface figure which shows the weighting coefficient set according to the determination conditions. It is explanatory drawing which shows the imaging state of a target object. It is explanatory drawing which showed the mark etc. which show the polarity of components.

Explanation of symbols

2 Electronic parts 12 Image processing device 13 Machine control device 40, 41 Image of object 42 Model circle

Claims (2)

A method for determining the shape of an object provided on an electronic component mounted on a substrate,
When the object is imaged, and the image data is enlarged by digital interpolation according to the size of the object, and the enlargement process is performed, a scanning line is set radially from the center of the image of the object and the scanning is performed. When sampling line image data and separating the sampled section into two, a point having the highest degree of separation is extracted as a contour point, and the perimeter of the contour line obtained by connecting each contour point Find the area of the part surrounded by the outline,
A first evaluation value indicating circularity is calculated from the obtained perimeter and area,
A model of the object is set, and a second evaluation value indicating the degree of coincidence between the gradient vector at the contour point of the object and the gradient vector at the contour point of the model corresponding to the contour point is set as the object. Calculated by a weighted average obtained by increasing the weight of the gradient vector of the portion where the difference between the shape of the model and the shape of the model is large ,
In accordance with the shape size of the object and the shape determination condition, the weighting coefficient is set to vary for each evaluation value, and the first and second evaluation values are respectively weighted with the set weighting coefficient,
A method for discriminating the shape of an object, comprising discriminating whether or not the object is circular. An apparatus for determining the shape of an object provided on an electronic component mounted on a substrate,
An imaging means for imaging an object;
Image enlarging means for enlarging the image data by digital interpolation according to the size of the object;
When the enlargement process is performed, the scanning line is set radially from the center of the image of the object, the image data of the scanning line is sampled, and the degree of separation is the highest when the sampled section is separated into two. Contour point extraction means for extracting a point that becomes higher as a contour point;
Calculating means for calculating the perimeter of the contour line obtained by connecting each contour point and the area of the portion surrounded by the contour line;
Calculating means for calculating a first evaluation value indicating circularity from the obtained perimeter and area;
A model of the object is set, and a second evaluation value indicating the degree of coincidence between the gradient vector at the contour point of the object and the gradient vector at the contour point of the model corresponding to the contour point is set as the object. Calculating means for calculating a weighted average obtained by increasing the weight of the gradient vector of a portion where the difference between the shape of the model and the shape of the model is large ;
Weighting means for setting a weighting coefficient for each evaluation value in accordance with the shape size and shape discrimination condition of the object, and weighting each of the first and second evaluation values with the set weighting coefficient ; Have
An object shape discriminating apparatus for discriminating whether or not an object is circular.