JP5852641B2 - Automatic inspection device and alignment method in automatic inspection device - Google Patents

Description

  The present invention relates to an automatic inspection apparatus that can be inspected in alignment with a reference image when inspecting the quality of an inspection object.

  In general, when inspecting the formation state of an inspection object, an image of the inspection object is acquired and compared with a reference image stored in advance to inspect the quality of the formation object.

  By the way, when inspecting the formation state of such an inspection object, it is necessary to accurately align the acquired image of the inspection object with the reference image, but conventionally, when performing such alignment, A method of searching for the positions of a plurality of reference marks (for example, pads) of the inspection object in advance and aligning the positions with the reference marks of the reference image (Patent Document 1, Patent Document 2, etc.) A method (for example, Patent Document 3) of gradually aligning the entire image of an object by changing the resolution is adopted.

JP 2007-333664 A JP 2003-086919 A JP 2009-162718 A

  However, the above-described alignment method causes the following problems.

  That is, in the method of searching for and aligning the positions of a plurality of reference marks as in the former, the reference marks can be aligned, but in other areas, they are aligned with the reference image. I do not know if. In particular, if the inspection object is warped or distorted, it will be difficult to perform accurate alignment at the intermediate part between the reference mark and the reference mark. Since it will be inspected at the time, there is a possibility of erroneous determination. In addition, in the latter method, where the resolution of the image of the inspection object is changed and the position is gradually aligned, the calculation amount increases as the resolution increases, and similarly, the inspection object is warped. When distortion or the like occurs, even if alignment is performed in one area, there is a possibility that a positional deviation occurs in another area. For this reason, it is necessary to perform inspection by aligning at a certain resolution.

  In addition, in any of the alignment methods, alignment is performed using characteristic portions. For example, in a region where there is not much change such as a substrate portion of a printed board or a resist-only portion. It becomes impossible to align with the reference image. For this reason, even if the region where accurate alignment could not be performed is inspected against the reference image, an erroneous determination may occur.

  Therefore, the present invention has been made paying attention to the above-mentioned problems, and is inspected by an automatic inspection apparatus and an inspection apparatus that can perform high-precision inspection by accurately aligning an image of an inspection object with a reference image. It is an object of the present invention to provide a method for correcting the position of an image of an inspection object.

That is, in order to solve the above-described problems, the present invention provides an automatic inspection apparatus in which an image acquired from an inspection object and a reference image are aligned to determine whether or not the inspection object is good. The first image aligning means for aligning the entire image with the reference image and the entire image aligned by the first position aligning means, and the amount of change in luminance value in the divided area is a predetermined value. A change amount detecting means for determining whether or not the above is detected, a second alignment means for aligning a corresponding area of a reference image with respect to an area where the change amount of the luminance value is equal to or greater than a predetermined value, and the position The variation of each pixel is calculated based on the combined image, the allowable luminance width at the position of the pixel is determined from the variation, and the allowable luminance width is stored in the polar coordinate system. It is obtained by so and a reference data generation unit that.

According to the present invention, when aligning the image of the inspection object with the reference image, the entire image is roughly aligned, and then the region having a large change in luminance value (i.e., the image has a change in the image). Since the alignment is performed for a large number of regions), it is possible to reduce the amount of calculation, and it is possible to perform the accurate alignment for a region with many changes that requires a particularly accurate inspection. In addition, when generating the reference data, the reference data can be generated accurately, and if the permissible luminance width is expressed in a polar coordinate system, only the brightness changes and the ratio of saturation and hue changes. Even if not, the reference data can be easily generated.

Functional block diagram of an automatic inspection apparatus according to an embodiment of the present invention The figure which shows the state aligned by the 1st position alignment means in the form The figure which shows the area | region where the variation | change_quantity of the luminance value in the same form is more than predetermined value, and an area below predetermined value The figure which shows the state which aligns the area | region where the variation | change_quantity of the luminance value in the same form is more than predetermined value The figure which shows the state which aligns the area | region where the variation | change_quantity of the luminance value in the same form is below a predetermined value Flowchart showing the position correction process of the automatic inspection apparatus Flowchart for generating reference data by automatic inspection device in the same form The flowchart which shows the process in the case of test | inspecting with the automatic test | inspection apparatus in the same form

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings. The automatic inspection apparatus 1 according to this embodiment can inspect the quality of the formation state of pads, wiring patterns, silk, resist, through holes, etc. formed on a printed circuit board 8, for example. Image acquisition means 2 for acquiring an image of the substrate 8, reference image storage means 5 for storing a reference image necessary for inspection in advance, and alignment for aligning the image acquired by the image acquisition means 2 with the reference image Means 3 are provided. Characteristically, with respect to the alignment means 3, the first alignment means 31 that aligns the entire image with a coarse resolution and the whole that is roughly aligned by the first alignment means 31. A second alignment unit 32 that divides an image into small areas and performs alignment only for areas where the luminance change is large among the divided areas; And a third alignment means 33 for performing alignment based on the correction information of the region aligned in step (b). Then, by performing alignment in this way, reference data is generated by the reference data generation means 7, and the quality of the printed circuit board 8 is determined by the determination means 6 using the reference data. Hereinafter, the automatic inspection apparatus 1 according to the present embodiment will be described in detail.

  First, the image acquisition means 2 acquires the surface image from the printed circuit board 8 which is a test object. When acquiring a surface image from the printed circuit board 8, the surface image may be acquired by a gray scale, but if it is a gray scale, a pad area where metal is exposed, an area where a resist is applied on the pad, Luminance changes such as areas where resist is directly applied on the substrate are not clear. For this reason, the surface image of the printed circuit board 8 is preferably acquired by RGB. When the surface image of the printed circuit board 8 is acquired by the image acquisition means 2, light is emitted from an illumination device arranged obliquely above the printed circuit board 8, and the reflected light is reflected on the line sensor or the The image is acquired by the area sensor, and the image is temporarily stored in the image memory. Then, the difference between the image stored in the image memory and the image when the printed board 8 is not placed (for example, the image of the stage) is taken, and the image of only the printed board 8 is extracted.

  On the other hand, the reference image storage means 5 stores an RGB image of the printed circuit board 8 serving as a reference in advance. Here, the “reference image” means an image of an inspection object that has been determined as a non-defective product by visual inspection or another inspection device in advance, or image data created from CAD data or the like. When the reference image is stored in the reference image storage unit 5, at least an image having a resolution higher than the resolution used in the first alignment unit 31 and the second alignment unit 32 described later is stored. To.

  The alignment unit 3 includes a first alignment unit 31, a second alignment unit 32, and a third alignment unit 33.

  Of these, the first alignment means 31 aligns the entire image of the printed circuit board 8 acquired by the image acquisition means 2 in a rough state, as shown in FIG. Specifically, the entire image of the printed circuit board 8 is made to correspond to the first resolution, which is a low resolution, while being shifted in the vertical direction, the horizontal direction, and the rotation direction, and the position where the degree of coincidence with the reference image becomes the maximum is correlated (Fig. 2 (a)). Then, the entire image of the printed circuit board 8 is corrected in the vertical direction, the horizontal direction, and the rotation direction based on the position where the degree of coincidence is greatest. Next, an image of the printed circuit board 8 is extracted at a second resolution, which is higher than the first resolution, and the position in each of the vertical direction, the horizontal direction, and the rotation direction is similarly based on the image. The position where the degree of coincidence is maximized is calculated by cross-correlation. Then, based on the position, the image of the second resolution is corrected in the vertical direction, the horizontal direction, and the rotation direction (FIG. 2B). In the present embodiment, the first and second resolutions are roughly aligned. However, the overall alignment may be performed with a higher resolution image, or the first The entire alignment may be performed with only the resolution. In this way, the entire image is aligned with a relatively low resolution.

  Next, the second alignment means 32 divides the image of the printed circuit board 8 aligned in a rough state into small areas (FIG. 3), and extracts only areas where the luminance change amount is large among the areas. Then, the areas are aligned independently. In general, in the printed circuit board 8, a region with little luminance change such as a resist-only portion (A1 in FIG. 3) and a region (A2) consisting only of a wiring pattern and a pad portion occupies most of the region. There is very little about the area | region which consists of the edge of a big wiring pattern, and the edge (A3) of a pad. For this reason, the entire calculation amount can be reduced by performing the alignment only for the region having a small ratio.

  When it is detected whether or not the luminance change is large for each of the divided areas, the luminance information of each pixel in the divided area is output to the luminance change detecting means 4, where the luminance change amount is a predetermined threshold value. Is detected. When detecting the amount of change in luminance, as a first method, the luminance for each RGB of each pixel in the region is extracted, and the variance value is calculated. At this time, if the variance value is larger than a predetermined threshold value, it is determined that the region has a large change amount. Conversely, if the variance value of luminance in the region is smaller than the predetermined threshold value, Judged as “small area”. As this “region with a large amount of change”, for example, a region (A3) including an edge of a pad or a wiring pattern can be considered, and as the “region with a small amount of change”, for example, a resist-only region, a substrate Only the central region of the pad and the wiring pattern can be considered (A1, A2). Here, although the variation in luminance is detected using the variance value, the width of the maximum luminance value and the minimum luminance value, or the number of pixels having a luminance value equal to or higher than the predetermined luminance value and the predetermined luminance value. The number of pixels having a luminance value equal to or lower than the luminance value may be used.

  The second alignment means 32 extracts only the region determined as “region having a large change amount” in this way, and aligns it with the region of the reference image corresponding to the region. In this alignment, a portion with a large luminance change (for example, an edge portion of a pad) is extracted from the extracted region, and the portion is aligned with a small region of the reference image using that portion as a reference. Specifically, since silk generally has a large positional misalignment, a color (RGB) or luminance area excluding silk is extracted, and the pixels in the extracted area are expanded by several pixels vertically and horizontally, An inflated edge region (A4) as shown in FIG. 4 is extracted. On the other hand, for the corresponding region of the reference image, contour regions that have been inflated by the same method are extracted, and the position having the highest matching rate is cross-correlated while shifting in the vertical direction, the horizontal direction, and the rotational direction. calculate. Then, alignment in the vertical direction, the horizontal direction, and the rotation direction is performed from the position with the highest matching rate. These processes are performed independently only for the areas determined as “regions with a large amount of change”, and these processes are not performed for the areas determined as “regions with a small amount of change”.

  Next, the third alignment means 33 performs alignment processing for the area determined as “an area with a small change amount”. The region determined to be the “region with a small amount of change” does not have a characteristic portion, and therefore cannot be aligned by the method such as the second alignment unit 32. Therefore, here, correction is performed using the position information corrected by the second alignment means 32. Specifically, if the region to be noticed is a “region with a small amount of change”, the closest “region with a large amount of change” is searched for, and correction information for the region (vertical correction, horizontal direction) And correction of the rotation direction). Then, the correction information is used to correct the area in the vertical direction, the horizontal direction, and the rotation direction. At this time, when using the correction information of one area having the largest change amount, the correction information of the area is used as it is, or when the correction information of the nearest area and the next area is used, You may make it correct | amend using the average value of correction information. In FIG. 5, two “regions with a large amount of change” having different directions with respect to the region to be corrected are searched (regions surrounded by thick solid lines), and the vertical correction is performed based on the correction information of the regions. The correction in the horizontal direction and the correction in the rotation direction are performed.

  When inspecting the inspection object, the determination unit 6 determines whether each area aligned in this way is compared with the area of the reference image, and determines whether the area is defective. The area that is being processed is output. Various methods can be used for this pass / fail judgment. For example, with reference to the position of the pixel of the aligned image, the luminance within a certain range from the luminance value of the pixel within a predetermined search distance in the reference image. If such a pixel exists, it is determined as “excellent pixel”, and if such a pixel does not exist, it is determined as “defective pixel”. Then, the same processing is performed for all the pixels, and when a predetermined number or more of defective pixels are continuous, the region is output as a “defective portion”.

  Further, the reference data generation means 7 acquires images from a plurality of inspection objects determined to be non-defective products, performs alignment of each, and generates reference data necessary for inspection. In this embodiment, as reference data, an allowable luminance width at the pixel position, a search distance for searching for a pixel corresponding to the pixel of the inspection object, and the like are generated and stored from the variation of the aligned pixels. I shall let you.

  Next, an inspection object positioning method in the automatic inspection apparatus 1 configured as described above will be described with reference to the flowchart of FIG.

  First, when inspecting the printed circuit board 8 that is the inspection object, a case where the reference image is generated first will be described. When the reference image is generated, the printed circuit board 8 that is a non-defective inspection object is placed on the stage. The image acquisition means 2 acquires the image. At this time, an image of a stage for placing the printed circuit board 8 is acquired in advance, and an image consisting only of the printed circuit board 8 is acquired by taking a difference from the image of the stage (step S1).

  Next, the acquired image is converted into a first resolution image having a low resolution, a reference image corresponding to the converted low resolution image is read, and overall alignment is performed. In this alignment, the vertical cross-correlation is used to detect the vertical shift, the horizontal cross-correlation is used to detect the horizontal shift, and the image is rotated to rotate. The amount of deviation in the rotational direction is also detected by the cross-correlation of directions (when n = 1 in step S2). Then, based on these deviation amounts, the deviation of the image of the printed circuit board 8 as the inspection object is corrected (step S3).

  Next, the image corrected at the first resolution is aligned at a second resolution that is relatively higher than the first resolution (step S4: No). Similarly, in the alignment at the second resolution, a reference image corresponding to the second resolution is read, and a vertical shift amount, a horizontal shift amount, and a rotational shift amount are detected by cross-correlation. (Step S2). Then, based on the detected deviation amount, the deviation amount of the image of the printed circuit board 8 as the inspection object is corrected (step S3).

  Next, a division process is performed on the image of the inspection object aligned with such a rough image (step S5). Then, a luminance change amount indicating how much the luminance change of each pixel is in the area is obtained (step S6). When calculating this amount of change in luminance, the variance value of the luminance of each pixel in that region is obtained. If the variance value is equal to or greater than a predetermined threshold, it is determined that the region has a large amount of change, and on the contrary If the value is less than the predetermined threshold value, it is determined as “a region with a small amount of change” (step S7).

  Next, an area determined to be the “region with a large amount of change” is extracted (step S7: Yes), and a portion with a large amount of change, such as an edge portion of the pad, is extracted from the region and inflated. (Steps S8 and S9). On the other hand, for the reference image corresponding to the region, the portion with a large amount of change is extracted and inflated, cross-correlation is performed for each image, and the degree of coincidence calculated by the cross-correlation is calculated. The amount of deviation is detected from the information of the high position (step S10), and the region of the inspection object is independently corrected (step S11).

  On the other hand, when it is determined that the region is a region with a small amount of change (step S7: No), the closest region with a large amount of variation and the next region with a large amount of variation are extracted ( Step S12), the deviation amounts in the vertical direction, the horizontal direction, and the rotation direction in each region are extracted. Then, the position of the region is corrected from the average value of the respective shift amounts (step S13).

  Then, the same processing is performed on the plurality of printed circuit boards 8, and the luminance for each RGB is extracted for each pixel of the aligned image (step T1 in FIG. 7), and allowed for each pixel position. The brightness range to be determined is determined. When determining the luminance width, a dispersion value or standard deviation that is a variation in RGB luminance of pixels of the printed circuit board 8 that has read a plurality of sheets is obtained (step T2). The allowable range from the luminance is set large and stored as reference data for the pixel. When setting the permissible luminance width, the RGB luminance is converted into a polar coordinate system so that each luminance can be expressed by a distance r from the origin and an angle θ formed with a straight line passing through the origin (step T3). Then, the allowable luminance width is set by changing the distance r and the angle θ (step T4). If the allowable luminance width is set using the distance r and the angle θ in this way, the parameters are reduced and the hue (ratio of RGB luminance) is changed as compared with the case where the allowable luminance width is set in the orthogonal coordinate system. There is an advantage that only brightness and saturation can be changed without any change.

  On the other hand, corresponding to the pixel, the search distance of the corresponding pixel of the printed circuit board 8 in the inspection object is stored as a table (step T5), and the pixel having the luminance within the allowable luminance width within the search distance from the position of the pixel. It is possible to check whether or not exists.

  Next, processing in the case of inspecting the printed circuit board 8 to be inspected in the state where the reference image and the reference data are stored in this way will be described with reference to FIGS.

  When inspecting the printed circuit board 8 that is an inspection object, similarly, the printed circuit board 8 is placed on the stage, and the image acquisition unit 2 acquires an image including only the printed circuit board 8. (Step S1).

  Next, the acquired image is converted into an image of a first resolution which is a low resolution, and a reference image corresponding to the converted low resolution image is read out to cross-correlate in the vertical, horizontal and rotational directions. To correct the amount of deviation in each direction (steps S2 to S3). Similarly, the image corrected at the first resolution is aligned at a second resolution that is relatively higher than the first resolution (step S4, step S2 to step S3).

  Next, the image of the inspection object aligned with such a rough image is divided (step S5), and the luminance change amount in the region is obtained to obtain “region with large change amount” and “change amount”. Are divided into “small regions” (steps S6 and S7).

  Then, an area determined as “an area having a large change amount” is extracted (step S8), an area excluding silk color and luminance is extracted for the area, and the change amount of the pixel in the extracted area is extracted. A large portion is extracted and inflated (step S9), and a reference image corresponding to the region is also extracted for a portion with a large amount of change, inflated, and a cross-correlation is performed for each image to obtain a deviation amount. Is detected (step S10). Then, the region of the inspection object is independently corrected from the position information with the highest degree of coincidence calculated by the cross-correlation (step S11), and the region is inspected using the reference image and the reference data (FIG. 8). When the inspection is performed for each region, the allowable luminance width and the search distance stored in advance are extracted from the position of the reference image corresponding to the position of each pixel of the inspection object (step U1), and the search distance of the reference image is extracted. It is checked whether or not there is a pixel having a luminance (luminance for each RGB) within the allowable luminance width (step U2). If such a pixel exists, it is determined as “good pixel” (step U3), and if such a pixel does not exist, it is determined as “defective pixel” (step U4). If a predetermined number or more of defective pixels are continuously present, a defective area is output (step U5).

  Next, an area determined to be a “region with a small amount of change” is extracted (step S12), and the “region with a large amount of change” closest to the region and the “region with a large amount of change” next to each other are extracted. Then, the shift amounts in the vertical direction, the horizontal direction, and the rotation direction in each region are extracted. Then, the position of the region is corrected from the average value of the respective shift amounts (step S13). Similarly, this region is inspected using the reference image and the reference data (step U1). Even when the inspection is performed for each region, the permissible luminance width and the search distance are extracted from the position of the reference image corresponding to the position of each pixel of the inspection target (step U1), and the permissible within the search distance of the reference image. It is checked whether or not there is a pixel within the luminance width (step U2). When such a pixel exists, it is determined as “good pixel” (step U3), and when such a pixel does not exist, it is determined as “defective pixel” (step U4). If a predetermined number or more of defective pixels are continuously present, a defective area is output (step U5).

As described above, according to the above embodiment, in the automatic inspection apparatus 1 that determines the quality of the inspection object by aligning the image acquired from the inspection object and the reference image, the entire image is obtained with a predetermined resolution. First alignment means 31 for alignment with the reference image, and a partial area of the entire image aligned by the first alignment means 31 are extracted, and the amount of change in luminance value in the area is a predetermined value A change amount detecting means for determining whether or not the difference is present, and a second alignment for aligning the corresponding area of the reference image with respect to an area where the change amount of the luminance value is equal to or greater than a predetermined value by the change amount detecting means. Means 32 for calculating a variation of each pixel based on the aligned image, determining an allowable luminance width at the position of the pixel from the variation, and the allowable luminance width; Since the reference data generating means for storing the data in the polar coordinate system is provided, the amount of calculation can be reduced and accurate alignment can be performed. Moreover, when generating the reference data , the reference data can be generated accurately, and if the permissible luminance width is expressed in a polar coordinate system, only the brightness changes and the ratio of saturation and hue changes. Even if it is not, the reference data can be easily generated.

  In addition, this invention is not limited to the said embodiment, It can implement in a various aspect.

  For example, in the above embodiment, the region determined to be “the region with small change amount” is aligned based on the information of the two adjacent regions. At this time, depending on the distance between the respective regions You may make it correct the deviation | shift amount of a vertical direction, a horizontal direction, and a rotation direction. That is, when the distance to the “region with a large amount of change” is long, there is a possibility that the influence of the shift amount in that region may be reduced, so the amount of change may be corrected in inverse proportion to the distance.

  In addition, when calculating the shift amount of the “region with a small change amount” based on the shift amounts of the plurality of regions in this way, the “region with the large change amount” closest to one direction and a direction different from the direction ( For example, the closest “region with a large change amount” in the orthogonal direction may be extracted, and the shift amount of the region may be calculated using these shift amounts. In this way, there is an advantage that it is possible to accurately correct the shift amount and rotation amount between the vertical direction and the horizontal direction.

  Furthermore, in the above embodiment, each of the “region with a large amount of change” and the “region with a small amount of change” is inspected, but only the “region with a large amount of change” is extracted and inspected. May be. Alternatively, in a “region with a small amount of change”, for a relatively unimportant portion such as a region consisting only of resist, the region is extracted from a predetermined RGB luminance value, and the average luminance value of the region is the luminance. If it is included in the value range, the area may not be inspected.

  In addition, in the above embodiment, two types of images of the first resolution and the second resolution are stored as the reference image, but a high-resolution image is stored and lower than that. When positioning with a resolution image, the high resolution image may be converted into a low resolution and used.

  In the above embodiment, when the second alignment unit 32 performs alignment, a characteristic part of the region is extracted and aligned. You may make it align by taking a correlation.

  In addition, when performing the second alignment in the above embodiment, an area excluding silk color and luminance is extracted, and a portion with a large luminance change near the contour is generated by expansion / contraction processing and synthesis thereof. Alternatively, a difference between the RGB luminance values of adjacent pixels may be obtained, and the same processing may be performed while leaving pixels whose difference is equal to or greater than a predetermined value.

  The present invention is used in the field of a printed circuit board, a liquid crystal substrate, and an automatic inspection apparatus that inspects the formation state of patterns and characters formed on the surface of an object.

DESCRIPTION OF SYMBOLS 1 ... Automatic inspection apparatus 2 ... Image acquisition means 3 ... Positioning means 31 ... 1st positioning means 32 ... 2nd positioning means 33 ... 3rd positioning Means 4 ... Brightness change calculation means 5 ... Reference image storage means 6 ... Determination means 7 ... Reference data generation means 8 ... Printed circuit board

Claims (2)

In the automatic inspection device that can determine the quality of the inspection object by aligning the image acquired from the inspection object and the reference image,
First alignment means for aligning the entire image with a reference image at a predetermined resolution;
A change amount detection unit that divides the entire image aligned by the first alignment unit and determines whether or not the change amount of the luminance value in the divided region is equal to or greater than a predetermined value;
A second alignment means for aligning the corresponding area of the reference image with respect to the area where the amount of change in the luminance value is equal to or greater than a predetermined value;
Reference data generating means for calculating a variation of each pixel based on the aligned image, determining an allowable luminance width at the position of the pixel from the variation, and storing the allowable luminance width in a polar coordinate system;
An automatic inspection device characterized by comprising: In the alignment method in the automatic inspection apparatus that aligns the image acquired from the inspection object and the reference image and determines the quality of the inspection object,
Aligning the entire image with a reference image at a predetermined resolution;
Extracting a partial region of the entire image aligned by the step, and determining whether or not the amount of change in luminance value for the region is equal to or greater than a predetermined value;
Aligning the corresponding region of the reference image with respect to the region where the amount of change in the luminance value is equal to or greater than a predetermined value by the determination;
Calculating a variation of each pixel based on the aligned image, determining an allowable luminance width at the position of the pixel from the variation, and storing the allowable luminance width in a polar coordinate system;
An alignment method in an automatic inspection apparatus, comprising:

Images