JPH10288505A - Both work position detection and appearance inspection methods and respective devices thereof - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make any processing performable accurately and speedily at a time when each work position, for example, the position of a semiconductor package (a position of each ball), where regular sized members are plurally arranged in a fixed rule on the top of a work or an object of inspection is roughly specified in advance of the full-scale inspection. SOLUTION: A top face of a work 1 is imaged in a picture taken, and a first detection area 12 to be circumscribed with plural pieces of arrayal members 2 is fed out of this pickup picture image 10. Then, a second detection area 15 showing these members 2 in a line is fed out of this first detection area 12. In succession, well-known relative position data of these plural arrayal members 2 in relation to the work 1 and the first detection area 12 showing these arrayal members 2 fed out are butted to each other, whereby an absolute center position C of the work 1 is detected, while the well-known relative position data and the second detection area 15 showing a line of these members 2 fed out are butted to each other, whereby an attitude of the work 1 is detected as well.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for detecting and inspecting the position of a work such as a semiconductor package, and more particularly to a method in which a plurality of members having a fixed size are arranged on a top surface of a work according to a fixed rule. A work position detection method and apparatus for detecting the position of a work by image processing, and an image of a work position in which a plurality of members of a fixed size are arranged in a lattice on the upper surface of the work and positions of the respective members on the work. The present invention relates to a method and an apparatus for inspecting the appearance of a work in which members detected on a work are detected in advance as processing, and the detected positions are used as a reference position of the work and a reference position of each member.

[0002]

2. Description of the Related Art BGA (Ball Glid Array) or CS
As shown in FIG. 1, a semiconductor package called a P (Chip Size Package) has two ball-shaped solder bumps 2 (hereinafter referred to as balls 2) on the back surface of the semiconductor package 1.
The semiconductor package 1 is mounted in such a manner that the semiconductor package 1 is directly soldered to the printed wiring board by the balls 2 and mounted.

A plurality of semiconductor packages 1 are placed on a tray 7 and the semiconductor package 1 to be inspected is transferred to a position directly below the camera 3 by an XYZ stage 8 as a transfer device. An appearance inspection is performed on the basis of the image pickup result to determine whether or not the height of each ball 2 on the semiconductor package 1 to be inspected is constant.

That is, in FIG. 1, a semiconductor package 1 is placed on a tray 7 upside down, and an illumination device 9 for obtaining illumination for cameras 3 and 4 is provided above the semiconductor package 1. It is arranged. Upper camera 3
Captures the back surface image of the semiconductor package 1 as a planar image. The oblique camera 4 captures an image of the back surface of the semiconductor package 1 on which the balls 2 are disposed from obliquely above.

In this way, three-dimensional information (plane position (x, y) and height position z) of each ball 2 is obtained based on the image data captured by the two cameras 3 and 4, and a predetermined inspection is performed.

However, when the semiconductor package 1 to be inspected is transported and positioned directly below the upper camera 3 by the XYZ stage 8, the accuracy of the XYZ stage 8, the accuracy of the tray 7, and the accuracy of the The semiconductor package 1 is not always transported to the same position and is not always positioned due to a shift due to play or the like.

Therefore, prior to performing a full-scale inspection,
In order to shorten the inspection time, it is necessary to roughly specify the position of the semiconductor package 1 to be inspected (the position of each ball 2).

FIG. 19 shows a conventional technique for roughly specifying the position of the semiconductor package 1 to be inspected (the position of each ball 2).

At the time of setting, that is, as shown in FIG. 19A, a partial image (corner portion) of the semiconductor package 1 existing at a reference position (ideal position) in the captured image 10 of the upper camera 3 in advance. Is stored as the model image 40. Further, the relative position between each ball 2 to be inspected and the model image 40 is stored as a vector 41.

[0010] At the time of inspection execution As shown in FIG.
The position of 'is shifted from the reference position 1. Therefore, the same image 40 ′ as the model image 40 is searched from the captured image 10 of the upper camera 3 using, for example, pattern matching using a correlation coefficient. The actual position of each ball 2 on the semiconductor package 1 'is determined by the movement amount (vector) 42 of the model image 40 and the relative position (vector) 41 of the ball 2 with respect to the model image 40 stored at the time of the above setting. It can be specified based on

However, in this prior art, since the search is performed using only one model image 40, the position cannot be accurately specified when the actual semiconductor package 1 'is inclined. In other words, since the correlation coefficient at the time of putter matching is significantly reduced due to the inclination, the accuracy of pattern matching is reduced. Even if the pattern matching succeeds, only the amount of parallel movement is measured and the amount of inclination of the semiconductor package 1 is not measured, so that the position of each ball 2 includes an error corresponding to the amount of inclination. Will be.

In order to compensate for such a disadvantage, FIG.
As shown in FIG. 0, a search using two types of model images is also performed.

At the time of setting, that is, as shown in FIG. 20A, images (corner portions) of different portions of the semiconductor package 1 existing at a reference position (ideal position) in the captured image 10 of the upper camera 3 in advance.
Are stored as model images 43A and 43B. Further, the position and the inclination of the line segment 44 connecting the model images 43A and 43B are stored. Furthermore, this line segment 44
Is stored as a vector.

At the time of inspection execution As shown in FIG.
The position of 'is shifted from the reference position 1 by a predetermined amount of parallel movement and an inclination. Therefore, the model image 43A
An image 43 ′ A that is the same as the image 43 ′, and an image 43 ′ B that is the same as the model image 43 B
Are searched from among the patterns using, for example, pattern matching using a correlation coefficient. At this time, the searched image 4
The position and inclination of a line segment 44 'connecting 3'A and 43'B are measured.

The actual position of each ball 2 on the semiconductor package 1 'is stored in the position of the line segment 44' thus measured, the position of the line segment 44 when the inclination is set, the difference with respect to the inclination, and at the time of the above setting. It can be specified based on the relative position of each ball 2 to be inspected with respect to the set line segment 44.

In this way, the position of each ball 2 is specified more accurately than the method shown in FIG. 19, taking into account the actual amount of inclination of the semiconductor package 1.

[0017]

However, since the lighting device 9 used for the above inspection uses lighting that lifts only the ball 2 on the semiconductor package 1, the actual position of the semiconductor package 1 'is I have to search depending on the existence of. That is, the ball 2 must be included in the model image 43A and the like used in the above-described conventional technology.

However, in the actual semiconductor package 1 'to be inspected, the ball 2 is often defective because the ball 2 itself is missing or greatly deformed. For this reason, when pattern matching is performed on a semiconductor package 1 ′ having such a defective ball 2 using a model image including the ball 2, the accuracy of the pattern matching is reduced, and in some cases, the pattern matching itself may fail. is there.

If the pattern matching taking into account the inclination shown in FIG. 20 is performed on the image 17 'taken by the oblique camera 4, the position of the ball 2 may not be accurately grasped.

That is, as shown in FIG. 21, when the semiconductor package 1 is rotated by the XYZ stage 8 as shown by the arrow E and the image is taken by the oblique camera 4 at each rotation position, FIGS. 22 (a) and 22 (b) As shown in FIG.
The semiconductor packages 1a and 1b in 7 'are deformed into a trapezoidal shape with respect to the actual square shape, and are acquired in different shapes 1a and 1b for each rotation position. Such deformation may reduce the accuracy of pattern matching. Further, regarding the optical system, it is necessary to make a correction in consideration of how the semiconductor package 1 'is deformed according to the rotational position, so that the arithmetic processing becomes complicated, and the position of the ball 2 can be quickly specified. There is also the problem of not being able to do so.

The present invention has been made in view of such a situation, and prior to a full-scale appearance inspection, a plurality of members of a predetermined size are arranged on a top surface of a work to be inspected according to a predetermined rule. It is an object of the present invention to solve the problem in that when a position of a work (for example, a position of a semiconductor package (a position of each ball)) is roughly specified, the processing can be performed accurately and promptly.

[0022]

SUMMARY OF THE INVENTION Therefore, in the main invention of the present invention, a work for detecting, by image processing, the position of a work in which a plurality of members of a fixed size are arranged on a top surface of the work according to a fixed rule. In the position detection method, an imaging step of imaging the upper surface of the workpiece and a first image processing step of cutting out a first detection area circumscribing the plurality of array members from an image captured in the imaging step And a second image processing step of cutting out a second detection area indicating a row of members from the first detection areas indicating the plurality of array members cut out in the first image processing step; and By comparing the known relative position data of the plurality of array members with respect to the first detection region indicating the plurality of array members cut out in the first image processing step, it is possible to cut off the workpiece. A work for detecting a posture of the work by detecting a center position and matching the known relative position data with a second detection area indicating a row of members cut out in the second image processing step. A position and posture detection process is provided.

That is, as shown in FIGS. 6 and 10, the upper surface of the work 1 is imaged.
The first detection region 12 circumscribing the plurality of array members 2, 2,...
Is cut out. Are cut out of the first detection area 12 showing the plurality of arranged members 2, 2,.... Then, a plurality of arrangement members 2, 2,.
The absolute center position C of the workpiece 1 is detected by matching the known relative position data with the first detection region 12 indicating the plurality of cut-out arrangement members 2, 2,. The position φ of the work 1 is detected by matching the position data with the second detection area 15 indicating the cut-out members 2, 2,....

As described above, the first detection area 12 is set in units of the entire members 2, 2,... Of the actual work 1 to be inspected.
Are cut out, and the second detection area 15 is cut out in units of one row of members 2, 2..., And the position C and the posture φ of the work 1 are detected based on these. Even if there is a defect in which some of the members 2 are missing (Em indicates a missing portion in FIG. 10) or the member 2 is greatly deformed, it is affected by the partially defective members. Without this, the position C and posture φ of the work 1 can be accurately detected.

According to a second aspect of the present invention, there is provided a work position detecting method for detecting, by image processing, the position of a work in which a plurality of members of a predetermined size are arranged in a grid on the upper surface of the work. An imaging step of imaging the upper surface of the image, and a detection area setting step of cutting out a detection area including the entire work from the image captured by the imaging step,
In each of the X-axis direction and the Y-axis direction of the detection area cut out in the detection area setting process, the position on the image and the sum of the brightness of all the pixels corresponding to this position or the value obtained by evaluating the sum A calculation process for obtaining the relationship, a peak value extraction process for extracting a peak value of a value obtained by evaluating the added value of the brightness or the added value in the relationship obtained in the calculation process, and an X-axis extracted in the peak value extraction process. Each position in the X-axis direction corresponding to each peak value in the direction is defined as an arrangement position in the X-axis direction of the member, and corresponds to each peak value in the Y-axis direction extracted in the peak value extraction process. An arrangement position detection process in which each position in the Y-axis direction is an arrangement position in the Y-axis direction of the member; a member arrangement position in the X-axis direction obtained in the arrangement position detection process;
By matching the member arrangement position in the axial direction,
A member position detection step of detecting each X, Y position of the plurality of arrangement members, known relative position data of the plurality of arrangement members with respect to the workpiece, and a plurality of arrangement members obtained in the member position detection step By comparing the X and Y positions, a work position for detecting the position and posture of the work, and a posture detection process are provided.

That is, as shown in FIGS. 11, 12, and 13, the upper surface of the work 1 is imaged.
From the detection areas 18A and 18B including the entire work 1
(20C, 20D) is cut out. Then, for each of the X-axis direction and the Y-axis direction of the cut-out detection region, a position 22 on the image and a value Px obtained by evaluating the added value of the brightness of all the pixels 22A corresponding to the position 22 or the added value Is required. Then, the peak value 23 of the value Px obtained by evaluating the added value of the brightness or the added value in this relationship is extracted. Then, each of the extracted peak values 231, 232, 233, 234, 23 in the X-axis direction is extracted.
Each position x1, x2, x3 in the X-axis direction corresponding to 5, 236
It is assumed that x4, x5, and x6 are the array positions of the member 2 in the X-axis direction, and the respective positions in the Y-axis direction corresponding to the extracted peak values in the Y-axis direction are the positions of the members 2 in the Y-axis direction. It is assumed to be an array position.

Then, as shown in FIG.
By abutting the member arrangement positions 23A and 23B in the axial direction with the member arrangement positions 23C and 23D in the Y axis direction, the X and Y positions (x, y) of the plurality of arrangement members 2, 2,.
Is detected.

As described above, the member arrangement position is obtained for each of the X axis and the Y axis, and the obtained member arrangement position 2 in the X axis direction is obtained.
3A, 23B and member arrangement positions 23C, 23 in the Y-axis direction
D, the plurality of array members 2,
Since the X and Y positions (x, y) of 2... Are detected, the work 1 is rotated by, for example, the XYZ stage 8, and is imaged by the oblique camera 4 at each rotation position. Is deformed and has a different shape for each rotational position, the position of each member 2 can be quickly and accurately specified without being affected by these.

[0029]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of a method and an apparatus according to the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an inspection device for a semiconductor package 1 assumed in this embodiment.

As shown in FIG. 1, a plurality of semiconductor packages 1, 1,... Are placed on a tray 7, and a semiconductor package 1 to be inspected is moved by a camera by an XYZ stage 8 as a transfer device. 3 is transported to just below.

Here, the semiconductor package 1 to be inspected is
Is a BGA (Ball Glid Array) or CSP (Chip Si
1 and 6, a ball-shaped solder bump 2 (hereinafter referred to as a ball 2) is formed on the back surface of the semiconductor package 1 as shown in FIGS.
The semiconductor package 1 is dimensionally arranged in a lattice, and the semiconductor package 1 is directly soldered to the printed wiring board by these balls 2 and mounted.

Then, based on the imaging results of the cameras 3 and 4, an appearance inspection is performed to determine whether or not the height of each ball 2 on the semiconductor package 1 to be inspected is constant.

That is, in FIG. 1, the semiconductor package 1 is placed upside down on a tray 7, and above the semiconductor package 1, an illuminating device 9 for obtaining illumination for the cameras 3 and 4 is provided. It is arranged. Upper camera 3
Captures the back surface image of the semiconductor package 1 as a planar image. The oblique camera 4 captures an image of the back surface of the semiconductor package 1 on which the balls 2 are disposed from obliquely above.

In the image processing computer 5, based on the image data captured by the two cameras 3 and 4,
The three-dimensional information (plane position (x, y) and height position z) of each ball 2 is calculated, and a predetermined visual inspection is performed.
The image processing result is displayed on the TV monitor 6 as appropriate.

Prior to performing such a full-scale visual inspection, it is necessary to roughly specify the position of the semiconductor package 1 to be inspected (the position of each ball 2) in order to reduce the inspection time. This processing is also executed by the image processing computer 5. This is because the XYZ stage 8 moves the semiconductor package 1 to be inspected to the upper camera 3
XYZ stage 8 when transporting and positioning just below
This is because the semiconductor package 1 is not always conveyed to the same position and positioned every time due to the accuracy of the tray 7, the accuracy of the tray 7, the gap between the semiconductor package 1 and the tray 7, and the like.

FIGS. 2 to 5 are flowcharts showing processing contents for roughly specifying the position of the semiconductor package 1 to be inspected (the position of each ball 2). Hereinafter, description will be made with reference to FIGS.

Processing based on the image picked up by the upper camera 3 (FIGS. 2 and 3) This corresponds to the semiconductor package 1 to be inspected.
Is a process based on the imaging result of the upper camera 3 when the captured image is not deformed even when rotated as shown in FIG.

Processing for Finding Center Position C of Semiconductor Package 1 That is, as shown in FIG. 6, when the upper surface of the semiconductor package 1 is imaged by the upper camera 3 and the captured image 10 is captured (step 101), Captured image 10
In the position where the semiconductor package 1 is assumed to exist,
An inspection area 11 including the entire semiconductor package 1 is set. The size of the inspection area 11 is determined in consideration of a positioning error of the transport tray 7 and the like. Specifically, the x axis is set in the horizontal (width) direction of the drawing, and the top and bottom (height) of the drawing
When the y-axis is set in the direction, as shown in FIG. 7A, an inspection area in which the left end position in the width direction is xa1, the right end position in the width direction is xa2, the upper end position in the height direction is ya1, and the lower end position in the height direction is ya2. 1
1a is set. That is, the inspection area 11a whose width is defined by xa2-xa1 and its height is defined by ya2-ya1 is set (step 102).

Next, blob detection is performed on the inspection area 11a. In other words, a detection area of a ball arrangement candidate in which the outer edge portion of the brightness corresponding to the plurality of ball groups 2, 2,... Finally, the detection area of the ball arrangement candidate is narrowed down, and as shown in FIG.
The detection area 12 circumscribing the plurality of ball groups 2, 2,... (Ball arrangement) arranged in a lattice is cut out (steps 104 to 118).

More specifically, first, as shown in FIG. 7A, a detection area 12b having a width defined by xb2-xb1 and a height defined by yb2-yb1 is cut out (step 103).

The width xb2-x of the detection area 12b
It is determined whether or not b1 matches the width of the known ball arrangement (step 104).

Similarly, the height yb2−
It is determined whether or not yb1 matches the height of the known ball arrangement (step 111).

As a result, the width xb2-xb1 of the detection area 12b is obtained.
However, if it is determined that the detected area does not match the width of the known ball array, the detection area is determined to not accurately circumscribe the width direction of the ball array, and the detection area is corrected in the width direction. The process is executed (steps 105 to 110), and the height yb2-yb1 of the detection area 12b is set.
However, if it is determined that the detection area does not match the height of the known ball array, the detection area is determined to not accurately circumscribe the ball array in the height direction, and the detection area is determined in the height direction. Correction processing is executed (step 11
2 to step 117).

For example, as shown in FIG. 8, when a foreign object 13 to be detected such as a pattern on a substrate or dust is present at the time of detecting a blob, the detection region 1 is circumscribed.
This is because 2 may be cut out and may not circumscribe the ball arrangement that should originally circumscribe.

Hereinafter, the width xb2-xb1 of the detection area 12b is calculated as follows.
A case where it is determined that the width does not match the width of the known ball arrangement will be described as a representative.

In this case, the procedure shifts to step 105, where xc1 = xb1 + dx xc2 = xb2-dx yc2 = yb1 yc2 = yb2 (1), the width is xc2-xc1, the height is yc2-yc1, A specified width re-inspection area 11c is set in which the width of the detection area 12 is reduced by a certain amount dx (see FIG. 7A). Here, the fixed amount dx is equal to the above-mentioned object 1 to be detected.
3 is slightly larger than the diameter of the ball 2 and is set to about half of the ball 2. This is because the diameter of the detection object 13 is usually smaller than half of the ball 2 (step 105). And this width re-inspection area 11
Blob detection similar to that in step 103 is performed on c, and the detection area is narrowed down. In this way, as shown in FIG. 7A, a width re-detection area 12d whose width is defined by xd2-xd1 and whose height is defined by yd2-yd1 is cut out (step 1).
06).

Next, it is determined whether or not the contour line in the width direction narrowed down in the width re-detection area 12d matches the contour line in the width re-inspection area 11c.

Specifically, it is determined whether or not xd1−xb1 ≦ dx (2) xb2−xd2 ≦ dx (3) (step 107,
Step 108).

If the above equation (2) or (3) is not satisfied, the outer line of the first detection area 12b is assumed to be inaccurate, and xa1 = xd1. (4) xa2 = xd2 (5), and processing for correcting the outer line of the first inspection area 11a is executed (steps 108 and 110).

Specifically, as shown in FIG. 9, the position x of the narrowed outer line in the re-detection area 12 '(12d)
If the distance between d2 and the position xb2 of the outer line of the first detection area 12 (12b) matches the fixed amount dx (when the expression (3) is satisfied), the outer line of the detection area As the detection area 12 (12
Although the position xb2 of b) is not corrected as being correct (determination "Yes" in step 109), the re-detection area 12 '
When the distance between the narrowed outer line position xd1 of (12d) and the outer line position xb1 of the first detection area 12 (12b) is larger than the fixed amount dx (when the expression (2) is not satisfied) ) Is the outer line of the detection area,
The position xb2 of the detection region 12 (12b) cut out first
Is incorrect (circumscribes the object 13 to be detected), and the position xd1 of the re-detection area 12 '(12d) is corrected (estimated) and corrected (Equation (4)).

Although the narrowing down of the detection area in the width direction has been described, the same processing is executed for the narrowing down of the detection area in the height direction (steps 112 to 11).
7).

In the next step 118, it is determined whether or not the detection area has been corrected. That is, as described above, as shown in the position xd1 of FIG. 9, the outer line of the detection area is not the outer line of the first detection area 12b, but the position xd2 of the narrowed outer line of the re-detection area 12d is correct. If there are any modifications,
The determination at 18 is "No", and as shown in Steps 108, 110, 115 and 117, the contour line position narrowed down in the re-detection area 12d is set to the contour line position of the first inspection area 11a. After the correction, the process of narrowing down the detection area is executed again (steps 103 to 117).

As a result, the detection area finally converges to an area circumscribing the ball arrangement (step 10).
(Yes in step 4 and Yes in step 111)
In addition, the determination in step 118 is “Yes”.

Therefore, as shown in FIG. 6, the coordinate position (xp,
yp) is obtained based on xp = (xb1 + xb2) / 2 yp = (yb1 + yb2) / 2 (6) and the finally obtained outer line positions xb1, xb2, yb1, and yb2 of the detection area 12b (step). 119).

As described above, according to the present embodiment, the ball arrangements 2, 2 of the actual semiconductor package 1 to be inspected are
The detection area 12 is cut out in units of..., And the center position C of the semiconductor package 1 is determined based on the detection area 12. Therefore, there is a defect that the ball 2 of the ball arrangements 2, 2,. Even if the center position C of the work 1 is accurately determined without being affected by the partial defect,
Can be detected.

Process for Determining Attitude Angle φ of Semiconductor Package 1 Once the center position C of the semiconductor package 1 to be inspected is determined as described above, the process for determining the attitude angle, that is, the inclination, is shown in FIG. Is performed as follows.

That is, as shown in FIG. 10A, the detection area 12 circumscribing the cut ball arrangements 2, 2,...
Alternatively, a search area 15 for detecting the inclination of the semiconductor package 1 is created for the area 14 corresponding to this. This search area 15 is cut out from the image so that a row of balls 2, 2,... (Only) is included therein. Since there is a high possibility that the balls 2, 2,... In one row exist around the piece 12a of the outer line of the detection area 12,
A search area 15 is set around this position. Search area 1
5 is slightly larger than the length of one piece 12a of the outer line of the detection area 12, and the width thereof is as shown in FIG.
The pitch is set to about twice the pitch of the ball arrangement in consideration of the case where the ball arrangement is inclined as shown in FIG.
.. Can be reliably included in the search area 15 (Step 201).

The search for each of the balls 2 in a row is performed, for example, by pattern matching using a correlation coefficient (step 202), and the position (x, y) of each ball 2 in the row is detected. Is performed (step 203).

Then, based on the positions (x, y) of the balls in the row searched, a straight line 16 connecting these positions is provided.
(See FIG. 10A), for example, is obtained as a least-squares straight line (regression line), and the slope of this straight line is calculated (step 204). Then, the inclination of this straight line is defined as the inclination φ of the semiconductor package 1 (step 205).

As described above, the search area 15 is cut out in units of one row of balls 2, 2,... Of the actual semiconductor package 1 to be inspected, and the attitude angle (tilt) φ of the semiconductor package 1 is detected based on this. So that
As shown by Em in FIG. 10B, even if there is a defect that some of the balls in the row are missing or greatly deformed, they are affected by these defects. And the inclination Δy / Δx of the straight line indicating the ball row based on the positions of the other non-defective balls 2a and 2b.
And the attitude angle φ of the semiconductor package 1 can be detected.

Although the search area 15 is set in the vicinity of the piece 12a of the outer line of the detection area 12, the search area 15 may be set in the vicinity of the center of the ball array.
For example, as shown in FIG. 10A, in the case of a ball array having four rows (even number) in the vertical direction, it is known in advance that the ball row exists at a position half a pitch above the center 12a of the inspection area 12. Therefore, the center 1 of the search area 15 is located at this position.
What is necessary is just to set the search area 15 so that 5a may correspond.

As described above, the center position of the ball array and the inclination of the ball row are directly obtained as the center position C and the inclination φ of the semiconductor package 1.

If the ball arrangement above the semiconductor package 1 is offset or has a relative inclination, the known relative position data of the ball arrangements 2, 2,. The center position C and the inclination φ of the semiconductor package 1 can be obtained based on the relative attitude angle data, the center position of the ball array detected as described above, and the inclination of the ball row.

In the present embodiment shown in FIG. 10A, the detection area 12 (or the area 1 corresponding thereto) is used.
.. In 4) are arranged in a grid pattern at an equal pitch. However, depending on the type of the semiconductor package 1, as shown in FIG. 23, the ball arrangement in the detection area 12 (or the area 14 corresponding thereto) is not at a constant pitch. In some cases, 2 is missing.

If the information that the ball arrangement is not equal pitch is known in advance, the search area 15 'is set to a position where the row of balls 2, 2,... Can be set.

For example, with respect to the detection area 12 in which the balls 2 in some central rows shown in FIG. 23 are missing, “the balls are missing in the second to fourth rows from the left, Column (first column from the left) and the rightmost column (fifth column from the left) ", the search in the column direction surely includes the balls 2, 2,. The area 15 'can be set, and the straight line 16 indicating a row of balls can be obtained accurately. This prevents the search area 15 ″ in the row direction where the ball is missing from being set by mistake.

Processing based on the image captured by the oblique camera 4 (FIGS. 4 and 5) This is because when the semiconductor package 1 to be inspected rotates as shown in FIG. 21, the image captured as shown in FIG. This processing is performed when (the outer shape of the semiconductor package in the captured image) is deformed, and the processing contents are shown in FIGS.

That is, as shown in FIGS. 11 and 12,
The oblique camera 4 captures an image of the upper surface of the semiconductor package 1 to be inspected from an oblique direction, and captures the captured image 17. Then, from the captured image 17, the detection regions 18A and 18B (20C,
20D) is cut out (step 301).

Then, in each of the x-axis direction and the y-axis direction of the extracted detection area, the position x (or y) on the image and the brightness of all the pixels 22A corresponding to the position x (or y) are determined. The added value or the relationship with the evaluated value Px is obtained as shown in FIG.

More specifically, as for the above relationship in the x-axis direction, as shown in FIG. 11, the detection areas 18A and 18B are divided into an upper area 18A and a lower area 18B in the y-axis direction. , 18B (step 3
05, step 306). Then, as for the above relationship in the y-axis direction, as shown in FIG.
D is divided into a left region 20C and a right region 20D in the x-axis direction, and is obtained for each of the divided regions 20C and 20D (steps 302 and 303).

The left area 20C, the right area 20D, and the upper area 18
A, since the processing performed on the lower area 18B is the same as shown in FIG. 5, the processing on the lower area 18B will be described below as a representative.

That is, when the process proceeds to step 306 in FIG. 4, the procedure shifts to the ball row detection subroutine shown in FIG.

Then, in step 401 of FIG.
As shown in FIG. 13A, for each position x of the lower region 18B, a process of adding the brightness of the pixels 22 corresponding to the position x, that is, all the pixels 22A in one row in the y-axis direction is executed.

As described above, the relationship between the position x on the image and the added value (hereinafter referred to as pixel value) Px of the brightness of all the pixels 22A in one row in the y-axis direction corresponding to this position x is shown in FIG.
3 (b). Instead of the brightness value, an equivalent value, for example, an average value of the brightness value may be used.

Then, as shown in FIG. 13B, the peak value 23 of the pixel value Px is extracted. That is, the peak values 231, 232, 233, 234,
235 and 236 are obtained, and these peak values 231 and 2
The positions x1, x2, x3, x4, x5, and x6 on the x-axis corresponding to 32, 233, 234, 235, and 236 correspond to the lower region 18B.
Is the array candidate position of the ball 2 in the x-axis direction. The peak value can be detected by differentiating the curve shown in FIG. 13B (step 402).

Thus, the lower region 18 shown in FIG.
When the upper and lower balls 2 and 2 in B are regarded as a single ball 19, an arrangement position of the single ball 19 in the x-axis direction (hereinafter, referred to as a ball row: indicated by an arrow) is detected.

However, the ball row obtained at this stage is:
It does not necessarily indicate only the ball row on the semiconductor package 1 to be inspected. This is because, for example, the ball 2 on another semiconductor package 1 (mounted on the same tray 7) adjacent to the inspection target may be erroneously detected. In addition, even for the semiconductor package 1 to be inspected, dust or wiring patterns on the semiconductor package 1 or a part (edge) of the tray 7 for transport may be captured as a peak value. is there. Further, since the ball 2 is imaged obliquely, as shown in FIG. 16, the top portion 2 'of the ball 2 and the pad 30 supporting the ball 2 (see FIG. 16B) are displayed on the captured image. This is because the peak value may be increased from the actual value by being captured as two separate images (see FIG. 16A).

Then, the peak-to-peak distance is obtained, the mode d of the peak-to-peak distance is obtained (step 403), and the mode d of the peak-to-peak distance and the other peak-to-peak distances d 'and d ".
Are compared with each other, and when the difference is equal to or greater than a certain threshold value, it is determined that the peak belongs to another group. For example, in the case of FIG. 14, peaks whose peak-to-peak distance is d (mode) are grouped into one group, and peaks 23 'and 2' whose peak-to-peak distances are d 'and d "
3 "is determined to belong to another group (step 404).

Thus, each peak 23 is classified into each group (step 405), and for each group, the number of peaks included in the group is compared with the number of known ball rows of the semiconductor package 1 to be inspected. (Step 40
6).

As a result, when the number of peaks included in the group matches the actual number of ball rows, the peak obtained in step 402 accurately captures the ball row on the semiconductor package 1 to be inspected. It is determined that the peak positions x1, x2, x3... Shown in FIG. 13B are the ball row positions 23B (see FIG. 17A) (step 417).

However, the number of peaks included in the group is
If the number of ball rows is one or two more than the actual number of ball rows, the peak obtained in step 402 is regarded as not accurately capturing the ball row on the semiconductor package 1 to be inspected, and is included in this group. With respect to the included peaks at both ends, a process of confirming whether or not the peak is the ball 2 is executed. That is, as shown in FIG.
For the actual number of 4 (4 rows) ball rows,
Assuming that a group of five peaks 23 is detected, pattern matching is performed on the regions 24a and 24b corresponding to the peaks 23a and 23b at both ends of the group by using the shape of the ball 2 as a model. It is determined whether or not 23a and 23b are balls 2 (step 407). This determination is performed for both ends of the group (step 41).
0).

As a result of the pattern matching, if the ball 2 is found at each end of the group (determination of “Yes” in steps 408 and 411), the corresponding peak is included in the number of peaks in the group. In addition to counting, if the ball 2 is not found at each end of the group (decision “NO” in step 408 and step 411), the corresponding peak is not counted in the number of peaks in the group, Execute processing to delete from the group (step 4
09, step 412).

Finally, the number of peaks in the group is determined (step 413). If the number of peaks included in this group matches the actual number of ball rows, the peak obtained in step 413 is Assuming that the ball row on the semiconductor package 1 to be inspected is accurately captured,
The positions of these peaks on the x-axis are represented by the ball row position 23B.
(See FIG. 17A). For example, in FIG. 15, the peak 23a at one end is deleted,
When the peak 23b at the other end is counted as a peak in the group as indicating the ball 2, the number of peaks (4) in the group matches the actual number of ball rows (4), and the peak 23b is determined. The positions on the x-axis of the peaks except for are set as the ball row position 23B (step 417).

However, the number of peaks included in the group is
If the number does not match the actual number of ball rows, the peak obtained in step 413 is accurately displayed on the semiconductor package 1 to be inspected unless another group exists (determination “No” in step 414). The processing is terminated assuming that the ball row has not been captured and that the detection of the ball row has failed (step 415). Step 40
In step 6, even if the number of peaks included in the group is smaller than the actual number of ball rows, the detection of the ball row fails unless another group exists (determination "No" in step 414). And the process is terminated (step 415).

If another group exists (determination "Yes" in step 414), the same processing (steps 406 to 415) is executed for this group (step 416).

As described above, the ball row 23B is obtained for the lower area 18B as shown in FIG. 17A (step 306), and similarly, the ball row 23A is obtained for the upper area 18A (step 305). ).

As a result, these upper and lower ball rows 23A,
A straight line 26 connecting 23B is obtained.
7).

Similarly, a ball row 23C is obtained for the left area 20C (step 302), and the right area 20D is obtained.
The ball row 23D is obtained for (step 30)
3) Therefore, a straight line 25 connecting these left and right ball rows 23C and 23D can be obtained (step 304).

The intersection of these straight lines 25 and 26 is the actual candidate position of the ball 2 of the semiconductor package 1.

However, actually, the balls 2 on the semiconductor package 1 do not exist on all the intersections, but have a certain arrangement pattern as shown in FIG. Exists only in.

Therefore, as shown in FIG. 17C, data indicating a known ball arrangement pattern (FIG. 17B)
And the intersections of the straight lines 25 and 26 are matched to extract only the intersections indicated by the ball arrangement pattern, and each of the extracted intersections is the position (approximate position) of each ball 2 to be inspected.
(X, y) is determined (step 309).

In the present embodiment, the detection area is
The reason why the ball array is divided into upper and lower or left and right regions and a ball row is detected for each divided region is as follows. That is, as shown in FIG. 18, when trying to detect a ball row for the entire detection area 18A, 18B, since the detection areas 18A, 18B are inclined with respect to the semiconductor package 1 on the captured image, the ball row is This is because the ball 2A included and the ball 2B included in another row adjacent to the ball 2A may be erroneously detected as a single ball 19 at the same position x.

However, if such a problem does not occur,
The detection of the ball row may be performed in each of the x-axis direction and the y-axis direction without dividing the detection area.

The detection area is divided into two parts vertically and two parts left and right.
Instead of dividing the entire divided area as a ball row detection target, only the divided area, for example, about / of a portion near the outer line in the lower area 18B may be the ball row detection target.

If the detection of the ball row fails, the ball row detection target (for example, the lower area
The area (about / of B) may be further narrowed by one pitch, and the detection of the ball row may be retried again.

In the present embodiment shown in FIGS. 11 and 12, it is assumed that the ball rows 2, 2,... In each of the detection areas 18A, 18B, 20C, 20D are arranged at an equal pitch. However, as described in FIG.
Depending on the type of the semiconductor package 1, as shown in FIG. 24, for example, the ball rows in the detection area 20C (left area) are not at equal pitches, but are ball rows in which some middle balls 2 are missing. There is.

When the arrangement of the ball rows is not equal pitch, the mode d of the peak-to-peak distance is compared with the other peak-to-peak distances d ′ and d ″, and the difference is equal to or larger than a certain threshold value. Then, the processing based on the assumption that the ball rows have the same pitch, such as the processing of determining that the peak belongs to another group (step 404), cannot be applied as it is.

Furthermore, if the processing of FIG. 5 is applied to the semiconductor package 1 having such a non-equal pitch ball row, it is expected that grouping after peak detection will often fail. The reason is that a plurality of groups exist in one semiconductor package 1 in many cases. Also, because the distance between adjacent balls in the ball row is large, it is difficult to distinguish the adjacent semiconductor package 1 that is not the inspection target, distinguish it from the end of the tray 7, and remove dust from the pattern.

Therefore, in the case of the semiconductor package 1 having a ball row having a non-equal pitch, it is necessary to perform the special processing shown in FIG. 26 instead of applying the processing shown in FIG. 5 as it is.

In the processing shown in FIG. 26, when it is known in advance that the ball rows are not at the same pitch, a true ball row is searched for based on information on the number of balls in the ball row and the distance between each ball. Things.

Hereinafter, the flowchart shown in FIG. 26 will be described with reference to FIGS. 25 (a), 25 (b) and 25 (c). In this embodiment, a left area 20C shown in FIG. 24 is assumed as a detection area in which the ball rows are not at the same pitch.

In steps 501 and 502 in FIG. 26, the same processing as in steps 401 and 402 shown in FIG. 5 is executed. That is, y in the left area 20C shown in FIG.
By detecting each peak value in the axial direction, the arrangement candidate position of the ball 2 in the y-axis direction in the left area 20C is detected (Step 501, Step 502).

However, the position of the ball row obtained at this stage is an array candidate position, and does not necessarily indicate only the ball on the semiconductor package 1 to be inspected. this is,
For example, another semiconductor package 1 adjacent to the inspection target
This is because there is a possibility that the ball 2 on the same tray 7 is erroneously detected. Further, even for the same semiconductor package 1 to be inspected, dust and wiring patterns on the semiconductor package 1 and a transfer tray 7
In some cases (edge) may be caught as a peak value.
(Ball 2 of semiconductor package 1 on adjacent tray 7
(Steps 504 to 508) are repeatedly executed for each peak (step 503).

That is, as shown in FIG. 24, a search area 30 large enough to include the ball 2 around the peak position 23c obtained in step 502 is set for each peak (step 504). Each search area 30
Are subjected to pattern matching using the shape of the ball 2 as a model, and it is determined whether or not each peak is the ball 2 (steps 505 and 506).

For example, in FIG.
When pattern matching is performed on the search area 30 'set around' c, the search area 30 '
It is determined that the ball 23 does not include the ball 2 and the peak 23'c does not indicate the ball 2 (determination "No" in step 506). As a result, it is determined that the peak 23'c does not represent a ball row, and is excluded from ball row candidates (step 507). On the other hand, when it is determined that the peak 23c indicates the ball 2 as a result of the pattern matching (determination “Yes” in step 506), it is determined that the peak 23c represents the ball row,
Leave as a ball row candidate (step 50
8).

Thus, combinations of peaks finally left as ball row candidates are set, and it is repeatedly determined whether or not each combination represents a true ball row. (Step 509).

Here, as shown in FIG. 25 (b), the number of balls in the true ball row (four) and the distances (d1, d2, d3) between adjacent balls in the true ball row
Is stored in advance as known information.

Thus, as shown in FIG. 25A, combinations of peaks (1), (2), (3), and (4) corresponding to the number of balls (four) in this true ball row are used. Is set (step 510).

Next, the distance between adjacent balls is calculated for each combination. For example, for the combination (2), as shown in FIG. 25C, the distance between adjacent balls is calculated as d1 ', d2', and d3 '(step 51).
1).

Next, the calculated inter-ball distance di '(i = 1, 2, 3) is compared with the inter-ball distance di (i = 1, 2, 3) in a true ball row.

Specifically, calculation, Thus, the sum of squares Δ of the difference between the distances between the balls is obtained (step 512).

For each of the combinations (1) to (4), that is, for each ball row candidate, the sum of squares Δ of the difference between the balls is obtained (step 513).

Then, the sum of squares Δ of the difference in the distance between the balls obtained for each ball row candidate is compared, and the combination having the smallest value is determined to be the true ball row. FIG.
In the case of (a), it is determined that the combination (3) is a true ball row, that is, a ball row to be inspected (step 5).
14).

As described above, according to the present embodiment, the x-axis,
A ball row (ball row position) is acquired for each y-axis, and the obtained ball row (ball row position) 23 in the x-axis direction is obtained.
A, 23B and ball row (ball row position) 2 in the y-axis direction
3C, 23D, the x, y coordinate position (x, y) of each ball 2, 2,... Is detected.
The semiconductor package 1 is rotated by the YZ stage 8 and imaged by the oblique camera 4 at each rotational position. Even if the semiconductor package 1 in the captured image is deformed and has a different shape for each rotational position, it is affected by these. Without this, the position of each ball 2 can be quickly and accurately specified.

As described above, the center position C and the attitude angle φ of the semiconductor package 1 to be inspected are obtained based on the imaging result of the upper camera 3, and the x of each ball 2 is obtained based on the imaging result of the oblique camera 4. , Y positions (x, y) are obtained, the reference positions of the semiconductor package 1 and the respective balls 1 on the semiconductor package 1 which are necessary for the appearance inspection of the semiconductor package 1 to be inspected are determined based on these. .

Therefore, an appearance inspection of the height of each ball 2 is performed with reference to the reference position of the semiconductor package 1 thus determined and the reference position of each ball 2 on the semiconductor package 1.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an apparatus configuration according to an embodiment of the present invention.

FIG. 2 is a flowchart showing a processing procedure executed by the image processing computer shown in FIG. 1, which is executed based on an imaging result of an upper camera.

FIG. 3 is a flowchart showing a processing procedure executed by the image processing computer shown in FIG. 1, which is executed based on an imaging result of an upper camera.

FIG. 4 is a flowchart showing a processing procedure executed by the image processing computer shown in FIG. 1, which is executed based on an imaging result of an oblique camera.

FIG. 5 is a flowchart showing a processing procedure executed by the image processing computer shown in FIG. 1, which is executed based on an image pickup result of an oblique camera.

FIG. 6 is a diagram used to explain the processing contents shown in FIG. 2;

FIGS. 7A and 7B are diagrams used to explain the processing contents shown in FIG. 2;

FIG. 8 is a diagram used to explain the processing contents shown in FIG. 2;

FIG. 9 is a diagram used to explain the processing contents shown in FIG. 2;

FIGS. 10A and 10B are diagrams used to explain the processing contents shown in FIG. 3;

FIG. 11 is a diagram used to explain the processing contents shown in FIG. 4;

FIG. 12 is a diagram used to explain the processing contents shown in FIG. 4;

FIGS. 13A and 13B are views used to explain the processing contents shown in FIG. 5;

FIG. 14 is a diagram used to explain the processing contents shown in FIG. 5;

FIG. 15 is a diagram used to explain the processing contents shown in FIG. 5;

FIGS. 16A and 16B are diagrams used to explain the processing contents shown in FIG. 5;

FIGS. 17A, 17B, and 17C are diagrams used to explain the processing contents shown in FIG. 4;

FIG. 18 is a diagram used to explain the processing contents shown in FIG. 4;

FIGS. 19A and 19B are diagrams used to explain a conventional technique.

FIGS. 20 (a) and 20 (b) are diagrams used for explaining a conventional technique.

FIG. 21 is a diagram used for explaining a conventional technique.

FIGS. 22 (a) and 22 (b) are diagrams used to explain a conventional technique.

FIG. 23 is a diagram for explaining processing when a ball array having a non-equal pitch is handled;

FIG. 24 is a diagram for explaining processing when a ball array having a non-equal pitch is handled;

FIGS. 25 (a), (b) and (c) are views for explaining processing when a ball array having a non-equal pitch is handled.

FIG. 26 is a flowchart illustrating a procedure of processing when a ball array having a non-equal pitch is handled.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Semiconductor package 2 Ball 3 Upper camera 4 Oblique camera 5 Image processing computer

Claims (20)

[Claims] 1. A work position detection method for detecting, by image processing, the position of a work in which a plurality of members of a fixed size are arranged on a top surface of a work according to a predetermined rule, wherein an imaging process of imaging the top surface of the work is performed. A first image processing step of cutting out a first detection region circumscribing the plurality of array members from an image captured in the imaging step; and a plurality of image processing steps cut out in the first image processing step. From the first detection area showing the array members, the second showing the row members
A second image processing step of cutting out the detection region of the above, known relative position data of the plurality of array members with respect to the workpiece, and a first detection indicating the plurality of array members cut out in the first image processing step The absolute center position of the work is detected by matching the area, and the known relative position data is matched with a second detection area indicating a row of members cut out in the second image processing step. A work position detection method comprising a work position and a posture detection step for detecting the posture of the work. 2. The method according to claim 1, wherein the work is a semiconductor package in which ball-shaped solder bumps are arranged in a grid on the upper surface. 3. The first image processing step includes: an inspection area setting step of cutting out an inspection area including the entire work from an image captured by the imaging step; and an inspection area setting step of cutting out the inspection area cut out by the inspection area setting step. From the inside, a detection area narrowing step of cutting out the outer edge portions of the brightness corresponding to the plurality of array members as a detection area of an array member candidate to be an outer line, and an array member candidate detection area cut out in the detection area narrowing step A determining step of determining whether the length in the X-axis direction and the length in the Y-axis direction match the known lengths of the array member in the X-axis direction and the known length in the Y-axis direction; When it is determined that the length of the array member candidate detection area does not match the known length of the array member, the outer line is set in the X-axis and Y-axis directions that do not match. A re-inspection area setting step of setting a re-inspection area shifted by a small distance in a direction in which the arrangement member candidate detection area is smaller than the width of the member; and the re-inspection area set in the re-inspection area setting step. A re-detection step of cutting out a re-detection area by performing the same processing as the detection area narrowing step, and when an outer line of the re-detection area cut out in the re-detection step does not match the outer line of the re-inspection area. Performs the same process as the above-described determination process on the array member candidate detection region corrected in the correction process, using the outer line of the non-matched re-detection region as the outer line of the array member candidate detection region. If the length of the array member candidate detection area is finally determined to be equal to the known length of the array member by the determination process, 2. The method according to claim 1, further comprising the step of determining the supplementary detection area as an area circumscribing the plurality of array members. 4. A work position detection method for detecting, by image processing, the position of a work in which a plurality of members of a predetermined size are arranged in a lattice on an upper surface of the work, wherein: A detection area setting step of cutting out a detection area including the entire work from the image captured by the imaging step; and an X-axis direction and a Y-axis direction of the detection area cut out in the detection area setting step. And a calculation step for finding the relationship between the brightness value of all pixels corresponding to this position and the value obtained by evaluating the addition value, and evaluating the brightness addition value or the addition value in the relationship obtained in the calculation step. A peak value extraction step of extracting a peak value of the calculated value, and each position in the X-axis direction corresponding to each peak value in the X-axis direction extracted in the peak value extraction step. X of the member
The position in the Y-axis direction corresponding to each peak value in the Y-axis direction extracted in the peak value extraction step is defined as the arrangement position in the Y-axis direction of the member. An X-axis direction member arrangement position obtained in the arrangement position detection step, and an X-axis direction member arrangement position obtained in the above-described arrangement position detection step. A workpiece position detecting method for detecting a workpiece position. 5. The work position detecting method according to claim 4, wherein the work is a semiconductor package in which ball-shaped solder bumps are arranged in a grid on the upper surface. 6. The calculation step includes, for the relationship in the X-axis direction, dividing the detection area into a plurality of pieces in the Y-axis direction and obtaining each of the divided areas, and regarding the relationship in the Y-axis direction, The detection area is divided into a plurality of parts in the X-axis direction, and is obtained for each of these divided areas. The member position detection process is performed by using X obtained for each of the divided areas.
A straight line indicating the member arrangement in the Y-axis direction is generated based on the member arrangement position in the axial direction, and a straight line indicating the member arrangement in the X-axis direction is generated based on the member arrangement position in the Y-axis direction obtained for each divided region. 5. The work position detecting method according to claim 4, wherein the X and Y positions of the plurality of array members are detected as intersections of these straight lines. 7. The arrangement position detection step obtains an interval between positions in the X-axis direction corresponding to each peak value in the X-axis direction extracted in the peak value extraction step, and calculates an interval according to the size of the interval. Determining the ends of the plurality of array members on the work to be inspected in the X-axis direction, and determining the intervals between the respective positions in the Y-axis direction corresponding to the respective peak values in the Y-axis direction extracted in the peak value extraction step. 5. The work position detection method according to claim 4, wherein the end of the plurality of array members on the work to be inspected in the Y-axis direction is determined according to the size of the interval. 8. The arrangement position detecting step calculates the number of peak values in the X-axis direction extracted in the peak value extracting step, and determines the number of peak values in the X-axis direction in the X-axis direction of the workpiece. If it is different from the number of parts,
Pattern matching is performed on the image corresponding to the end in the X-axis direction, and it is determined whether or not the end in the X-axis direction is the member, and each peak in the Y-axis direction extracted in the peak value extraction process is determined. Determine the number of values, if the number of peak values in the Y-axis direction is different from the number of array members in the Y-axis direction of the work, perform pattern matching on the image corresponding to the end in the Y-axis direction, 5. The method according to claim 4, wherein it is determined whether or not an end in the Y-axis direction is the member. 9. The position of a work in which a plurality of members of a fixed size are arranged in a lattice on the upper surface of the work and the position of each member on the work are detected in advance by image processing, and the detected position is determined. In a work inspection method for inspecting each member on the work as a reference position of the work and a reference position of each member, a first imaging process of directly imaging an upper surface of the work, and an imaging process in the first imaging process A first detection region for cutting out a first detection region circumscribing the plurality of array members from the image thus obtained;
An image processing step; and a second detection area indicating a row of members from a first detection area indicating a plurality of array members cut out in the first image processing step.
A second image processing step of cutting out the detection region of the above, known relative position data of the plurality of array members with respect to the workpiece, and a first detection indicating the plurality of array members cut out in the first image processing step The absolute center position of the work is detected by matching the area, and the known relative position data is matched with a second detection area indicating a row of members cut out in the second image processing step. A work center position for detecting the posture of the work, a posture detection process, a second imaging process for obliquely imaging the upper surface of the work, and an entire work from an image captured by the second imaging process. A detection area setting step of cutting out the oblique detection area including the following, and X of the oblique detection area cut out in the detection area setting step.
In each of the axial direction and the Y-axis direction, an operation step for obtaining a relationship between a position on the image and an added value of the brightness of all pixels corresponding to this position or a value obtained by evaluating the added value; A peak value extraction process for extracting a peak value of a value obtained by evaluating the lightness value or the value obtained by evaluating the sum value in the relationship, and an X-axis direction corresponding to each peak value in the X-axis direction extracted in the peak value extraction process. Set the position to X
The position in the Y-axis direction corresponding to each peak value in the Y-axis direction extracted in the peak value extraction step is defined as the arrangement position in the Y-axis direction of the member. An array position detecting step to be performed, an array position of members in the X-axis direction obtained in the array position detecting step, and an array position of members in the Y-axis direction. A member position detection process for detecting a position, the work center position, the work center position and posture obtained in the posture detection process, and the X, Y positions of the plurality of members obtained in the member position detection process. A reference setting step of setting a reference position of the work and a reference position of each member on the work based on the reference position of the work and a reference position of each member on the work set in the reference setting step. Inspecting each member on the work according to a standard. 10. The work according to claim 9, wherein the work is a semiconductor package in which ball-shaped solder bumps are arranged in a grid on the upper surface, and the inspection of the work is an inspection of the solder dump for defects. Inspection method. 11. A work position detecting device for detecting, by image processing, a position of a work in which a plurality of members of a fixed size are arranged on a top surface of the work according to a predetermined rule, wherein: an imaging unit for imaging the top surface of the work A first image processing unit that cuts out a detection region circumscribing the plurality of array members from an image captured by the imaging unit; and a plurality of array members cut out by the first image processing unit. From the first detection region shown, a second member showing a row of members is shown.
A second image processing unit that cuts out the detection region of the above, a known relative position data of the plurality of arrangement members with respect to the workpiece, and a first detection indicating the plurality of arrangement members cut out by the first image processing unit The absolute center position of the workpiece is detected by matching the area, and the known relative position data is matched with a second detection area indicating a row of members cut out by the second image processing means. A work position detecting device comprising a work position and posture detecting means for detecting the posture of the work. 12. The work position detecting device according to claim 11, wherein the work is a semiconductor package in which ball-shaped solder bumps are arranged in a grid on the upper surface. 13. An inspection area setting means for cutting out an inspection area including an entire workpiece from an image taken by the imaging means, wherein the first image processing means comprises: From the inside, a detection area narrowing unit that cuts out the outer edge portions of the brightness corresponding to the plurality of array members as a detection region of an array member candidate to be an outer line, and an array member candidate detection region cut out by the detection region narrowing unit. Determining means for determining whether the length in the X-axis direction and the length in the Y-axis direction match the known lengths in the X-axis direction and the Y-axis direction of the array member; When it is determined that the length of the array member candidate detection area does not match the known length of the array member, the outer line is determined in the X axis and the Y axis that do not match. A re-inspection area setting means for setting a re-inspection area shifted by a small distance in a direction in which the arrangement member candidate detection area is smaller than the width of the member, and the re-inspection area set by the re-inspection area setting means A re-detection unit that cuts out a re-detection region by performing the same processing as the detection region narrowing-down unit, and when an outer line of the re-detection region cut out by the re-detection unit does not match an outer line of the re-inspection region. Performs the same processing as the determining means on the array member candidate detection area corrected by the correcting means, using the outer line of the non-matched re-detection area as the outer line of the array member candidate detection area. If the determining unit finally determines that the length of the array member candidate detection area matches the known length of the array member, 12. The work position detecting device according to claim 11, further comprising means for determining a supplementary detection region as a region circumscribing the plurality of array members. 14. A work position detecting device for detecting, by image processing, the position of a work in which a plurality of members of a predetermined size are arranged in a lattice on the upper surface of the work, wherein: A detection region setting unit that cuts out a detection region including the entire work from an image captured by the imaging unit; and an X-axis direction and a Y-axis direction of the detection region that is cut out by the detection region setting unit. Calculating means for determining the relationship between the position of (i) and the value of the brightness of all the pixels corresponding to the position or the value obtained by evaluating the value of the pixel; and evaluating the value of the brightness or the value of the value of the relationship obtained by the calculating means. Value extracting means for extracting a peak value of the obtained value, and each position in the X-axis direction corresponding to each peak value in the X-axis direction extracted by the peak value extracting means. , X of the member
In addition to the arrangement position in the axial direction, each position in the Y-axis direction corresponding to each peak value in the Y-axis direction extracted by the peak value extraction means is the arrangement position in the Y-axis direction of the member. Array position detecting means, and the X and Y member arrangement positions obtained by the array position detection means are compared with the member arrangement position in the Y axis direction, whereby the X and Y positions of the plurality of arrangement members are compared. And a member position detecting means for detecting the position of the workpiece. 15. The work position detecting device according to claim 14, wherein the work is a semiconductor package in which ball-shaped solder bumps are arranged in a grid on the upper surface. 16. The arithmetic means divides the detection area into a plurality of parts in the Y-axis direction for the relation in the X-axis direction, obtains each of these divided areas, and obtains the relation in the Y-axis direction: The detection region is divided into a plurality of regions in the X-axis direction, and the divided region is determined for each of the divided regions.
A straight line indicating the member arrangement in the Y-axis direction is generated based on the member arrangement position in the axial direction, and a straight line indicating the member arrangement in the X-axis direction is generated based on the member arrangement position in the Y-axis direction obtained for each divided region. 15. The work position detecting device according to claim 14, wherein X and Y positions of the plurality of array members are detected as intersections of these straight lines. 17. The arrangement position detection means obtains an interval between each position in the X-axis direction corresponding to each peak value in the X-axis direction extracted by the peak value extraction means, and according to the size of this interval. Determining the ends of the plurality of array members on the work to be inspected in the X-axis direction, and determining the intervals between the respective positions in the Y-axis direction corresponding to the respective peak values in the Y-axis direction extracted by the peak value extracting means. Is determined, and the Y of a plurality of array members on the work to be inspected is
15. The work position detecting device according to claim 14, wherein the end in the axial direction is determined. 18. The arrangement position detection means obtains the number of each peak value in the X-axis direction extracted by the peak value extraction means, and determines the number of peak values in the X-axis direction of the work in the X-axis direction. If the number is different from the number of members, pattern matching is performed on the image corresponding to the end in the X-axis direction, and it is determined whether or not the end in the X-axis direction is the member, and the peak value extraction is performed. The number of peak values in the Y-axis direction extracted by the means is obtained, and if the number of peak values in the Y-axis direction is different from the number of array members in the Y-axis direction of the workpiece, the end in the Y-axis direction is determined. 15. The work position detecting device according to claim 14, wherein pattern matching is performed on an image corresponding to the part to determine whether or not an end in the Y-axis direction is the member. 19. The position of a work in which a plurality of members of a fixed size are arranged in a lattice on the upper surface of the work and the position of each member on the work are detected in advance by image processing, and the detected position is determined. In a work inspecting apparatus for inspecting each member on the work as a reference position of the work and a reference position of each member, a first image pickup unit for picking up an image of an upper surface of the work from directly above, and an image picked up by the first image pickup unit A first detection region for cutting out a first detection region circumscribing the plurality of array members from the image thus obtained;
An image processing unit, and a second detection unit indicating a row of members from a first detection area indicating a plurality of array members cut out by the first image processing unit.
A second image processing unit that cuts out the detection region of the above, a known relative position data of the plurality of arrangement members with respect to the workpiece, and a first detection indicating the plurality of arrangement members cut out by the first image processing unit The absolute center position of the workpiece is detected by matching the area, and the known relative position data is matched with a second detection area indicating a row of members cut out by the second image processing means. A work center position and posture detection means for detecting the posture of the work, a second imaging means for obliquely imaging the upper surface of the work, and an entire work from an image taken by the second imaging means. A detection area setting unit that cuts out the oblique detection area including the following: X of the oblique detection area cut out by the detection area setting unit
Calculating means for obtaining a relationship between a position on the image and an added value of the brightness of all pixels corresponding to the position or a value obtained by evaluating the added value in each of the axial direction and the Y-axis direction; A peak value extraction unit for extracting a peak value of a value obtained by evaluating the lightness addition value or the value of the addition value in the relationship, and an X-axis direction corresponding to each peak value in the X-axis direction extracted by the peak value extraction unit. Set the position to X
In addition to the arrangement position in the axial direction, each position in the Y-axis direction corresponding to each peak value in the Y-axis direction extracted by the peak value extraction means is the arrangement position in the Y-axis direction of the member. An array position detecting unit that performs alignment of the X and Y members in the X axis direction obtained by the array position detecting unit with an array position of the members in the Y axis direction, thereby obtaining X and Y of each of the plurality of members. A member position detecting means for detecting a position; a work center position and a work center position and posture obtained by the work position detecting means; and X and Y positions of the plurality of members obtained by the member position detecting means. Reference setting means for setting a reference position of the work and a reference position of each member on the work based on the reference position of the work and a reference position of each member on the work set by the reference setting means. Means for inspecting each member on the work according to a standard. 20. The work according to claim 19, wherein the work is a semiconductor package in which ball-shaped solder bumps are arranged in a grid pattern on an upper surface thereof, and the inspection of the work is an inspection of a defect of the solder dump. Inspection equipment.