JP3189604B2 - Inspection method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a method for manufacturing an IC lead or the like having a uniform width.
Perform image recognition for patterns arranged at equal intervals,
The present invention relates to an inspection method and an apparatus for performing a pass / fail determination.

[0002]

2. Description of the Related Art There is a type in which an IC pattern is formed at a predetermined position on a continuous belt-shaped resin tape by a semiconductor process or the like. That is, as shown in FIG.
IC pattern forming regions R are set at equal intervals on the top, and various I / Os for configuring a desired circuit are formed in the region R.
A C pattern and a wiring are formed. As a result, a large number of the same patterns are repeatedly formed at predetermined intervals on the tape 1 in units of the region R.

By the way, the I formed as described above
For the C pattern and the like, it is necessary to finally inspect the presence / absence of a defect to determine the quality. In the present invention, a pattern portion in which patterns of equal width are arranged at equal intervals, such as an IC lead pattern, is particularly targeted for inspection.

That is, the IC lead pattern formed on the tape carrier 1 is formed on a resin tape by a metal film and surface plating. For this reason, the defect includes a defect in the pattern shape on the plane of the resin tape, a defect in surface plating, and a discoloration defect caused by adhesion of dirt and foreign matter caused by contact with another object until the main inspection step.

According to the conventional inspection method, these defects are:
It is visually identified through the naked eye or through a magnifying glass.
In the case of performing automatically, a normal pattern is provided as a model pattern in advance, and pattern matching between image data obtained by imaging an IC lead pattern to be inspected and the model pattern is performed. In some cases, a degree (conformity) is obtained, and a value equal to or higher than a predetermined threshold value is determined to be good.

[0006]

However, in the above-described conventional inspection method based on visual inspection, the load on the inspector is large, and there is a possibility that the criterion varies from inspector to inspector and that an inspection error such as oversight or the like occurs. There is also. Further, the processing speed also becomes slow.

On the other hand, in the latter automated method, the possibility of inspection errors such as variation and oversight is suppressed as much as possible, but it is complicated to form a model pattern and time is required for matching processing. There is a limit to speeding up.

In particular, when a high-precision matching process is to be performed, an image of the lead is magnified and an image can be taken, and a small difference can be checked. However, ICs include various types of lead widths and require an operation for changing the zoom each time, so that processing time is required.

On the other hand, imaging is performed without enlarging the lead,
When trying to detect, the proportion of the image signal of the camera pixel including the lead pattern edge in one pixel increases, and the information to be inspected becomes extremely small,
It becomes difficult for the information of the defective part to be distinguished from the others in the image signal.

The present invention has been made in view of the above background, and an object of the present invention is to provide a chip generated on a pattern arranged at equal intervals such as an IC lead pattern formed on a tape carrier. Detecting pattern shape defects due to shape defects such as surface defects and discoloration defects due to surface layer defects and dirt etc. without enlarging the leads, and enabling quick and accurate image processing to detect them. It is an object of the present invention to provide an inspection method and apparatus which can automatically set values and can follow changes in an external environment such as lighting.

[0011]

In order to achieve the above-mentioned object, in the inspection method according to the present invention, first, a plurality of pattern elements such as IC lead patterns are arranged at equal intervals.
Comprising a inspection method for performing quality determination with respect to the inspection object, from the multi-value image data of the object to be inspected picked up, the path
Scan the turn element arrangement direction to add neighboring pixel densities
And then extract the maximum or minimum concentration
Filtering processing is performed to match the pattern element existence candidate.
The corresponding peak value is determined, and the determined peak value is within a predetermined range.
The items in the box are determined as good candidates for the pattern element.
At the same time, the location of the pattern element of the good candidate is extracted.

Next, based on the extracted existence position,
Spacing adjacent to the arrangement direction of the pattern element of the good candidate
And if the interval is within a certain range, the test
The inspection target is judged to be good, and if it is out of the range, it is judged to be defective.
I decided to.

[0013] Further, the criteria for determining the good product candidate
A threshold value that defines the predetermined range is a first processed image obtained by performing an addition process of the neighboring pixel density on image data obtained by imaging an actual pattern element, It is more preferable to perform the setting by a predetermined calculation process based on the second processed image obtained by performing the local area density extraction for one processed image.

In the inspection apparatus according to the present invention, which is suitable for carrying out the above-described method, an inspection target comprising a plurality of pattern elements arranged at equal intervals, such as an IC lead pattern .
An inspection apparatus for determining the quality of an elephant , comprising: an imaging device for imaging the inspection object; and a good candidate for the pattern element from multi-valued image data obtained from the imaging device.
And determines a feature value extraction means for extracting the location of the pattern elements of the good candidates, out of the pattern elements of the good candidates extracted by the feature extracting unit, a spacing adjacent to the arrangement direction Based on the pitch measurement means to be obtained, based on the detection result by the pitch measurement means,
A pass / fail judgment means for judging whether or not the interval is within a predetermined range . Then, the feature amount extracting means
Add the pixel density of the neighboring pixels in the direction in which the pattern
A neighborhood pixel density adding means for calculating a calculated density value;
Receiving the output of the element density addition means,
Extracting means for extracting a minimum point, and an output of the extracting means.
And a peak value corresponding to the pattern element existence candidate
Is obtained, and the obtained peak value is within a predetermined range.
Is determined as a good candidate for the pattern element.
Means for determining the location of the pattern element .

[0015] and preferably said imaging device, Ru der that one pixel aperture ratio of used image sensor almost 100%.

[0016]

According to the present invention, a predetermined process is performed on a multi-valued image obtained by capturing an image of an object to be inspected in which pattern elements are arranged, and the position of a pattern element of a good candidate is extracted. At this time, more specifically, the pixel (density) information on the pattern element is shifted to the center by adding the pixel density of the neighboring pixels, so that the density becomes higher (or lower) at the center. As a result, peaks (or valleys) of the density data appear corresponding to the portions where the pattern elements exist. The peak of each peak (maximum value) or the peak of a valley (minimum value) is located at the center of the pattern element, and if normal, each peak is constant. Therefore, the feature quantity extracting means estimates that each peak is within a certain range as a peak based on the pattern element with the good candidate, and sends the data to the pitch measuring means.

The pitch measuring means obtains an interval between adjacent good candidate positions on the basis of the given position of the good candidate, and judges whether or not the distance is within a certain range by the good / bad judgment means. That is, if there is a chip in the pattern element or the like, there is no peak in that part, and the interval becomes longer, so that it is determined to be abnormal. When an abnormal pattern having a similar density or the like exists between normal pattern elements, the abnormal pattern is also extracted as a good product candidate, so that the distance between adjacent good product candidates is shorter than the design value. Is determined to be abnormal.

In the case where the threshold value used for determining a good product candidate is set using a pattern actually used for determination, a background portion is determined based on an image obtained by adding the density of neighboring pixels. (For example, a histogram of each density is obtained, the mode is set as the most frequent part, or an average is obtained, and it can be determined by various methods). Further, a standard density of the pattern element is obtained based on the image data when obtaining the density of the local maximum or the local minimum (for example, a histogram of each density is obtained, the mode is set as a mode, or an average is obtained. ). Then, the threshold can be determined from the two standard densities. That is, a value is set such that the density of the background portion does not fall within the margin when a margin of ± with respect to the standard density of the pattern element is taken. Then, a value obtained by adding the margin thus obtained to the standard density of the pattern element becomes a threshold value.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the inspection method and apparatus according to the present invention will be described below in detail with reference to the accompanying drawings. As shown in FIG. 1, the tape carrier 1 is continuously pulled out from the raw material 1a wound in a roll shape, guided to an inspection processing area of the present apparatus, and then wound up again.

As shown in FIG. 2, IC pattern forming regions R are set at predetermined positions on a band-shaped resin tape carrier 1 by a semiconductor process or the like, and a desired circuit is provided in the region R. Are formed with various IC patterns and wirings. As a result, a large number of the same patterns are repeatedly formed at predetermined intervals on the tape 1 in units of the IC pattern formation region R. Since the IC lead pattern formed on the tape carrier 1 is formed on the resin tape by the metal film and the surface plating, the IC lead pattern is brighter than the adjacent part where the pattern is not formed. Therefore, as shown in FIG.
Part and unformed part 3 are alternately arranged.
Band portions (white portions in the drawing) having the same width and high brightness corresponding to the formation portions of the lead patterns 2 are arranged in parallel at predetermined intervals.

In the inspection apparatus according to the present embodiment, an image sensor 5 serving as an imaging device for imaging a predetermined position on the surface of the tape carrier 1 located in the inspection processing region, and an image signal output from the image sensor 5 are used. I
A feature amount extraction unit 10 for extracting a C lead pattern portion;
A pitch measuring circuit 12 for detecting a distance between adjacent lead patterns extracted by the feature extracting unit 10, and an I / O present in an area imaged by receiving an output of the pitch measuring circuit 12.
A pass / fail judgment circuit 14 for judging pass / fail of the C lead pattern.

The feature amount extraction unit 10 includes a neighborhood pixel addition circuit 1
0a, a local maximum concentration extraction circuit 10b, and an arithmetic circuit 10c.
4a and a comprehensive judgment circuit 14b. The details of each circuit will be described later.

Further, in order to clarify the contrast of the image (the surface to be inspected) on the tape carrier 1 picked up by the image sensor 5, the entire inspection region (the region including one IC pattern forming region R in this example) is formed. A light projecting device 20 for uniformly irradiating light is provided. The light emitting device 20 includes a light source 2
1, a diffuser 22 for uniformly irradiating the light to be inspected with the light emitted from the light source 21, and the diffuser 22
Lens 23 that converts light transmitted through the lens into a parallel light flux
And a half mirror 24 that reflects the parallel light beam, changes its optical path, and irradiates a predetermined position on the tape carrier 1 with light.

The light reflected by the half mirror 24 is
The angles of the members 21 to 24 are adjusted so that the tape carrier 1 is irradiated perpendicularly, so that the light reflected on the surface of the tape carrier 1 has an optical path of one.
After being converted by 80 degrees, it reaches the half mirror 24 again, transmits as it is, enters the image sensor 5 installed above the half mirror 24, and forms an image there.

Further, since the reflected light incident on the image sensor 5 is a part of the IC pattern forming region R, it is necessary to move the image sensor 5 and the tape carrier 1 relative to each other. The carrier 1 is intermittently moved by a driving means (winding motor) (not shown) at each pitch of the IC pattern formation region R so that the IC lead pattern to be processed present in one IC pattern formation region R is moved. The image acquisition is performed by moving the image sensor 5 horizontally.

Specifically, the image sensor 5 is mounted on the XY table 30, and the feed motor 31 for the XY table 30 is rotated forward and reverse based on a control signal from the drive control circuit 32. The drive control circuit 32
Operate according to a control command from the overall control circuit 33.

Next, each part will be described. The image sensor 5 uses a CCD as a detection element, and in this example, uses a high-performance one in which the aperture ratio of one pixel is almost 100%.

The feature extraction unit 10 extracts an IC lead pattern portion. There are two cases where the IC lead pattern is present in one pixel in the image sensor 5.
There is a case where no IC chip exists, or a case where an edge of an IC lead pattern is included in one pixel. In the case where such an edge is included, the detection density differs depending on the existence ratio of the IC lead pattern and the tape portion. In the present invention, in order to finally obtain the distance between the adjacent IC lead patterns, the feature position is first extracted from the existence position of each IC lead pattern. In this embodiment, the center of the lead pattern is obtained.

Therefore, first, the neighborhood pixel addition circuit 10a performs a spatial filtering as a neighborhood pixel addition process for adding the densities of a predetermined number of adjacent pixels in the arrangement direction of the lead pattern. Specifically, N × N (3 in this example) shown in FIG.
× 3), using a mask M including a local area, the central pixel is defined as a0, and the content (density) of each pixel in the column including a0 in the lead pattern arrangement direction (the left-right direction in this example) is (a
0), (a1) and (a2). Then, the circuit 10
In a, the summation operation is performed, and the added density value A at the central pixel to be processed is

[0030]

(Equation 1) Is calculated.

Thus, the central portion of the IC lead pattern has the highest density value (brighter the higher). Actually, N is determined by the ratio between the width of the lead pattern and the length of one pixel. That is, for example, when the width of the lead pattern is set to be equal to the width of N pixels, when the center of the pattern coincides with the center pixel (pixel to be processed) of the mask M, all the pixels to be added arranged in a line are reflected light from the lead pattern. , The density of each pixel shows a high value,
On the other hand, if even one pixel shifts, at least one pixel becomes an edge portion or a tape portion, and the density becomes low (dark). Therefore, the added density value A at the center of the lead pattern
Is highest, and becomes lower toward the end. Therefore, when the scanning is sequentially performed in the lead arrangement direction and the change of the added density value A at that time is seen, the peak shape becomes a mountain shape and the vertex position becomes the center position of the lead pattern. When N is reduced, the additional density value A is obtained as long as the lead pattern exists in one horizontal row of pixels.
Is the maximum (vertex), so that the vertex has a flat mountain shape at the top. In such a case, the center of the candidate points (the location where the added density value is the maximum) is the position of the lead pattern to be obtained. Becomes In addition, in order for detection to perform simple processing,
The former (in which the width of N pixels is matched with the pattern width) in which the vertex position can be easily detected is more preferable. The edge portion of the lead pattern is often inclined (chamfered) due to the requirements of the manufacturing process. Therefore, when determining N, the width of the pattern is the width of the flat upper surface excluding the inclined portions at both ends. More preferably,

Thus, for example, when a normal IC lead pattern as shown in FIG. 3A is scanned at a certain position (indicated by a dashed line), as shown in FIG.
A peak having a high density appears at a position corresponding to the lead pattern portion, and a peak (indicated by an arrow) appears at a central portion.

The pixels used in the mask M are
In the above-described example, the center pixel and the pixels that exist in the left and right direction are included. However, if the lead arrangement direction is orthogonal to the direction shown in FIG. Pixels existing in the vertical direction will be used.

The local maximum extracting circuit 10b detects the peak (local maximum) position of each peak of the maximum density value A in the scanning direction obtained by the above-described neighboring pixel adding circuit 10a, and performs spatial filtering. Alternatively, extraction is performed by a known method such as rank-value filtering and operation between images. Thus, by passing the characteristic of the added density value shown in FIG. 3B through the extraction circuit 10b, a characteristic in which only the apex portion protrudes as shown in FIG. 3C is obtained. The value of each peak is the same as the value of the maximum density value A. The higher the maximum density value A, that is, the brighter the peak, the higher the peak value.

The arithmetic processing section 10c determines a peak portion corresponding to a non-defective IC lead pattern candidate among the peaks extracted by the local maximum portion extracting circuit 10b, and outputs the location of the peak portion. That is, for example, if a foreign substance (having a low reflectance) is attached to the upper surface of the IC lead pattern, or if the pattern is lacking and the underlying tape portion is exposed, the portion becomes dark and the peak value is low. Become. Conversely, if the surface plating portion is lost and the metallic glossy surface having a high reflectance is exposed, the amount of reflected light at that portion increases, and the peak value increases.

Therefore, it is abnormal if the peak value is higher than the peak value corresponding to the normal IC lead pattern portion by a certain value or more, and if it is lower than the peak value, it is abnormal. Therefore, if the IC read pattern is normal, the peak value corresponding to the normal read pattern should be within a certain range. In this example, two thresholds VTH1 and VTH2 that determine the normal range are set ( The arithmetic processing unit 10c extracts the extraction circuit 10b (see FIG. 5).
Are compared with both threshold values, a peak value existing within the range is detected, and the position (coordinate data) of the peak value is output. As a result, non-defective IC lead pattern candidates whose peak values are within the normal range are extracted. Then, as shown in FIG.
When all the IC lead patterns are normal, FIG.
As shown in the figure, all three peaks are within the normal range (VTH
2 to VTH1).

In this embodiment, the threshold values VTH1, VTH
2 are provided manually from an external device via an input device (keyboard, numeric keypad, touch panel, etc.) (not shown) (automatically set based on image data obtained from a model pattern, as will be described later). It is also possible).

The above processes in the circuits 10a to 10c are performed for each line while scanning the image data obtained by the image sensor 5 in a raster system. Therefore, the characteristic of the above-described added density value (see FIG. 3B) and the peak value based on it (see FIG. 3C) are obtained for each line, and a good-quality IC lead pattern candidate point is extracted. Will be.

As described above, the feature quantity extracting unit 10 (the arithmetic processing unit 10c) sends the coordinate data of the non-defective IC lead pattern candidate to the pitch measuring circuit 12 at the next stage. The distance between the IC lead pattern candidates is measured.

More specifically, as shown in the flowchart of FIG. 6, first, coordinate data of a candidate for a good product provided from the feature quantity extraction unit 10 is received, and the received coordinate data is stored in a memory in a pair with a pattern No. I do. At this time,
They are stored in the order of appearance when scanning in the raster direction, that is, for each line, and on the same line in the order of coordinate values (ST1). Note that this storage memory is RAM
Independent hardware as a storage element such as
Alternatively, it may be in a CPU (a memory (buffer) on a program).

Next, specific pitch measurement is performed.
That is, first, the coordinates of the pattern candidate existing for the first line and the coordinates of the next pattern candidate are read, and the difference between the two is obtained and output (ST2, S2).
T3). Since the numbering is performed for each pattern candidate for convenience, the data to be output is performed in association with the calculation result and the number.

Then, if the pitch between certain adjacent pattern candidates is obtained. The pitch with the coordinates of the next pattern candidate is obtained. That is, the number (pattern No.) assigned to the pattern candidate is incremented and increased by 1 (ST4). Then, it is determined whether or not there is a pattern candidate (a pattern candidate having a number one higher than the pattern No. incremented in step 4) in the traveling direction (ST5). A pitch between them is required.

If there is no next pattern candidate as a result of the branch determination in step 5, the line is incremented (S
T6) Then, the above processing is performed from the beginning of the next line. Then, the above-described processes are repeated until there is no more processing area (unprocessed data).

As a result, the distance (pitch) between adjacent non-defective lead pattern candidates extracted by the characteristic amount extraction unit 10 is obtained, and the calculation result is used as the next-stage pass / fail judgment circuit 1.
4 given.

The pass / fail judgment circuit 14 includes a comparison circuit 14 in the preceding stage.
According to a, it is first determined whether or not each given pitch is within a predetermined non-defective range. This comparison circuit 1
4a has, for example, a window comparator as a basic configuration,
Data that can determine whether or not the input data (the calculated pitch) exists between two preset thresholds (reference values) is used. In this example, when the input data is out of the range, an abnormality (High) ).

That is, assuming that there is no single lead pattern on a scanning line having a missing IC lead pattern, the pitch is twice as long as the normal pitch (the pattern existing on both sides of the missing pattern). (The distance between the two is measured), so that the abnormality is detected. Conversely, if any high-reflectance foreign matter is attached between the normal lead patterns, the pitch is shortened, so that an abnormality is detected.

It should be noted that the reference value serving as the above criterion is also:
The setting is made externally via an input device (not shown). Since the arrangement pitch of the IC lead pattern is clearly known in advance by design, this value can be set relatively easily by inputting a value having a certain margin.

The result of the comparison is output to the comprehensive judgment circuit 14.
b, and a final pass / fail judgment is made. In this pass / fail determination, if all the IC lead patterns present in one IC pattern formation region R are normal, it is determined to be non-defective, and if at least one is abnormal (there is a detection output from the comparison circuit 14a), The region R is set to be defective (the allowable amount is 0 in this example). If the number of abnormal locations has a certain allowable amount depending on the inspection target, the number of detected abnormalities is integrated, and it is determined whether the total number is within an allowable range to make a pass / fail determination.

When the judgment processing for the image data picked up by the image sensor 5 is completed, the comprehensive judgment circuit 14b outputs a predetermined control signal to the overall control circuit 33. I have. Accordingly, the overall control circuit 33 sends a movement command signal of the image sensor 5 to the table control circuit 32, and in response to this, the table control circuit 32 changes the imaging region of the image sensor 5 to the next inspection in the same region R. A control command required to move the XY table 30 so as to move the XY table 30 over the target is output. It should be noted that a specific drive control can be performed using a known drive mechanism and control flow, and thus detailed description thereof will be omitted.

When the inspection of all the IC lead patterns in the same area R is completed, the comprehensive judgment circuit 14b makes a final pass / fail judgment and sends the judgment result to the overall control circuit 33. The overall control circuit 33 stores the pass / fail judgment result in the built-in storage means in combination with the area number (continuously attached from the beginning of the tape).
At the same time, a drive command is output to a tape driving unit (not shown) so that the tape is moved by the pitch of the region R in order to perform processing for the next region R.

In the above embodiment, the arithmetic processing unit 1
The output from 0c is such that only coordinate data which is a candidate for a good product is output. However, the present invention is not limited to this. For example, the density region between VTH1 and VTH2 is assumed to be a normal IC lead pattern "1". The other parts may be set to “0”, and “1/0” may be determined and output while scanning each line.

Thus, the image data to be processed is
Since it is encoded into “1/0”, the comparison circuit 14a
The pitch measurement is performed by extracting “1” existing on the same line and calculating the distance between adjacent “1” pixels.

Next, the operation of the above-described apparatus, which is one embodiment of the inspection method according to the present invention, will be described in detail (particularly with respect to defects), together with specific examples of IC lead patterns. First, on the tape carrier 1, for example, FIG.
It is assumed that there is a cut 40 defect on the central pattern 2b of the IC lead patterns 2a, 2b, 2c as shown in FIG. FIG.
As shown in (A), the IC lead patterns 2a, 2b, 2
It is assumed that there is a defect 41 in the center pattern 2b of c, and the tape 1 as a base is exposed.

Then, the line indicated by the dashed line is scanned,
When the addition processing is performed by the neighboring pixel addition circuit 10a, a cut 4
In the case of a defect of 0, there is no IC lead pattern, and therefore, as shown in FIG.
In the case of the defect of the notch 41, the peak of the peak becomes smaller as shown in FIG. 8B, and the peak position is also smaller than the original position, as shown in FIG. Are also shifted to the right (since the peak appears in the center of the actual pattern portion). Therefore, when such added density value data is sent to the next-stage maximum location density extraction circuit 10b, a peak value is extracted as shown in FIG.

Then, each peak value is compared with the threshold values VTH1 and VTH2 in the arithmetic processing section 10c, and finally the central I
Since the peak value corresponding to the C lead pattern 2b is equal to or less than the lower threshold value VTH2, coordinate data for the IC lead patterns 2a and 2c on both sides is output. Therefore, in the pitch measuring circuit 12, the intervals between the adjacent IC lead patterns are both the same as those shown in FIGS. 7C and 8C.
Since the interval D between the patterns on both sides is obtained as shown in (1), the comparison circuit 14a determines that it is longer than the reference value. Then, the integrated determination circuit 14b determines that there is an abnormality (the same applies hereinafter).

Further, as shown in FIG. 9A, there is a defect 42 between the left and center patterns 2a, 2b of the IC lead patterns 2a, 2b, 2c, and the underlying tape portion to be originally exposed. Is hidden.

Then, the line indicated by the dashed line is scanned,
When the addition process is performed by the neighboring pixel addition circuit 10a, in the case of the defect of the connection 42, the IC lead pattern portion having a high brightness exists continuously, and as shown in FIG. A mountain-like characteristic in which the summit where the added density values equivalent to the following are continuous is flat. Therefore, when such added density value data is sent to the next-stage maximum-density density extraction circuit 10b, a peak value as shown in FIG. That is, for the right IC lead pattern 2c, a peak of a predetermined height appears at a position where it should be, but the connected IC lead pattern 2c has a predetermined height.
a and 2b are regarded as one peak, and one peak value is output as a representative of the central portion thereof. (If the added density values are equal, the two-dot chain line in FIG. Peaks may appear at each position, as shown by.

Then, each peak value is compared with the threshold values VTH1 and VTH2 by the arithmetic processing unit 10c, and each output peak value is within the range of both threshold values VTH1 to VTH2. The coordinate data is output. Therefore, since the position other than the peak value corresponding to the right IC lead pattern 2c appears at a position different from the original IC lead patterns 2a and 2b, the coordinate data also deviates from the actual position. Therefore, when one peak value appears for a series of peaks, the pitch measuring circuit 12 uses the same figure (C).
Since the interval D is obtained as shown in FIG.
Is determined to be longer than the reference value. As shown by the two-dot chain line, even if a plurality of peak values are output in a series of peaks, the interval D 'is too small, so that the comparison circuit 14
In a, it is determined that it is shorter than the reference value. Therefore, in any case, it is determined to be abnormal.

The above three examples show examples of shape defects on the plane of the tape carrier 1, and it has been described that in such a case, it is possible to reliably recognize (pass / fail judgment). The fact that the present invention can be applied to the generated discoloration defect will be described.

That is, as shown in FIG.
It is assumed that there is a defect of a stain (fingerprint or the like) 43 on the left and center patterns 2a, 2b of the C lead patterns 2a, 2b, 2c. Further, as shown in FIG.
a, 2b, 2c, the left and center patterns 2a,
It is assumed that there is a defect of the foreign matter adhesion 44 covering the upper surface 2b.

Then, the line indicated by the dashed line is scanned,
When the addition processing is performed by the neighboring pixel addition circuit 10a, the contamination 4
In the case of the defect of No. 3, since the reflectance at that portion decreases, darkening occurs in the image data picked up by the image sensor 5, and as shown in FIG.
A peak appears in the lead pattern portion with an increase in the added density value, but its peak becomes smaller. In addition, foreign matter adhesion 4
In the case of the defect No. 4, as shown in FIG. 11 (B), a mountain-like characteristic in which the summit is flat and the summit is flat at a low value appears. Therefore, when such added density value data is sent to the next maximum local density extraction circuit 10b, a peak value is extracted as shown in each diagram (C).

Then, each peak value is compared with the threshold values VTH1 and VTH2 in the arithmetic processing unit 10c. Since the peak value corresponding to the defective portion is finally equal to or smaller than the lower threshold value VTH2, the right IC read pattern is obtained. Only the coordinate data for 2c is output. Therefore, pitch measurement circuit 1
In No. 2, since the interval between adjacent IC lead patterns is obtained by comparing with the coordinate value of a pattern (not shown), the comparison circuit 14a determines that the interval is longer than the reference value. Also,
If there is no coordinate of another good IC lead pattern candidate existing in the line, no pitch is obtained (0 or infinity), and the comparison circuit 14a also recognizes that the outside of the normal range.

As shown in FIG. 12A, the center pattern 2 of the IC lead patterns 2a, 2b and 2c is used.
It is assumed that there is a plating defect or scratch (scratch etc.) 45 on b. Then, a metallic glossy surface having a high reflectance is exposed due to a plating defect or the like, and whitening occurs with an increase in the amount of reflected light, which is reflected on an image signal.

Therefore, if the line indicated by the dashed line is scanned and added by the neighboring pixel addition circuit 10a, the reflectance at the portion of the stain 43 is reduced in the case of a defect of the dirt 43. As shown in ()), a peak accompanying an increase in the added density value appears in an IC lead pattern portion which should be, and the central peak becomes large. Therefore, when such data of the added density value is sent to the local maximum density extraction circuit 10b at the next stage, a peak value is extracted as shown in FIG.

Then, each peak value is compared with the threshold values VTH1 and VTH2 in the arithmetic processing unit 10c. Since the central peak value corresponding to the defective portion finally becomes equal to or more than the upper limit threshold value VTH1, the coordinates of the relevant portion are obtained. No data is output, and coordinate data on the IC lead patterns 2a and 2c on both sides is output. Therefore, in the pitch measurement circuit 12, the interval between adjacent IC lead patterns is
Since the interval D between the patterns on both sides is obtained as shown in FIG. 2 (C), the comparison circuit 14a determines that it is longer than the reference value. As described above, any of the defects can be reliably detected, and a defective product can be accurately detected. In each of the above examples, the case where there is a defect has been described. However, in the case of a normal pattern, FIGS. 3A to 3C described while describing the above-described apparatus.
Is correctly determined to be a non-defective product through the processing shown in FIG.

Next, another embodiment of the inspection apparatus according to the present invention including a threshold value setting function (apparatus) for automatically setting two threshold values VTH1 and VTH2 serving as judgment criteria in the arithmetic processing section 10c. An example will be described. As shown in FIG. 13, the configuration is basically the same as that of the above-described embodiment, and the corresponding components are denoted by the same reference numerals.

Here, in the present embodiment, the neighboring pixel addition circuit 1
The output of 0a is also provided to the arithmetic circuit 10c. Then, in the arithmetic circuit 10c, the neighboring pixel adding circuit 1
Non-defective I based on the image data after addition value filtering given from 0a and the data about the maximum value (peak) given through the maximum location density extraction circuit 10b.
Two thresholds V for determining C lead pattern candidates
TH1 and VTH2 are set.

Specifically, an area including an IC lead pattern to be processed is imaged by the image sensor 5, and the obtained image is processed according to the flow shown in FIG. As a pattern to be read at this time, it is preferable to use a non-defective pattern. However, a pattern including a defect may be used as long as the number of defects is small.

First, a histogram (b) of each density is obtained from the output data of the neighboring pixel density adding circuit to obtain the center of the density value of the background (ST11). Further, a histogram (c) of each density is obtained from the output data of the local maximum density extraction circuit to obtain the density value center of the IC lead pattern (ST12). The processing of both steps is
The processing may be performed by hardware for speeding up.

Then, a graph shown in FIG. 15A is obtained by the processing of step 11, and a graph shown in FIG. 15B is obtained by performing the processing of step 12. In FIG. 15A, d is the density range of a pixel other than the IC lead pattern, e is the density range of a pixel including an IC lead pattern edge in one pixel, or the density range of a dark defect, and f is The density range of the pixel of only the IC lead pattern corresponds to the density range of the whitish defect.

Next, a temporary density value P for separating the IC lead pattern from the background is obtained from the histogram (b) (ST).
13). That is, the ratio of the left and right histogram areas divided by the density value P is the width of the IC lead pattern and the distance (lead pitch−IC
P is set at a position that is equal to the ratio of the lead pattern width. If the lead width is 0.3 mm and the lead pitch is 0.1 mm.
In the case of 6 mm, a position is selected such that the areas of the left and right histograms are equal (1: 1).

Then, the density values 0 to 0 of the histogram (b)
The most frequent density value Q up to P is obtained (ST14), and the density value P to 255 (M
AX) is determined (ST15). Each mode density value Q, R is specifically obtained by executing the following equation.

[0073]

(Equation 2) Next, the lower limit S and the upper limit T of the non-defective product density value are obtained by substituting the obtained Q and R into the following equation (FIG. 1).
5 (C)). Then, since the calculated T becomes VTH1 and S becomes VTH2, each value is stored and used as a threshold value in a subsequent inspection. As a result, the threshold value can be automatically determined, and even after it has been determined once, it can follow a certain degree of illuminance fluctuation.

[0074]

(Equation 3) When the two threshold values determined as described above are determined, a normal inspection process is performed. The specific configuration and operation for performing the inspection process are described in the first embodiment. Therefore, the same reference numerals are given and the description is omitted.

In each of the above embodiments,
Although an example in which the present invention is applied to an inspection apparatus for a C-lead pattern has been described, the present invention is not limited to this, and can be applied to an inspection for an arrangement in which equal-width patterns arranged in parallel at the same pitch are arranged.

Further, in the above-described apparatus, since the reflectance of the IC lead pattern, which is the pattern element to be detected, is higher than that of the background (tape) portion, the density of the IC lead pattern portion becomes high, so that the maximum value is obtained. Although the value (peak) is obtained, if the relationship of the shading is reversed, the pattern element portion becomes a valley shape with a low level. Therefore, it is a matter of course to obtain the minimum density of the peak.

[0077]

As described above, in the inspection method and apparatus according to the present invention, when there is a defect such as a chipped pattern element, the element portion is eliminated by the feature amount extracting means.
Only a good candidate portion is extracted. The extraction can be performed at a high speed and accurately, for example, by a simple process such as addition of a density value, extraction of a maximum / minimum value, and comparison with a threshold value. Furthermore, in the final pass / fail judgment, if there is a non-defective product, all the pattern element existence positions are extracted as non-defective products, so that the distance (pitch) between adjacent candidates is as designed, but there is a certain defect, If not extracted by the quantity extracting means, the interval becomes longer than the design value, and this can be performed simply and at high speed by comparison with the reference value.

Further, the aperture ratio of the imaging device is almost 100.
%, Even when an edge of a pattern element exists in one pixel, the edge portion can be accurately detected (reflected to the density addition), and the accuracy is higher. Become.

Further, since there is no need to perform the enlargement processing (of course, it does not matter), there is no time loss associated with changing the zoom and no complicated processing. In addition, when adding the neighboring pixel densities, if the number of pixels to be added is substantially equal to the width of the pattern element, the density at the center of the pattern element becomes a peak, so that not only the center position is easily extracted, but also the actual Are different, the peak values and the shapes of peaks and valleys that appear are different, so that a good candidate can be easily extracted.

When the threshold value for extracting a good product candidate is set by using the actual pattern, the threshold value can be set easily, and the change of the inspection environment (particularly, the influence of illumination or the like). (A change in illuminance applied to the inspection area) can be easily and accurately followed, and a more accurate inspection can be performed.

[Brief description of the drawings]

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

FIG. 2 is a diagram illustrating a tape to be inspected.

FIG. 3 is a diagram illustrating an example of a normal IC lead pattern.

FIG. 4 is a diagram illustrating an example of a mask used in a neighboring pixel addition circuit.

FIG. 5 is a diagram illustrating the operation of an arithmetic processing unit.

FIG. 6 is a flowchart illustrating functions of a pitch measurement circuit.

FIG. 7 is a diagram illustrating the operation of the present embodiment.

FIG. 8 is a diagram illustrating the operation of the present embodiment.

FIG. 9 is a diagram illustrating the operation of the present embodiment.

FIG. 10 is a diagram illustrating the operation of the present embodiment.

FIG. 11 is a diagram illustrating the operation of the present embodiment.

FIG. 12 is a diagram illustrating the operation of the present embodiment.

FIG. 13 is a configuration diagram showing another preferred embodiment of the inspection apparatus according to the present invention.

FIG. 14 is a flowchart showing a part of the function of an arithmetic processing unit (threshold value determination).

FIG. 15 is a diagram illustrating the operation of a part of the function of the arithmetic processing unit (threshold value determination).

[Explanation of symbols]

2, 2a, 2b, 2c IC lead pattern (pattern element) 5 Image sensor (imaging device) 10 Feature amount extraction unit 10a Nearby pixel addition circuit 10b Local maximum density extraction circuit 10c Arithmetic processing unit 12 Pitch measurement circuit 14 Quality determination unit 14a Comparison circuit 14b Comprehensive judgment circuit

Claims (5)

(57) [Claims] 1. A test method for performing quality determination for IC lead patterns to be inspected composed of a plurality of pattern elements which are arranged at regular intervals, such as, the multi-valued image data of the object to be inspected picked up, the Pa
Scan the turn element arrangement direction to add neighboring pixel densities
And then extract the maximum or minimum concentration
Filtering processing is performed to match the pattern element existence candidate.
The corresponding peak value is determined, and the peak value within the predetermined range is determined.
And it determines the good candidate turn elements, extracts the location of the pattern elements of the good candidates, putter of the good candidates based on the location of the extracted
An interval adjacent to the component in the arrangement direction is determined. If the interval is within a certain range, the inspection target is determined to be a non-defective product, and if the interval is out of the range, it is determined to be a defective product. Inspection method to be characterized. 2. A criterion for determining the good product candidate.
A first processing image obtained by performing an adding process of the neighboring pixel density on image data obtained by imaging an actual pattern element; 2. The inspection method according to claim 1, wherein a predetermined calculation process is performed based on a second processed image obtained by extracting the maximum or minimum portion density. Wherein when performing the addition processing of said neighboring pixel density, the number of pixel addition target, also claim 1, characterized in that it has substantially equal to the number of pixels corresponding to the width of the pattern element
Is the inspection method described in 2 . 4. A testing apparatus that performs quality determination on a plurality of inspection object consisting of pattern elements arranged at equal intervals of IC lead pattern and the like, an imaging device for imaging the object to be inspected Target, the imaging multi-valued image data obtained from the device, the path
Determine good candidates for the turn element and determine the good
A feature amount extracting section which extracts the location of the pattern elements of the complement, of the pattern elements of the good candidates extracted by the feature extracting unit, a pitch measuring means for determining the intervals adjacent to the arrangement direction, the Good or bad judgment means for judging whether or not the interval is within a certain range based on the detection result by the pitch measurement means, wherein the feature quantity extraction means adds the neighboring pixel density in the direction in which the pattern elements are arranged. and a neighboring pixel density addition means for obtaining a sum concentration values, receiving an output of said neighboring pixel density addition means, an extraction means for extracting a maximum or minimum point of the addition concentration values, receiving an output of the extraction means, the pattern Element presence
The peak value corresponding to the complement is found, and the found peak value is
Those within a certain range are considered as good candidates for the pattern element.
Is determined, and the location of the pattern element of the good product candidate is determined.
Inspection apparatus characterized by comprising a means for determining. 5. The inspection device according to claim 4, wherein the imaging device is an image sensor having an aperture ratio of one pixel of about 100%.