JPS61153503A - Detecting method of position of work - Google Patents

Abstract

PURPOSE:To easily detect the position of a work even when the ratio of the work size and work interval is large by positioning a corner part of the work almost at the center part of the image pickup field of a television camera when images of plural rectangular works of specific size arrayed almost in a line is picked up by the television camera. CONSTITUTION:An image of one corner part of a pellet 1a is picked up, an XYtheta table is rotated so that two adjacent sides of the corner part is parallel to reference lines on the XYtheta table, and the X and Y directional arrays of the pellet 1a are specified, thereby finding the positions x6 and y5 of the one corner part C. Then, the size (l) of the pellet 1a is already known, so the center positions x6 and y6 of the pellet 1a are found; and the center positions x0 and y0 of the image pickup visual field are already known, so the positions are corrected by (x0-x6) in the X direction and by (y0-y6) in the Y direction, so that the center of the pellet 1a comes onto the takeout position (x0, y0). Then, the position of the pellet 1a is detected to take out the pellet 1a and the table is moved so that a corner part of a pellet expected to be taken out next is almost at the center part of the image pickup field.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention supports, for example, workpiece semiconductor berets and feet arranged at substantially constant intervals in orthogonal directions so as to be movable in the X and Y directions, and the semiconductor beret at the take-out position is The present invention relates to a method of detecting the position of a workpiece by capturing an image with a camera and calculating the center position of the workpiece from the captured image data.

BACKGROUND ART In the process of manufacturing semiconductor devices, a single wafer is divided into individual semiconductor chips and chips, which are taken out one by one and mounted on a substrate.

Here, each of the divided semiconductor chips and chips is slightly misaligned, so if you place them on the board as they are, the electrode positions on the chips and chips will shift by a wide range during wire bonding, and bonding will occur. Since this may not be possible, the position of the pellet is detected when the pellet is taken out, and the position is corrected when the pellet is taken out to minimize misalignment on the board.

Here, a method for detecting the position of the pellet will be explained with reference to FIGS. 4 to 6. In the figure, 1 is a semiconductor bevel and throat, which cuts the semiconductor wafer stuck on a sheet z at predetermined intervals in the orthogonal direction, and a cylindrical holding ring 3 stretches the sheet 2. This is supported by a feeding device (xyθ table) (not shown) and sent in the direction in which the pellets 1 are lined up.
The center of the pellet 1 is detected by taking an image of the pellet 1 to be taken out and calculating the position of the pellet 1. The center of the pellet 1 is detected by the corelid 5 and supplied onto a substrate (not shown).

Incidentally, the image of the foot 1 is taken by the television camera 4 as shown in FIG. That is, even if the feeding device is partially pitch-feeded in order to take out one bere/lid and take out the next bere/y)la, the position of the bevel (y) on the sheet 2 will shift and it will be out of the effective field of view of the camera 4. Since there is a tendency for the berets to come off, the magnification is determined so that a portion of the berets 1b, ICs, etc. that are close to the beret 1a to be taken out can be placed in the center. The dotted line in the figure indicates the position of the pellet that has already been taken out.

Suppose that the center of the beret 1a to be taken out is shifted from the take-out position 'Xs, Fo, as shown by the dotted line in Figure 6, and at an angle of 07? :If it is rotating, beret 1a
Since the length of one side of is known in advance, for example, by detecting the diagonal positions χ++II+ and χ8,V of the beret soto 1a, the center χ1, /j and the angle U of the beret can be found. I can do it. Also, θ formed by the length L from the rotation center 0 of the ring 3 to the center χs, ysl of the bevel, the reference line Sl, the point χ1, V, and a line passing through the center O.
Since 1 is known, when the beret is rotated by an angle θ around the rotation center 0, the − side of the beret y) becomes parallel to the reference line SL, as shown by the dashed line in the figure. Also, the center χ1,r of the bullet at this time is the position x, J4 and the angle θ9. It is determined from the angle θ and the length L. Removal position χ
Since O and FO are known, (χ0-χ4) in the X direction,
If the position is corrected by (FOF4) in the Y direction, the edges will be located at the take-out position shown by the solid line.

In this way, the tip is removed with its positional deviation corrected.

Problems to be Solved by the Invention In the above method, it is necessary to ensure that the beret to be taken out is within the effective field of view of the television camera.

On the other hand, in order to reliably detect the end position of the beret and the foot, it is necessary to make the interval between the beret and the foot sufficiently long.

However, the size of chips for semiconductor integrated circuits has increased as circuit elements have become more highly integrated, and on the other hand, the method of dividing wafers into chips and strips has changed from the method of scribing and then braking, to the method of dividing wafers into chips and strips. Power, power,
The 11th bere is moving to a method of taking out Soto.

For this reason, in the past, in addition to having a small beret and V-toe, by stretching the seat after braking and increasing the pele-toe spacing, the ratio of the beret and foot dimensions to the spacing was no more than 10, but recently, the beret and vee distance has been reduced to 10. Throat size is 1-1
For 0m+, the vertical and horizontal spacing is 20-30μ,
Since the ratio is as high as 80 to 500, if you try to fit one beret within the effective field of view of a TV camera,
The spacing between the sheets appears narrow, and even slight shadows can cause blurring.
The spacing between the pads was sometimes unclear.

Further, when using a COD camera as a television camera, for example, it is necessary to detect the interval at least 2 bits. .

However, if the giJ recording ratio is as high as 500, the commonly used TV cameras with a resolution of 256 x 256 bits or 512 x 512 bits will not be able to handle it. The present invention was proposed in view of the above problems, and it is possible to use a television camera to capture a plurality of rectangular workpieces of predetermined dimensions arranged approximately in a straight line. When capturing an image and calculating the center position of the workpiece from the captured image data to detect the position of the workpiece, the corner of the workpiece is positioned at the imaging center of the television camera, and the ratio between the workpiece dimensions and the workpiece interval is determined. To provide a method capable of detecting the position of a workpiece even when the position of the workpiece is extremely large.

EXAMPLES Below, an example of the present invention will be explained with reference to FIGS. 1 to 3 by applying it to the apparatus shown in FIG. The same reference numerals as in FIG. 4 indicate the same parts, and the explanation will be omitted.

As shown in FIG. 1, the present invention includes a semiconductor pellet la to be taken out, a semiconductor pellet 1b adjacent to or in contact with this pellet P l-1a, and a corner of an IC at approximately the center of the imaging field of view of a television camera 4. 7) 1
a, 1b are sufficiently far apart, beret y) 1 a
The area indicated by the dotted line in the figure is enlarged and imaged so that the position of the corner can be sufficiently determined. Therefore, the image shown in Fig. 2 was obtained by the television camera 4, and the beret q to be taken out was obtained.
The other eight corners of la are out of the imaging field of view.

In this way, one corner of the verge, y ) 1 a is imaged, and the vcXYθ table is rotated so that the two adjacent sides at the corner are respectively parallel to the reference line on the XYθ table. Bere that has been fully powered and lifted by dicing,
・There is less disorder in the arrangement of the belt than that of a pellet separated by the scribe method, and it is sufficient to correct the transposition shift once at the beginning.

After regulating the arrangement of the bere and bottom in the X and Y directions in this way, and finding the positions χ1 and χ of one corner C of the bere foot 1a as shown in Fig. 3, the bere = IR TR
Since the dimension t of is known, the center position χ of the pellet 1a
1χ, can be found. That is, χ6=χs'
/'2. ya=F11+'/2. When this center position χ6, V6 is found, 1. @center position χ of image field of view
o, y.

Since is already known, if the position is corrected by (χ0 - x6) in the X direction and (FO - pg) in the Y direction, the center of the vertex (la) will be located on the extraction position (χO, Vo).

In this way, move the XYθ table so that the corner of the beret 1 (after detecting the position of Z and taking it out by this and soto, the next beret y to be taken out) is in the approximate center of the imaging field of view. and repeat the above steps.

It should be noted that the present invention is not limited to the above-mentioned embodiments, but can be applied not only to semiconductor pellets but also to rectangular workpieces in general as long as the workpieces are arranged substantially in a straight line. Further, the square workpiece may be not only square but also rectangular or polygonal.

Effect As shown above, the ratio between the workpiece dimensions and the workpiece spacing is extremely large, and if you try to capture the entire view of the workpiece to be taken out within the imaging field of view, the workpiece spacing becomes extremely narrow and the boundary becomes unclear. Even if it is not possible to detect the position of the workpiece, according to the present invention, the position can be detected by imaging the distance between the workpieces with a wide enough width that the boundaries between the workpieces can be sufficiently determined.

Furthermore, even if the distance between the workpieces is narrow, it is sufficient to enlarge the boundaries of the workpieces to the extent that they can be distinguished. For example, in the case of a COD camera, it is sufficient to make the distance between the workpieces correspond to 2 to 3 bits of image pickup pixels, and 256 x 256 bits. of

[Brief explanation of the drawing]

Fig. 1 is a plan view showing an embodiment of the present invention, Fig. 2 is an enlarged view of the shaded area in Fig. 1 showing a workpiece image within the field of view of a television camera, and Fig. 8 is a determination of the center position of a workpiece according to the present invention. Drawings explaining the method, Figure 4 is a sectional side view of the main part of the workpiece take-out device, Figure 5 is a diagram showing the image of the workpiece within the field of view of the television camera according to the conventional method, and Figure 6 is the conventional method for detecting the position of the workpiece. FIG. 1, l (Z... Work, 4...
・・・・・・TV camera, C・・・・・・・・・・・・
・Corners of workpieces.

Claims (1)

[Claims] When a plurality of rectangular workpieces of predetermined dimensions arranged approximately in a straight line are imaged by a television camera, and the center position of the workpiece is calculated from the captured image data to detect the position of the workpiece, an abbreviation of the imaging field of view of the television camera is used. A position detection method characterized in that a corner of the workpiece is located in the center.