TWI618925B - Defect inspection method and defect inspection system - Google Patents

Abstract

With a simple structure, defects of the object to be inspected are easily and reliably detected.

The defect inspection method includes a process of photographing an object to be inspected (W) to obtain a photographic portrait, and arranging two color components of the color components of the pixels of the photographic portrait in red, green, and blue to become the vertical axis and the horizontal axis. Project on dimensional scatter diagram. The pixels of the defect distribution area and the noise distribution area are divided on the two-dimensional scatter diagram, and the pixels of the divided defect distribution area are selected. The selected pixel is applied to the place where the pixel exists to generate a limited photographic image, and the limited photographic image is used to perform defect inspection.

Description

Defect inspection method and defect inspection system

This embodiment relates to a defect inspection method and a defect inspection system for inspecting defects in photographic portraits using an image processing technique, and more particularly to a defect inspection method and a defect inspection system that simply distinguish between defects and noises in an image and detect real defects.

Various methods have conventionally been known as defect inspection methods for photographing the surface of an object to be inspected using an imaging device such as a CCD camera, and inspecting defects from photographic images.

However, any defect inspection method requires complicated image processing, and the problem of defects is not detected sharply.

The present embodiment is created in consideration of such a point, and aims to provide a defect inspection method and a defect inspection system that can easily and reliably detect defects of an inspection object with a simple structure.

This embodiment is a defect inspection method, which is characterized in that it includes: a lighting project that irradiates light to the surface of the object to be inspected; a photography project that photographs the surface of the object to obtain a photographic image; Extract the defect distribution area from which the defect should be detected and the noise distribution area where no defect is required to be detected from the photographic portrait. At the same time, the color components of the pixels existing in each area are arranged in two color components of red, green and blue. Become the two-dimensional scatter diagram on the vertical and horizontal axis; the segmentation project is to divide the pixels constituting the defect distribution area and the pixels constituting the noise distribution area on the two-dimensional scatter diagram by dividing straight lines; the selection project is selected in The pixels in the two areas that are divided by the straight line on the secondary scatter diagram belong to the side of the defect distribution area; the limited project is that the pixels selected in the selection process are applicable to the presence of the A limited photographic image is generated at a pixel location; and an inspection implementation process is performed using a limited photographic image to perform defect inspection.

This embodiment is a defect inspection method, which uses light as white light as a feature.

This embodiment is a defect inspection method, which is characterized by two or more different colors of light.

This embodiment is a defect inspection method, which is characterized by the first color light, the second color light, and the third color light with different colors of light.

This embodiment is a defect inspection method, which is characterized in that the first color light, the second color light, and the third color light are any one of red, green, and blue, respectively.

This embodiment is a defect inspection method, which is characterized by different heights of the positions where the first color light, the second color light, and the third color light are irradiated to the inspection object by the lighting process.

This embodiment is a defect inspection method. When the vertical axis of the two-dimensional scatter diagram is set to y, the horizontal axis is set to x, and a and b are set to real numbers, the linear equation y = ax + b means.

This embodiment is a defect inspection system, which is characterized by including: an illuminating device that irradiates light to the surface of the object to be inspected; a photography process that photographs the surface of the object to obtain a photographic image; and a defect inspection device, Defect inspection is performed by subjecting a photographic image from a photographic device to image processing. The defect inspection device is provided with a configuration section that extracts from the photographic image a defect distribution area in which a defect should be detected and noise that does not need to be detected. Distributing the area, and arranging the color components of pixels in each area on the two-dimensional scatter diagram of the two color components of red, green, and blue to form the vertical and horizontal and horizontal axes; the division section is based on the two-dimensional scatter diagram The pixels constituting the defect distribution area and the pixels constituting the noise distribution area are divided by the division line; the selection unit selects the defect distribution area side of the two areas that are divided by the division line on the secondary scatter diagram. Pixels; the limitation section, which applies the pixels selected in the selection process to the limitation where the pixels exist in the photographic portrait Movies portrait; and inspection execution unit that defines a system used to implement the defect inspection and portrait photography.

This embodiment is a defect inspection system, which uses light as white light as a feature.

This embodiment is a defect inspection system The same two or more colors are characteristic.

This embodiment is a defect inspection system, which is characterized by the first color light, the second color light, and the third color light with different colors of light.

This embodiment is a defect inspection system, which is characterized in that the first color light, the second color light, and the third color light are any one of red, green, and blue, respectively.

This embodiment is a defect inspection system, which is characterized by the difference in height of the positions where the first color light, the second color light, and the third color light are irradiated to the inspection object by the lighting process.

This embodiment is a defect inspection system. When the vertical axis of the two-dimensional scatter diagram is set to y, the horizontal axis is set to x, and a and b are set to real numbers, the linear equation y = ax + b means.

As described above, if this embodiment is used, it is possible to more easily and surely detect the defect of the detected object with a simple structure.

1‧‧‧frame

1b‧‧‧large opening

1n‧‧‧small opening

1s‧‧‧Concave inside

2a‧‧‧Light incident department

10a, 10b‧‧‧Lighting device

20‧‧‧ Camera

Dd‧‧‧Defect distribution area

Dn‧‧‧ Noise distribution area

LH‧‧‧High Position Lighting

LI, Liw‧‧‧ incident light

LIH‧‧‧ high-level incident light

LIL‧‧‧ low-level incident light

LIM‧‧‧Injecting light in the middle position

LM‧‧‧Intermediate position lighting

LL‧‧‧Low Position Lighting

LR‧‧‧Reflected light

Lw‧‧‧White Light Emitting Diode

W‧‧‧ Workpiece

FIG. 1 is a flowchart of a defect inspection method according to this embodiment.

2 (a), (b), and (c) are explanatory diagrams of the principle of the embodiment.

FIG. 3 is an explanatory diagram of the principle of the embodiment.

FIG. 4 is an explanatory diagram of the principle of the embodiment.

FIG. 5 is an explanatory diagram of the principle of the embodiment.

Fig. 6 is a perspective view of a wafer-type electronic component.

7 (a) and 7 (b) are explanatory diagrams of an illuminating device for irradiating a wafer-type electronic component with illumination light and a photographing device for the surface of the wafer-type electronic component.

FIG. 8 is a flowchart of a defect inspection method according to Comparative Example 1. FIG.

9 (a), (b), (c), and (d) are explanatory diagrams for obtaining red, green, and blue portraits from photographic portraits.

Figures 10 (a), (b), (c), (d), (e), (f), (g), (h), and (i) show the image processing in the defect inspection method of Comparative Example 1. Illustrating.

11 (a) and 11 (b) are explanatory diagrams of an illuminating device for irradiating a wafer-type electronic component with illumination light and a photographing device for obtaining a wafer-type electronic component using the same.

FIG. 12 is an explanatory diagram of image processing in the defect inspection method of Comparative Example 2. FIG.

FIG. 13 is an explanatory diagram of image processing in the defect inspection method of Comparative Example 3. FIG.

[Implementation type]

Hereinafter, the defect inspection method and the implementation form of the defect inspection system will be described with reference to the drawings. First, for wafers that are to be inspected An electronic component (hereinafter, also referred to as a “workpiece”) will be described with reference to FIG. 6. As shown in FIG. 6, the workpiece W has a hexahedral shape, has a body Wd made of an insulator, and electrodes Wa, Wb made of a conductor formed at both ends in the longitudinal direction of the body Wd. .

An element such as a resistor or a capacitor is formed in the body Wd, and the connection between the element and the external circuit is performed by connecting the electrodes Wa and Wb to the external circuit.

Next, a defect inspection method with respect to the surface of the workpiece W will be described. First, the work W is irradiated with white light as illumination light, and the surface is photographed.

Fig. 7 shows an illuminating device for illuminating illuminating light and an imaging device for imaging the surface of a workpiece. Here, FIG. 7 (a) is a perspective view of the lighting device viewed from below, and FIG. 7 (b) is a side view showing a state of photographing the surface of the workpiece using the lighting device and the imaging device.

In FIG. 7 (a), the lighting device 10a has a frame 1 formed in a hemispherical shape, and the center side of the hemisphere of the frame 1 has a lower surface 1d formed in a plane. Further, the inside of the hemisphere that becomes the inner side of the frame body 1 as viewed from the lower surface 1d forms an inner concave surface 1s having the same hemispherical shape as the outer shape of the frame body 1. In addition, the center side of the hemisphere formed in the housing 1 is a large opening portion 1b that is opened on the lower surface 1d. In addition, when the inner concave surface 1s of the frame body 1 is viewed from the large opening portion 1b side, at the center portion of the circular frame body 1, the small circular opening portion 1n having a larger diameter opening portion 1b penetrates the frame body 1 And formed. Here, the shape of the small opening portion 1n in FIG. 7 (a) is circular, and the shape of the small opening portion 1n is not limited to a circular shape, and it will be from a workpiece as described later. The reflected light may pass through the shape of the frame 1. On the surface of the inner concave surface 1s, white light emitting diodes (LEDs) Lw capable of radiating most of white light toward the large opening portion 1b are arranged at equal intervals in a concentric circle shape. The center of the concentric circles is the same as the center of the small opening 1n, and the number of the concentric circles is all seven.

In FIG. 7 (b), the illuminating device 10a is arrange | positioned so that the lower surface 1d and the large opening part 1b may become downward, and the small opening part 1n may become upward. Furthermore, the workpiece W is placed on a surface above the mounting table (not shown) at a position just below the center position of the large opening 1b, just below the center 1d of the frame 1 and a large distance Wu and the large opening. The parts 1b face each other.

In addition, directly above the lighting device 10a and above the small opening portion 1n, the photographing device 20 is arranged so that the light incident portion 2a faces the small opening portion 1n. The light emitted from the workpiece W is taken into the imaging device 20 through the light incident portion 2a, and a photographed image can be obtained by this.

The photographic image acquired by the imaging device 20 is then sent to the defect inspection device 20A, and the defect inspection device 20A performs defect inspection. A monitor 20B is connected to the defect inspection device 20A.

In this case, the lighting device 10a, the imaging device 20, the defect inspection device 20A, and the monitor 20B constitute a defect inspection system.

Next, as shown in FIG. 7 (b), a procedure of a defect inspection method for obtaining a photographic image of one surface Wu of the workpiece W and detecting a defect from the photographic image will be described using the flowchart of FIG. 1. First, like S101 (lighting process) in FIG. 1, a white light-emitting diode Lw arranged in the lighting device 10a shown in FIG. 7 (b) is used to photograph the workpiece W. Shoot white light. At this time, the incident light LI emitted from the white light-emitting diode Lw is irradiated to one surface Wu of the workpiece W, and is reflected there to become reflected light LR. The reflected light LR is taken into the imaging device 20 from the light incident portion 2a through the small opening 1n of the frame 1 located above the work W. Accordingly, as shown in S102 (photographic process) in FIG. 1, one surface Wu (upper surface) of the workpiece W is photographed to obtain a photographic image.

Next, the photographic image acquired by the imaging device 20 is then sent to the defect inspection device 20A, and the defect inspection device 20A performs defect inspection.

That is, as for the photographic image obtained by the photographing device 20, as in S33 (extraction process), from the photographic image, a defect distribution area where a defect should be detected and a noise distribution where no defect is required to be detected are extracted. region. This state is shown in Fig. 2 (a). The portrait shown in FIG. 2 (a) is displayed on the monitor 20B, and is a photographic portrait of the workpiece W obtained at S102 in FIG. While the operator visually displays the image displayed on the monitor 20B, the operator extracts the defect distribution area to be detected and marks it. Here, an area Ae surrounded by a one-dot chain line is extracted as a defect distribution area and marked. In addition, all areas other than the area Ae become noise distribution areas where no defect is required to be detected.

In this way, when two types of areas are distinguished, S34 (configuration process) of FIG. 1 is performed next. Here, the color components of all pixels in the defect distribution area and noise distribution area in the photographic image are arranged with the blue component as the vertical axis and the red color component as the horizontal axis. Dimensional scatter plot.

The result of performing S34 of FIG. 1 on the photographic image of FIG. 2 (a) is shown in FIG. 2 (b). The area enclosed by the one-dot chain line Dd in FIG. 2 (b) is for the pixels constituting the defect distribution area in FIG. 2 (a), that is, Ae, and is arranged on the second-dimensional scatter diagram. Furthermore, the area surrounded by the one-dot chain line Dn in FIG. 2 (b) is the one in which the pixels constituting the noise distribution area in FIG. 2 (a) are arranged on the second-dimensional scatter diagram. Here, because the pixels constituting the defect distribution area Dd in FIG. 2 (b) belong to the area Ae in FIG. 2 (a), directly above the secondary scatter diagram in FIG. 2 (b), the set is used. The symbol is written as Dd = {Ae}.

In this way, when S34 in FIG. 1 ends, then on the two-dimensional scatter diagram made in S34, as shown in S35 (segmentation process), the picture constituting the defect distribution area is divided by a straight line (divided line). Pixel and noise distribution area. This division is also shown in Fig. 2 (b). In FIG. 2 (b), the shapes of the defect distribution area Dd and the noise distribution area Dn are both elliptical shapes having a long axis in the approximate upper right direction. In the case where two regions have such regions, imagine that these two regions exist in the first quadrant formed by the positive direction of the vertical axis (B axis) and the positive direction of the horizontal axis (R axis), and draw A straight line dividing the defect distribution area Dd and the noise distribution area Dn as shown in FIG. 2 (b) is obtained. If the inclination of the straight line is a, and the coordinate where the straight line intersects the vertical axis (B-axis) is b, the expression that B and R express the straight line is expressed by B = aR + b (1). The b becomes a segment for dividing the defect distribution area Dd and the noise distribution. Threshold of area Dn. At this time, the expression system B> aR + b (2) indicating that the straight line is the defect distribution area Dd is, that is, B-aR-b> 0 (3)

It is obvious from FIG. 2 (b) that all pixels constituting the noise distribution area Dn in the area of the formula (3) are excluded.

Next, as shown in S36 (selection process) in FIG. 1, among the two regions divided by a straight line in FIG. 2 (b), all pixels belonging to the area side that does not include pixels belonging to the noise distribution area are selected. , That is, all pixels in the defect distribution area Dd (formula (3)).

Then, as shown in S37 (Limited Project) in FIG. 1, the pixels constituting the selected defect distribution area Dd are applied to the photographic portrait where the pixels exist, and the defects limited to the selection are generated. Limited photographic portrait of pixels in the distribution area Dd.

The limited photographic image generated from FIG. 2 (b) at S37 is shown in FIG. 2 (c). This limited photographic image is displayed on the monitor 20B. The limited photographic image obtained in FIG. 2 (c) does not include pixels other than the area Ae in FIG. 2 (a). Therefore, after the process of S37 in FIG. 1 is completed, the operator uses the limited photographic image of FIG. 2 (c) obtained in S37 to inspect the defect while visually monitoring the monitor 20B as in S38 (inspection execution process). Accordingly, it is possible to perform a high-precision defect inspection by excluding pixels constituting the noise distribution area Dn in FIG. 2 (b).

In addition, in S35 (segmentation process), in order to determine a straight line, the operator has a method of determining the two-dimensional scattergram visually and uses a discriminant analysis known from the past as image processing software. A straight line system in the case of using this discriminant analysis is determined as, for example, a straight line from which the Mahalanobis distance of any one of the defect category and the noise level category becomes equal.

Here, the principle of the present invention shown by S33 to S37 in FIG. 1 will be described using FIG. 3. Fig. 3 is a schematic diagram of Fig. 2 (b) showing the principle of the present invention. In the first quadrant formed by the positive direction of the vertical axis (B axis) and the positive direction of the horizontal axis (R axis), there is a noise distribution area Dd surrounded by a solid line and noise surrounded by a dotted line. Each of the distribution regions Dn has an ellipse shape having a long axis in a generally upper right direction.

In FIG. 3, imagine a straight line passing B = P as the origin and a straight line represented by a dotted line at the upper right of 45 °. It is assumed that the defect distribution area Dd and the noise distribution area Dn are divided by moving the straight line by a straight line (B = R + BT) of the solid line of the critical value BT (positive number) in the positive direction of the vertical axis. Above the straight line is a range of B> R + BT, and there is no noise distribution area Dn. Here, the pixels constituting the defect distribution area Dd existing in the range of B> R + BT can be said to be all pixels exhibiting true defects. That is, the range of B> R + BT becomes a target area for defect inspection. By selecting all the pixels in the area, as shown in S37 in FIG. 1, it is applicable to the place where the pixel exists in the photographic portrait, and a limited photographic portrait can be generated. Moreover, the limited photographic image generated in this way does not include pixels in the noise distribution area Dn in FIG. 3 at all. Therefore, the operator can look The video monitor 20B performs a high-accuracy defect inspection using a limited photographic image.

Thus, with this embodiment, it is possible to easily and reliably perform defect inspection without performing complicated image processing.

In addition, in the above description, for simplicity, although the inclination of the straight line that divides the defect distribution area Dd and the noise distribution area Dn is set to 45 °, the actual shape of the defect distribution area Dd and the noise distribution area Dn The inclination changes into various kinds. For example, the inclination of the straight line shown in FIG. 2 (b) is a. If a is a value other than 1, the inclination of the straight line is no longer -45 °.

In the above description, the LEDs arranged in the lighting device 10a shown in FIGS. 7 (a) and 7 (b) are described in white. However, the light irradiated to the work W in FIGS. 7 (a) and (b) is not limited to white. As another example, there are lighting devices using red, green, and blue LEDs as the three primary colors of light. FIG. 11 shows a photographing device for photographing the illumination device and the surface of the workpiece. Fig. 11 (a) is a perspective view of the lighting device viewed from below. In addition, FIG. 11 (b) is a side view of a state where the surface of the object to be inspected is imaged using the illumination device and the imaging device. 11 (a), (b) and FIGS. 7 (a), (b) differ only in the color of the LEDs, so detailed descriptions of the portions that do not have differences are omitted.

In FIG. 11 (a), light emitting diodes (LEDs) are arranged concentrically on the surface of the inner concave surface 1s of the lighting device 10b. The number of concentric circles is all seven. The red (R) LEDs are arranged on the four concentric circles close to the large opening 1b at approximately equal intervals to constitute a low-position lighting LL. with In a similar manner, green (G) LEDs are arranged at equal intervals on concentric circles No. 5 and No. 6 from the side of the large opening portion 1b to constitute an intermediate position lighting LM. In addition, the blue (B) LEDs are arranged at equal intervals on the concentric circle No. 7 from the side of the large opening 1b, which is the concentric circle closest to the position of the small opening 1b. Lighting LH.

In FIG. 11 (b), the lighting device 10b is arranged such that the lower surface 1d and the large opening portion 1b become downward, and the small opening portion 1n becomes upward. Furthermore, the shape of the hexahedron as the object to be inspected is placed just below the center of the large opening 1b, at a distance from the lower surface 1d of the frame 1, and placed on a mounting table (not shown). The wafer-type electronic component (workpiece) W is positioned such that one surface Wu thereof faces the large opening portion 1b.

The flowchart for the case of using the lighting device 10b and the photographing device 20 of FIG. 11 is the same as that of FIG. Here, effects produced by arranging three types of LEDs of red, green, and blue by the lighting device 10b will be described.

In FIG. 11 (b), the incident light LI emitted from each illumination (LED) is irradiated to one surface Wu of the work W. Here, due to the positions of the low-position illumination LL, the intermediate-position illumination LM, and the high-position illumination LH, among the incident light LI, the low-position incident light LIL emitted from the low-position illumination LL is to face the workpiece W from a low angle Wu shines. Similarly, among the incident light LI, the intermediate position incident light LIM emitted from the intermediate position illumination LM is irradiated to one surface Wu of the workpiece W from an angle where the incident light LIL is lower from a lower position. In addition, among the incident light LI, the high-level incident light LIH emitted from the high-position illumination LH is one of the workpieces W from an angle close to a vertical height. Surface Wu irradiation. In this way, the incident light LI is composed of different types of incident light having different incident angles. The incident light LI is reflected on one surface Wu of the workpiece W to become reflected light LR, and the photographed image is obtained by the imaging device 20 taking the reflected light LR. Therefore, three types of monochrome portraits obtained from photographic portraits are incident from different heights (angles) of the color light of each monochrome portrait onto one surface Wu of the workpiece W, which is a reflector. That is, each monochrome image shows the inherent reflection characteristic corresponding to the shape of the surface Wu. This characteristic is referred to as an angular response, and a monochrome image which is considered to best represent the angular response of the defect of one surface Wu of the workpiece W or an image obtained as a result of calculation between the monochrome images is selected as an inspection image (Fig. 8). S105 or S106), defect inspection can be performed. A lighting arrangement that can respond to such an angle has been used in the past.

Furthermore, when the principle of the present invention is described using FIG. 3, the defect distribution area Dd and the noise distribution area Dn do not have a common portion, so it is assumed that they can be completely divided by the straight line B = R + BT. However, when a two-dimensional scatter diagram like that shown in FIG. 2 (b) is created using a photographic image different from that shown in FIG. 2 (a), there are also defect distribution areas Dd and noise distribution areas Dn. An example of the method of dividing the two regions at this time will be described using FIG. 4 and FIG. 5.

FIG. 4 is a two-dimensional scatter diagram created for a photographic portrait. The defect distribution area Dd and the noise distribution area Dn have a common portion Dc. Therefore, the defect distribution area Dd and the noise distribution area Dn cannot be completely separated using one straight line. Here, it is assumed that the area that is the object of defect inspection is the pixels constituting the limited photographic image, even if it cannot be selected. A full portrait belonging to the defect distribution area Dd is also considered to be a case where the method of removing the full pixels belonging to the noise distribution area from the area to be subject to defect inspection is preferable. In this case, with the straight line B = R + BTa shown in FIG. 4, the common part Dc and the noise distribution area Dn are excluded from the selected objects at the same time. Furthermore, it is only necessary to select only pixels in a region above the straight line of the defect distribution region Dd, that is, a region of B> R + BTa, to generate a limited photographic image. Next, on the contrary, suppose that even if the full pixels belonging to the noise distribution area are removed from the area that is the object of defect inspection, the area that is considered as the object of the defect inspection is the pixels that constitute the limited photographic image to select One of the full portraits belonging to the defect distribution area Dd is preferable. In this case, the common portion Dc and the defect distribution area Dd are simultaneously selected by the straight line B = R + BTb shown in FIG. 5. Furthermore, it is only necessary to select pixels that exist in a region above the straight line in the noise distribution region Dn, that is, a region of B> R + BTa, and generate a limited photographic image.

In addition, an example of a method of setting a judgment criterion for dividing as shown in FIG. 4 or dividing as shown in FIG. 5 will be described below. For example, it is considered to consider characteristics such as the shape of a defect in a photographic image of a workpiece, and the probability of misjudgement of a defect to be a normal one at the time of defect inspection P1 and the probability of misjudgement of a normal one to a defect P2 and set the benchmark. In the case of defects having the characteristics of P1> P2, it is considered that the number of defects that cannot be misjudged when the area to be inspected for defects includes as many defect distribution areas as possible can be considered. In this case, the division may be performed as shown in FIG. 5. In addition, in the case of a defect having the characteristics of P1 <P2, it is considered that it can be increased if the defect detection becomes Examine the area of the object to exclude as many noise distribution areas as possible without being misjudged. In this case, it is sufficient to divide as shown in FIG. 4.

In the above description, although the two-dimensional scatter diagram in the present invention has been prepared for blue and red, it is possible to make a two-dimensional scatter diagram for green and blue or a two-dimensional scatter diagram for red and green. At this time, from the created scatter diagrams, a scatter diagram that can most easily divide the defect distribution area and the noise distribution area by a straight line is selected. In addition, it is only necessary to use the scatter diagram to select an image of a defect distribution area to create a limited photographic image. At this time, a straight line is generally expressed by the linear equation y = ax + b when the vertical axis of the two-dimensional scatter diagram is set to y, the horizontal axis is set to x, and a and b are set to real numbers. In the above-mentioned embodiment, since the vertical axis is blue (B) and the horizontal axis is red (R), this linear equation represents B = aR + b (1) as follows.

In the above description, although only one region in the photographic image is assumed to be the defect distribution region, even a region composed of two or more plural numbers in the photographic image may be assumed as the defect distribution region.

Moreover, in the above description, when the defect distribution area and the noise distribution area in the two-dimensional scatter diagram are divided by a straight line, at least all of the defect distribution area or all of the noise distribution area One side is described by including all of the two areas divided by a straight line. However, the division by straight lines is not limited to these. For example, in FIG. 4 or FIG. 5, a straight line may be drawn to further divide the common portion of the defect distribution area and the noise distribution area into two.

Furthermore, in the above description, although the low-level illumination of the workpiece is illuminated in red, the intermediate position is illuminated in green, and the high-level illumination is described in blue, the positions and colors of the three illuminations correspond to each other. Not limited to this.

Furthermore, in the above description, although the case where the light irradiated to the workpiece is composed of one color of white light and the case of three colors of red, green, and blue is described, the color of the irradiated light is different. The type is not limited to these, and any two colors or more may be used.

[Comparative Example 1]

Next, Comparative Example 1 in this embodiment will be described with reference to a flowchart of FIG. 8.

In the flowchart of FIG. 8, first, the lighting device 10 a shown in FIGS. 7 (a) and (b) is arranged on the upper side of the wafer-type electronic component, and the wafer-type electronic component is irradiated (S101). Next, the imaging device 20 is used to photograph the upper surface of the wafer-type electronic component.

Next, a defect inspection method is performed by the defect inspection apparatus 20A. A monitor 20B is connected to the defective device 20A. First, as shown in the monochrome image generation process (S103), each color component of red (R), green (G), and blue (B) is extracted from the photographic image, and a red image, a green image, and a blue image are generated as Three types of monochrome portraits. The extraction of each color component is implemented by filtering each color of the photographic image by software. FIG. 9 is an explanatory diagram of a monochrome portrait generation process. Figures 9 (b), (c), and (d) show each of the photographic images shown in Figure 9 (a). Red, green, and blue portraits of color components.

Next, an inspection image selection process is performed which selects an image that seems to be able to discriminate the defect from among the three types of monochrome images generated. The inspection image selection process is equivalent to S104 and S106 in FIG. 8. First, in S104, the images displayed on the monitor 20B by visual inspection are compared to check whether there is an image among the red, green, and blue images that seems to be able to discriminate the defect. At this time, it is better to prepare a plurality of workpieces that are considered to be non-defective and a workpiece that is considered to be defective. After implementing S101 to S103, it is better to compare and review the plural monochrome portraits generated as a result of each other.

When the determination result of S104 is YES, the process proceeds to S105, and the monochrome image is selected as a check image. If the determination result of S104 is NO, go to S106, perform some calculations between the monochrome portraits, compare the results, and choose to check the portrait.

A specific example of the calculation between monochrome images will be described below. Now, the green image is described as G, the blue image is described as B, and the difference-reduced regular reflection component is described as G-B. Here, since the blue image is almost only composed of regular reflection, the image (GB image) obtained by image processing such as subtracting the brightness (GB) of the blue image from the brightness of the green image is almost removed. Portrait of regular reflection. When a defect is extracted from this G-B image, since the regular reflection is almost removed, the accuracy of extraction is improved compared to the case where the defect is extracted from the photographic image itself. However, there is no guarantee that the defects extracted from the G-B portrait are all defects on the surface of the object.

Therefore, in addition to G-B portraits by visual comparison, for example For example, an R-B image, or an addition between the brightness of each monochrome image, such as a G + R image, is selected. The image that is judged to be able to optimally extract defects is used as the final inspection image.

After selecting the inspection image in S105 or S106, proceed to S107 which is the area extraction process. Here, the selected inspection image is subjected to image processing by software, and an inspection target area to be subjected to defect inspection is extracted.

Next, the process proceeds to S108, which is a luminance histogram, and a luminance histogram is created for pixels in the inspection target area. Further, in S109, which is a binarization threshold determination process, the brightness is visually determined by the monitor 20B to determine an appropriate luminance as the binarization threshold.

FIG. 10 is an explanatory diagram for the above S107 to S109. Fig. 10 (a) is an inspection image of a good workpiece W. Here, the inspection target area of the work W is the electrode Wa. At this time, the electrode Wa is extracted from the inspection image of FIG. 10 (a) as an inspection target area. This state is schematically shown in FIG. 10 (b). The electrode Wa indicated by a solid line is extracted, and the body Wd and the electrode Wb indicated by a broken line are not extracted.

In this way, the inspection images shown in Figs. 10 (a) to 10 (b) are generated as S107 in Fig. 8. Furthermore, in S108 of FIG. 8, a luminance histogram is created for all the images included in Wa of FIG. 10 (b), as shown in FIG. 10 (c). The distribution from low brightness to high brightness can be said to be slightly regular. In addition, for a defective workpiece W having a defect, the inspection image becomes the one shown in FIG. 10 (d), and the schematic diagram of the extracted inspection target area becomes the one shown in FIG. 10 (e). The histogram of the brightness of all the images in the area is shown in FIG. 10 (f).

As shown in FIG. 10 (d) and FIG. 10 (e), there is a defect WD1 in the electrode Wa of the defective workpiece W. Since the brightness of the electrode Wa is higher, the defect WD1 has a relatively low brightness. Therefore, in FIG. 10 (f), a peak of ΔF1 (= Fd1-Fn1) appears at the degree of Ld of low brightness. When comparing FIG. 10 (f) and FIG. 10 (c), in FIG. 10 (c) showing a histogram corresponding to the electrode Wa of the good-quality workpiece W, there is no peak in the degree of brightness Ld.

In this way, when the histogram of the pixels in the inspection target area is made for the good workpiece W and the bad workpiece W by S108 in FIG. 8, the process proceeds to S109, which is a binarization threshold decision process. In S109, the histograms shown in FIG. 10 (c) and FIG. 10 (f) are viewed on the monitor 20B. Furthermore, the optimum brightness of the defect candidate area is determined as a range for extracting a range of brightness Ld including a good product and a defective product with a clear difference in degree, and used as a binarization threshold. As the binarization threshold in this case, Tp1 is appropriate.

In addition, for comparison, examples of images that are not suitable as inspection images are shown in FIGS. 10 (g) to 10 (i). When comparing the defect WD2 in FIG. 10 (g) with the defect WD1 in FIG. 10 (d), it is difficult to visually identify the defect in FIG. 10 (a) because the defect WD2 is thinner than the defect WD1 regardless of the contour or range. Good product and bad product of Fig. 10 (g). That is, the image used as the judgment criterion of S104 in FIG. 8 and the defective workpiece W as seen in FIG. 10 (g) is not suitable as an inspection image. Actually, The inspection target area extracted from FIG. 10 (g) through S107 in FIG. 8 is shown in FIG. 10 (h), and the inspection target area in FIG. 10 (h) will be the luminance histogram generated by S108 in FIG. 8 The figure is shown in Figure 10 (i).

If the degrees of the brightness Ld are compared in FIG. 10 (i) and FIG. 10 (f), it is clear that the amount of the degree of the brightness Ld change in FIG. 10 (i) to the brightness change before and after can only be called as △ Small value of the degree of the peak value of F2 (= Fd2-Fn2). That is, from the corresponding histogram of brightness, it can also be seen that the image like FIG. 10 (g) is not suitable for inspection images.

In this way, when the execution proceeds to S109 in FIG. 8, the process proceeds to S110 as a defect candidate area extraction process. In S110, the critical image Tp1 shown in FIGS. 10 (c) and 10 (f) is used to binarize the image of the inspection target area. As a result, a defect candidate area composed of an image having a brightness lower than the brightness Tp1 is extracted.

Next, as shown in S111 of FIG. 8, the pixels of the extracted defect candidate area are subjected to surface morphology processing. Here, the surface morphology treatment will be described. Surface morphology processing is a general term for image processing of binary images (white and black images). The expansion and contraction described later are combined several times and performed. The purpose is to smooth the binary image (reducing the unevenness to make it smooth), remove the isolated points (buried holes), remove the protruding parts, separate the joint parts, and combine the cut parts. Dilation refers to the process of expanding the graphics in a white and black image by 1 pixel up, down, left, and right. Shrinking refers to the process of shrinking the graphics in the white and black image by one pixel up, down, left, and right, as opposed to the expansion process described above.

As a combination of expansion and contraction treatment, there are open and close Together. Shrinking N times, and then expanding N times to break off, the removal of the protruding part of the pattern or the separation of the bonding part is effective. Contrary to opening, N times expansion and N times contraction are used to close, and the combination in the buried or cut part of the figure has an effect. By performing surface morphology treatment, the periphery of the binary image becomes smooth, the originally integrated and cut off parts are connected to each other, and the noise that has nothing to do with the image can be removed.

That is, if a binary image obtained by surface morphology processing is used, as in S112, a region having a certain area or more can be determined as a defect. S111 and S112 constitute a defect inspection process.

However, the comparative example 1 has the following problems. The defect inspection device 20A composed of the lighting devices 10a and 10b and the photographing device 20 shown in FIGS. 7 (a), (b) or 11 (a), (b) is automated, and mass production of the workpiece W is performed. It is necessary to shorten the inspection time. In order to respond to the requirements, if the flowchart of FIG. 8 is implemented by software, the software may be installed in the defect inspection device 20A. When designing the software, the designer prepares a plurality of non-defective workpieces that are considered good, and defective ones that are considered bad. Then, after performing S101 to S103 in FIG. 8 on these workpieces, a plurality of monochrome portraits generated as a result thereof or images obtained as a result of calculation between the monochrome portraits are compared with each other, and an inspection is selected from among these. portrait.

Here, the shape of a workpiece as an inspection target and the types of defects to be detected during the inspection involve many aspects. Therefore, among the monochrome images and the images obtained as a result of the calculation between them, what kind of image is most suitable as the inspection image, which corresponds to the shape of the workpieces. And the types of defects vary. The calculation example such as the green image (G) -blue image (B) described above is a subtraction for obtaining the difference between two types of monochrome images. However, the calculation that can generate the best inspection portrait may be performed by combining different types of monochrome portraits such as red portrait (R)-green portrait (G) + blue portrait (B). That is, designers need to rely on intuition based on past experience to determine the content of calculations. In this way, a large number of calculations based on intuition are performed, and a large number of images generated by separate calculations are compared to select an inspection image. In addition, as described above, the image that is most suitable as the inspection image is changed into various types according to the shape of the workpiece and the type of the defect. Therefore, every time the shape of the workpiece to be inspected and the type of defect change, the calculation must be determined based on the intuition of past experience as described above, and the operation of selecting the most suitable inspection image must be re-executed. Therefore, the design of automated software for inspection requires a lot of labor and time.

On the other hand, if this embodiment is used as described above, it is possible to easily and surely perform defect inspection without performing complicated image processing.

[Comparative Example 2]

Next, as Comparative Example 2 with respect to this embodiment, in the defect inspection method of the prior art, extraction of defect candidate areas using the binary image of the inspection image (S107 to S110 in FIG. 8) will be described. An example of a check image selected by S104 to S106 of FIG. 8 is shown in FIG.

FIG. 12 shows a wafer-type electron from the workpiece W shown in FIG. 7 (b). The blue image obtained from the photographic image on the part (Wu in Figure 7 (b)). When the inspection image is binarized by an appropriate threshold value, a defect candidate region can be extracted. The white vertical regions visible on the left and right of the image are the electrode Wa and the electrode Wb, respectively. It is formed between the electrode Wa and the electrode Wb, and is held by the white edge E1 and the edge E2 (both are included by a one-dot chain line).

Here, in FIG. 12, the body Wd surrounded by a one-point chain line near the edge E1 and the edge E2 of the body Wd has a white portion in a thin horizontal band. Furthermore, in addition to the body Wd, there are three thin band-shaped regions from the lower left to the upper right. In FIG. 12, among these three, a one-point chain line surrounds a thin band-shaped region Bd1 existing at a slightly central portion in the up-down direction of the body Wd. For these areas E1, E2, and Bd1 as defects or noise, it is extremely difficult to identify them visually. The difficulty of identifying defects and noises in such an inspection image indicates that it is difficult to set a threshold value for binarizing the candidate defect areas. By adjusting the threshold to optimize the binary image, it is possible to obtain a binary image that is easy to inspect for defects. However, it takes a long time to reach the adjustment of the critical value. In some cases, it is not possible to extract a defect candidate area where defect inspection is easy even if the adjustment is performed. On the other hand, if this embodiment is used as described above, it is possible to easily and surely perform defect inspection without performing complicated image processing.

In the above description, although the inspection image is a blue image, it is sufficiently considered that even if the red image or the green image obtained from the same photographic image is set as the binarization threshold value, respectively, It is also difficult to extract the defect candidate area. In addition, even when an image generated by performing calculation between monochrome images is used as an inspection image, it may be difficult to set a critical value when the inspection image is binarized as described above. That is, it is possible to obtain a result that it is difficult to perform a high-accuracy defect inspection even if an inspection image is selected.

[Comparative Example 3]

Next, as a comparative example 3 with this embodiment, a case where a scatter diagram is prepared in Comparative Example 1 and a defect distribution region and a noise distribution region in the scatter diagram are divided by a straight line will be described using FIG. 13. In Comparative Example 1, as described above, a red image, a green image, and a blue image were obtained from the photographic image as the monochrome image (S103 in FIG. 8). From among the monochrome images obtained by performing the calculations between the monochrome images or the monochrome images thus obtained, the inspection images are selected (S104 to S106 in FIG. 8). Then, the inspection target area is extracted from the inspection image, and binarization is performed on the full pixels (S107 to S110 in FIG. 8). Here, FIG. 13 is a two-dimensional scatter diagram showing respective luminance distributions for defects and noises existing in defect candidate regions extracted by binarization, as in FIG. 3. The luminance B of the blue component is obtained on the vertical axis, and the luminance R of the red component is obtained on the horizontal axis. The area Dd with a solid line is a defect distribution area that is a set of luminance distributions of defects, and the area Dn with a dotted line is a noise distribution area that is a set of luminance distributions of noise. Since there is no common part in the defect distribution area Dd and the noise distribution area Dn, the two can be separated from each other in a visual range on the scatter diagram.

Here, the vertical axis of the scatter diagram of FIG. 13 corresponds to FIG. 9 (d) The luminance distribution of the blue image corresponds to the luminance distribution of the red image in FIG. 9 (b). That is, if the red image in FIG. 9 (b) is selected as the inspection image, a binary image obtained by binarizing the full pixels of the inspection target area extracted from the image is applied with a certain threshold RTT. When the defect distribution area Dd and the noise distribution area Dn in the scatter diagram in FIG. 14 are separated, the defect candidate area can be determined by such separation to perform the defect inspection with high accuracy.

However, in the scatter diagram of FIG. 13, when the critical value RTT is taken on the horizontal axis (red brightness) and a straight line R = RTT parallel to the vertical axis (blue brightness) is drawn, the noise distribution area Dn belongs to R> RTT, but the defect distribution area Dd belongs to both R> RTT and R <RTT. This system cannot accurately separate the defect distribution Dd and the noise distribution area Dn even if a threshold value is set for the red luminance R.

Similarly, in the scatter diagram of FIG. 13, when the critical value BTT is taken on the vertical axis (blue brightness) and a straight line B = BTT parallel to the horizontal axis (red brightness) is drawn, although the defect distribution area Dd belongs to The area of B> BTT, but the noise distribution area Dn belongs to both the area of B> BTT and the area of B <BTT. This system cannot correctly separate the defect distribution Dd and the noise distribution area Dn even if a threshold value is set for the blue brightness B.

That is, in Comparative Example 1, when a check image is selected from monochrome images and a two-dimensional scatter diagram is created, even if the red image of FIG. 9 (b) is used as a check image, or FIG. 9 (b) is used The blue image is used as an inspection image, and it is extremely difficult to determine the defect candidate area, and it is impossible to perform a high-accuracy defect inspection.

Hereinafter, although detailed description is omitted, the green image in FIG. 9 (c) is binarized even if the positional relationship between the defect distribution area Dd and the noise distribution area Dn in FIG. 13 is as described above. The case where the defect distribution area Dd and the noise distribution area Dn cannot be accurately distinguished. This corresponds to S104 in the flowchart in FIG. 8, which corresponds to the difficulty in selecting a portrait from which monochrome defects seem to be most discernible. As a result, as shown in S106 of FIG. 8, it is selected from among the images generated by performing calculations between monochrome images, and the image with the best discrimination can be selected. Moreover, in this case, since the same phenomenon as that described with reference to FIG. 13 occurs, it is still difficult to select an image that seems to be most capable of discriminating a defect from among the images generated as a result of calculation.

In this regard, if the present embodiment is used, the present invention divides pixels in a photographic image into defect distribution regions and noise distribution regions, and generates pixels generated by two colors for pixels constituting each region. The two-dimensional scatter diagram. Furthermore, on the scatter diagram, two regions are divided by a line, and all pixels belonging to the region to be the object of defect inspection are selected accordingly, and a limited photographic image is generated by the selected pixels. Because it is such a simple process, it is easy to eliminate the noise distribution area and implement defect inspection generated with high accuracy. In addition, the process is the same even if the shape of the workpiece to be inspected and the type of defect are changed. Therefore, it is not necessary to perform the operation of selecting the most suitable inspection image again as in Comparative Example 1. That is, the labor required for the design of automated software for inspection is reduced, and the time required for design is also shortened.

Claims (14)

A defect inspection method comprising: a lighting process for irradiating light to a surface of an object to be inspected; a photography process for photographing a surface of an object to be photographed; and a configuration process for obtaining a photographic image from the Extract the defect distribution area where the defect should be detected and the noise distribution area where no defect is required to be detected. At the same time, the color components of the pixels existing in each area are arranged in red, green, and blue. The two color components become vertical and horizontal. On the two-dimensional scatter diagram of the axis; the segmentation project is to divide the pixels constituting the defect distribution area and the pixels constituting the noise distribution area by dividing the straight line on the two-dimensional scatter diagram; the selection project is selected on the second scatter diagram The pixels in the two regions that are divided by the dividing line on the side of the defect distribution area; the limited project, which applies the pixels selected in the selection process to the place where the pixels exist in the photographic portrait A limited photographic portrait is generated; and an inspection implementation process is performed using a limited photographic portrait to perform defect inspection. The defect inspection method according to claim 1, wherein the light is white light. The defect inspection method according to claim 1, wherein the light is two or more different colors. The defect inspection method according to claim 1, wherein the light is a first color light, a second color light, and a third color light of different colors. The defect inspection method according to claim 4, wherein the first color light, the second color light, and the third color light are each of red, green, and blue. The defect inspection method according to claim 4 or 5, wherein the illumination process irradiates the inspection object with the first color light, the second color light, and the third color light at different heights. The defect inspection method according to claim 1, wherein when the vertical axis of the quadratic scatter diagram is set to y, the horizontal axis is set to x, and a and b are set to real numbers, the linear equation is y = ax + b means. A defect inspection system comprising: an illumination device that irradiates light to a surface of an inspection object; a photographing device that photographs a surface of an inspection object to obtain a photographic image; and a defect inspection device that The photographic image of the photographic device is subjected to image processing for defect inspection. The defect inspection device includes: The placement unit extracts the defect distribution areas where the defects should be detected and the noise distribution areas where no defects need to be detected from the photographic images, and arranges the color components of the pixels existing in each area in red, green, and blue. The two color components become the two-dimensional scatter diagram of the vertical and horizontal and horizontal axes; the segmentation section divides the pixels constituting the defect distribution area and the pixels constituting the noise distribution area by dividing the straight line on the two-dimensional scatter diagram; The selection unit selects pixels belonging to the defect distribution area side in the two regions that are divided by the straight line on the secondary scatter diagram; the limitation unit is used to apply the pixels selected in the selection process to photography There is a place in the image where the pixel exists to generate a limited photographic portrait; and an inspection enforcement unit that performs a defect inspection using the limited photographic portrait. The defect inspection system according to claim 8, wherein the light is white light. The defect inspection system according to claim 8, wherein the light is two or more different colors. The defect inspection system according to claim 8, wherein the light is the first color light, the second color light, and the third color light of different colors. The defect inspection system according to claim 11, wherein the first color light, the second color light, and the third color light are red, green, and blue respectively. Either. The defect inspection system according to claim 11 or 12, wherein the illumination process irradiates the inspection object with the first, second, and third color lights at different heights. The defect inspection system according to claim 8, wherein when the vertical axis of the second-order scatter diagram is set to y, the horizontal axis is set to x, and a and b are set to real numbers, the linear equation is y = ax + b means.