JP4849822B2 - Appearance inspection apparatus and appearance inspection method - Google Patents

Description

  The present invention relates to an appearance inspection apparatus and an appearance inspection method for classifying defects generated on an inspection sample from a captured image obtained by imaging the surface of the inspection sample, and in particular, formed on a semiconductor wafer in a semiconductor manufacturing process. The present invention relates to an appearance inspection apparatus and an appearance inspection method for classifying defects of a semiconductor circuit pattern and a liquid crystal display panel.

  It is widely practiced to image a formed pattern to generate image data, and to analyze the image data to inspect for the presence or absence of a pattern defect. In particular, in the field of semiconductor manufacturing, a photomask inspection apparatus that inspects a photomask, a semiconductor wafer, and an appearance inspection apparatus that inspects a pattern formed on a liquid crystal display panel are widely used. In the following description, an appearance inspection apparatus (inspection machine) that detects defects in a semiconductor circuit pattern formed on a semiconductor wafer in a semiconductor manufacturing process will be described as an example. However, the present invention is not limited to this. .

  A general visual inspection apparatus is a bright-field inspection apparatus that illuminates a target surface from the vertical direction and captures an image of reflected light, but a dark-field inspection apparatus that does not directly capture illumination light is also used. In the case of a dark field inspection apparatus, a sensor is arranged so that regular reflection is not detected by illuminating the target surface from an oblique direction or a vertical direction, and a dark field image of the target surface is obtained by sequentially scanning the irradiation position of the illumination light. . Therefore, the image sensor may not be used in the dark field device, but this is also an object of the invention. As described above, any apparatus and method can be applied as long as it is an appearance inspection apparatus and an appearance inspection method for inspecting the appearance of a sample from a captured image obtained by imaging the surface of the sample.

  FIG. 1 shows a block diagram of a conventional appearance inspection apparatus similar to the appearance inspection apparatus proposed by the applicant of the present patent application in Japanese Patent Application Laid-Open No. 2004-177397 (the following Patent Document 1). As shown in the figure, a sample stage (chuck stage) 2 is provided on the upper surface of a stage 1 that can freely move in two-dimensional or three-dimensional directions. On this sample stage, the semiconductor wafer 3 to be inspected is placed and fixed. An imaging device 4 configured using a one-dimensional or two-dimensional CCD camera or the like is provided above the stage, and the imaging device 4 generates an image signal of a pattern formed on the semiconductor wafer 3.

  As shown in FIG. 2, a plurality of dies 3A are repeatedly arranged in a matrix on the semiconductor wafer 3 in the X direction and the Y direction, respectively. Since the same pattern is formed on each die, it is common to compare images of corresponding portions of adjacent dies. If there is no defect in both dies, the gray level difference is less than the threshold, but if there is a defect in one, the gray level difference is greater than the threshold (single detection). Since this does not know which die is defective, it is further compared with adjacent dies on different sides, and if the difference in gray level of the same part becomes larger than the threshold, it can be seen that the die is defective (double Detection).

  The imaging device 4 includes a one-dimensional CCD camera, and moves the stage 1 so that the camera moves (scans) relative to the semiconductor wafer 3 at a constant speed in the X direction or the Y direction. The image signal is converted into a multi-value digital signal (gray level signal) and then input to the difference detection unit 6 and stored in the signal storage unit 5. When the gray level signal of the adjacent die is generated by scanning, the gray level signal of the previous die stored in the signal storage unit 5 is read out in synchronization with it and input to the difference detection unit 6. Actually, a minute alignment process is performed, but detailed description is omitted here.

  The difference detection unit 6 receives the gray level signals of two adjacent dies, calculates the difference between the two gray level signals (gray level difference), and outputs the difference to the detection threshold calculation unit 7 and the defect detection unit 8. . Here, the difference detector 6 calculates the absolute value of the gray level difference and outputs it as a gray level difference. The detection threshold calculation unit 7 determines a detection threshold from the gray level difference and outputs the detection threshold to the defect detection unit 8. The defect detection unit 8 compares the gray level difference with the determined threshold value and determines whether or not the defect is present.

  In general, a semiconductor pattern has a different noise level depending on a pattern type such as a memory cell portion, a logic circuit portion, a wiring portion, an analog circuit portion, and the like. Correspondence between semiconductor pattern portions and types can be understood from design data. Therefore, for example, the detection threshold value calculation unit 7 automatically determines a detection threshold value for each part according to the distribution of gray level differences of the part, and the defect detection unit 8 performs determination based on the threshold value determined for each part. .

  Further, some appearance inspection apparatuses include a defect classification unit 9 that automatically classifies each type of defect detected by the defect detection unit 8. The defect classification unit 9 identifies the type of the detected defect based on the appearance feature of the defect that appears in the captured image, and performs classification according to a predetermined defect classification. Such classification of defect information is used to identify the type of defect detected by the defect detection unit 8, investigate the cause of the defect, and identify the process that caused the defect.

JP 2004-177397 A

  However, when the captured image used for defect detection is also used for defect classification as described above, the resolution of the captured image required for defect classification is generally higher than the imaging resolution necessary for defect detection. If a captured image is acquired at a resolution required for defect detection 4, defect classification cannot be performed with sufficient accuracy. On the other hand, when a high-resolution image is used by SEM (scanning electron microscope) or the like in order to classify defects with sufficient accuracy, there is a problem that the inspection time for performing defect detection becomes longer than necessary.

  In view of the above problems, the present invention provides an appearance inspection apparatus capable of classifying a detected defect with higher accuracy than a conventional defect classification method using a captured image used for defect detection, and An object is to provide an appearance inspection method.

In order to achieve the above object, according to the present invention, in order to detect defects existing on the surface of the inspection sample from the captured image obtained by imaging the surface of the inspection sample and classify these defects, a predetermined standard is used in advance. The known defects detected in the sample are classified according to the defect classification of the first system and the defect classification of the second system by the predetermined defect classification means and the predetermined observation means, respectively, and the first and second systems Based on the result of each classification according to the defect classification, the classification according to the second defect classification by the predetermined observation means from the number of each type of defect classified according to the first defect classification by the predetermined defect classification means An approximate expression for calculating the number of defects of each class is derived.
Then, after the inspection sample is classified into the first system defect classification by a predetermined defect classification means, based on the number of defects of each class classified into the first system defect classification, the above According to the approximate expression, the number of defects of each class according to the defect classification of the second system is calculated.

  For example, the approximate expression is obtained by classifying the reference sample according to the second defect classification by the predetermined observation means in each class of defects classified according to the first defect classification by the predetermined defect classification means. Approximating the number of defects of each class according to the defect classification of the second system from the number of defects of each class classified according to the defect classification of the first system, according to the proportion of each class of defects accounted for It may be.

  Alternatively, the approximate expression may be, for example, the ratio of the number of defects classified according to the defect classification of the second system from the ratio of the number of defects classified according to the defect classification of the first system. It may be an approximation. Such an approximate expression includes the ratio of the number of defects of each class classified according to the first defect classification by the predetermined defect classification means for each of the plurality of reference samples, and the second observation means by the predetermined observation means. It is possible to derive based on the relationship between the ratio of the number of defects of each class classified according to the defect classification.

According to the present invention, other observation means (for example, SEM observation using an SEM which is a high-resolution imaging device) based on the result of defect classification by a conventional defect classification means using a captured image used for defect detection. It is possible to statistically approximate the result of defect classification. As a result, defect classification can be performed with higher accuracy than defect classification performed using a captured image used for conventional defect detection.
Further, according to the present invention, since it is not necessary to increase the resolution of a captured image used for defect detection in order to perform defect classification with high accuracy, it is possible to prevent adverse effects on the defect inspection throughput.

  Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. FIG. 3 is a block diagram showing a schematic configuration of the semiconductor pattern appearance inspection apparatus according to the first embodiment of the present invention. The semiconductor pattern appearance inspection apparatus shown in FIG. 3 has the same configuration as the semiconductor pattern appearance inspection apparatus described with reference to FIG. 1, and the same or similar components are denoted by the same reference numerals. In addition, detailed description of the same components is omitted.

As shown in the figure, a sample stage 2 is provided on the upper surface of a stage 1 that can move in a two-dimensional or three-dimensional direction. An imaging device 4 configured using a CCD camera or the like is provided on the upper part of the stage. The imaging device 4 generates an image signal of a captured image of a pattern formed on the semiconductor wafer 3.
The imaging device 4 includes a one-dimensional CCD camera such as TDI, and moves the stage 1 so that the camera moves (scans) relative to the semiconductor wafer 3 at a constant speed in the X direction or the Y direction. The image signal of the captured image is converted into a multi-value digital signal (gray level signal), and then input to the difference detection unit 6 and stored in the signal storage unit 5. When the gray level signal of the adjacent die is generated by scanning, the gray level signal of the previous die stored in the signal storage unit 5 is read out in synchronization with it and input to the difference detection unit 6.

The difference detection unit 6 receives the gray level signals of two adjacent dies, calculates the difference between the two gray level signals (gray level difference), and outputs the difference to the detection threshold calculation unit 7 and the defect detection unit 8. . The detection threshold calculation unit 7 determines a detection threshold from the gray level difference and outputs it to the defect detection unit 8. The defect detection unit 8 compares the gray level difference with the determined threshold value to determine whether the defect is a defect, detects the defect, and outputs predetermined defect information for each detected defect.
The defect information includes the defect identifier, position information, and information regarding the appearance characteristics of the defect appearing in the image captured by the imaging device 4 (for example, the brightness of the pixel determined to be a defect and the size of the defect). included.

The defect classification unit 9 identifies a type of the detected defect based on the appearance feature appearing in the captured image regarding the defect included in the defect information, and performs a predetermined defect classification (first system defect classification). According to the classification.
Here, the defect classification of the first system is composed of n classes of class 1 to class n (n is a natural number), and the defect classification unit 9 corresponds to, for example, appearance features appearing in the captured image of each defect. Then, the defects detected by the defect detection unit 8 are classified into classes 1 to n. For the following explanation, in the first type of defect classification, for example, class 1 has “high brightness and large” defects, class 2 has “high brightness and medium magnitude” defects, and class 3 It is assumed that a “high and small brightness” defect is classified into “4”, a “medium and large brightness” defect is classified into class 4, and so on.

  When defect detection by the defect detection unit 8 is completed for one or a plurality of wafers 3, the defect classification unit 9 detects all the defects that are detected on the wafer 3 and can be classified by the defect classification of the first system. The number of defects for each class classified into classes 1 to n according to the defect classification of the first system is output to the defect count calculation unit 10 described below.

Further, the semiconductor pattern appearance inspection apparatus is configured according to an approximate expression to be described later based on the number of defects for each class classified by the class 1 to n according to the first class of defect classification by the defect classification unit 9. 2 is provided with a defect number calculation unit 10 that calculates the number of defects for each class classified according to the defect classification of the second system.
Here, the defect classification of the second system is a defect classification consisting of m classes of class 1 to class m (m is a natural number). For example, the defect type related to each of classes 1 to m is arbitrary by the user. The defect classification defined in (1) may be used. For the following explanation, in the second type of defect classification, for example, class 1 has a “wiring short circuit” defect, class 2 has a “pattern defect” defect, and class 3 has an “interlayer” It is assumed that “defects caused by particles to be performed” are classified as follows.

  As shown in the following equation (1), the defect number calculation unit 10 determines the defects for each class 1 to n classified according to the first system defect classification according to the predetermined approximate expressions f1, f2,. Based on the number, the number of defects for each class 1 to m classified according to the defect classification of the second system is calculated.

  Here, N1, N2,... Nn indicate the number of defects for each class 1 to n classified according to the defect classification of the first system, and A1, A2,. The number of defects for each class 1 to m classified according to

Further, the appearance inspecting apparatus for a semiconductor pattern is an approximation that derives the above approximate expressions f1, f2,..., Fm in advance before calculating the number of defects of each class according to the defect classification of the second system by the defect number calculation unit 10. An expression deriving unit 11 and an approximate expression storage unit 12 that stores the derived approximate expression are provided.
The approximate expression deriving unit 11 uses the wafer 3 (reference sample) in which defects have already been detected and the defects are classified according to the defect classification of the second system by predetermined observation means such as SEM observation. Thus, the above approximate expression is derived.
Then, the approximate expression deriving unit 11 classifies the known defects detected in the wafer 3 (reference sample) according to the defect classification of the second system by the predetermined observation means (SEM observation in the above example). The approximate expression is derived from the result of classification according to the defect classification of the first system by the defect classification unit 9.

The approximate formulas f1, f2,..., Fm derived by the approximate formula deriving unit 11 are stored in the approximate formula storage unit 12, and the defect number calculation unit 10 is classified according to the defect classification of the second system. When calculating the number of defects for each class 1 to m, this approximate expression is read from the approximate expression storage unit 12 and used.
Hereinafter, referring to the flowchart shown in FIG. 4, a method for deriving the first example of the approximate expressions f1, f2,..., Fm by the approximate expression deriving unit 11, and the defect number calculating unit 10 based on the approximate expression. A method of calculating the number of defects of each class according to the second type of defect classification will be described in detail.

First, in step S1, each defect existing on the wafer 3 which is a given reference sample is detected by using the defect detection unit 8, and defect information of each defect is acquired.
Next, in step S2, each defect information detected in step S1 is classified by the defect classification unit 9 according to the first defect classification.
Similarly, in step S3, each defect information detected in step S1 is classified according to the second defect classification by a predetermined observation means such as SEM observation.
As shown in FIG. 3, the defect information classified by the defect classification unit 9 and the defect information classified by a predetermined observation unit are input to the approximate expression deriving unit 11.

  Returning to FIG. 4 again, in step S4, the approximate expression deriving unit 11 occupies the defects classified into classes 1 to n according to the first defect classification in step S2, and in accordance with the second defect classification in step S3. A ratio Pij (i = 1 to n, j = 1 to m) of defects classified into 1 to m is calculated. Here, Pij is the ratio of the defects classified into the i-th class according to the first defect classification to the defects classified into the j-th class according to the second defect classification. At this time, the approximate expression deriving unit 11 corresponds to which one of the defects classified according to the first defect classification in step S2 corresponds to each defect classified according to the second defect classification in step S3. Is determined based on the identifier and position information of each defect included in the defect information.

More specifically, the approximate expression deriving unit 11 occupies the defects classified into the i-th class according to the first defect classification in step S2, and each class 1 to 1 according to the second defect classification in step S3. The ratios Pi1 to Pim of the defects classified into m are calculated for the defects classified into classes 1 to m according to the first defect classification.
An approximate expression f1 to fm that approximates the number of defects A1 to Am of each class according to the defect classification of the second system from the number of defects N1 to Nn of each class classified according to the defect classification of the first system, Derived as shown in the following equation (2).

Then, the approximate expression deriving unit 11 stores the derived approximate expressions f1 to fm (or each ratio Pij (i = 1 to n, j = 1 to m)) in the approximate expression storage unit 12.
In step S5, the defect detection unit 8 is used to detect each defect present on the wafer 3 that is the inspection sample, and acquire defect information of each defect. In step S6, the defect classification unit 9 classifies the defect information detected in step S5 according to the first defect classification, and classifies each of the defect information classified into classes 1 to n according to the first defect classification. The defect numbers N1 to Nn are output to the defect number calculation unit 10.

  In step S <b> 7, the defect number calculation unit 10 reads out the approximate expressions f <b> 1 to fm (or the ratios Pij (i = 1 to n, j = 1 to m)) stored in the approximate expression storage unit 12, and calculates the above formula. According to (2), the respective defect numbers A1 to An classified into the classes 1 to m according to the second defect classification are calculated.

  FIG. 5 shows how the approximate expression deriving unit 11 derives the second example of the approximate expression, and how to calculate the number of types of defects according to the defect classification of the second system performed by the defect number calculating unit 10 based on the approximate expression. FIG. 6 is an explanatory diagram of the derivation method of the second example.

First, similarly to the derivation method of the first example described with reference to FIG. 4, in step S1, defect information of each defect existing on the wafer 3 which is a given reference sample is acquired, and step S2 is performed. In step S3, the defect information is classified by the defect classification unit 9 according to the first defect classification. In step S3, the defect information detected in step S1 is converted into the second information by predetermined observation means such as SEM observation. Classify according to defect classification.
The defect information classified by the defect classification unit 9 and the defect information classified by a predetermined observation unit are input to the approximate expression deriving unit 11.
These steps S1 to S3 are repeated for a plurality of reference samples.

In step S11, the approximate expression deriving unit 11 determines that the number of defects N1 to Nn of the defects classified into the classes 1 to n according to the first defect classification in step S2 is the total number of defects (ie, N = N1 + N2 + ... + Nn) is calculated separately for each reference sample.
The approximate expression deriving unit 11 also calculates the total number of defects (that is, A = A1 + A2 +... + Am) as the respective defect numbers A1 to Am classified into the classes 1 to m according to the second defect classification in step S3. Is calculated separately for each reference sample.

  Next, the approximate expression deriving unit 11 calculates the ratio Ni / N of the number of defects of each class i (i = 1... N) classified according to the first defect classification, calculated for each reference sample, and The ratio Aj / A is approximated from the ratio Ni / N on the basis of the relationship between the ratio Aj / A of the number of defects of each class j (j = 1... N) classified according to the two defect classifications. An approximate expression gij is calculated. This is shown in FIG.

FIG. 6A shows the ratio Ni / N of the number of defects of each class i (i = 1... N) classified according to the first defect classification, calculated for each reference sample, and the second. It is the figure which plotted the relationship with the ratio Aj / A of the number of defects of each class j (j = 1 ... n) classified according to the defect classification.
The approximate expression deriving unit 11 is based on a known method for deriving an approximate expression such as the least square method, and the approximate expression Aj / A = gij (Ni / N) as shown by the solid line in FIG. A correlation coefficient rij between the ratios Ni / N and Aj / A is calculated.

  Thereafter, the approximate expression deriving unit 11 calculates the number of defects of each class classified according to the second defect classification from the ratios N1 / N to Nn / N of the number of defects classified according to the first defect classification. An approximate expression that approximates A1 to Am is derived as shown in the following expression (3).

  Then, the approximate expression deriving unit 11 converts the derived approximate expression (or constant N, parameters defining each function gij, correlation coefficient rij (i = 1 to n, j = 1 to m)) into the approximate expression storage unit 12. To remember.

  Then, similarly to the derivation method of the first example described with reference to FIG. 4, in step S5, each defect present on the wafer 3 as the inspection sample is detected, and defect information of each defect is acquired. . In step S6, the defect classification unit 9 classifies the defect information according to the first defect classification, and sets the defect numbers N1 to Nn classified into the classes 1 to n according to the first defect classification as defects. It outputs to the number calculation part 10.

In step S12, the defect number calculation unit 10 obtains the ratio Ni / N of the number of defects of each class i (i = 1... N) classified according to the first defect classification for the wafer 3 as the inspection sample, and approximates it. The approximate expression (3) (or the constant N, the parameter defining each function gij, the correlation coefficient rij (i = 1 to n, j = 1 to m)) stored in the expression storage unit 12 is read.
And according to the above formula (3) based on the ratio Ni / N, the number of defects A1 to An classified into each class 1 to m according to the second defect classification is calculated.

  In the above-described embodiment, the defect classification of the second system is an example including each class forming the defect classification of the first system and a class indicating a different defect type. And the defect classification of the first system may be defect classifications composed of classes indicating the same defect type. That is, for example, the defect classifications of the first and second systems are classified as defect classifications consisting of m classes of class 1 to class m (m is a natural number), and the defect type related to each of classes 1 to m is determined by the user. Arbitrarily defined defect classifications, for example, class 1 includes “wiring short” defects, class 2 includes “pattern defects”, class 3 includes “interlayer particles” defects, It is good also as a defect classification | category classified like ....

  In addition, the defect classification of the second system and the defect classification of the first system, which are the same defect type in some classes, and which are different from each other in other classes, It is good.

  FIG. 7 shows a method of deriving the third example of the approximate expression by the approximate expression deriving unit 11 and the calculation of the number of types of defects according to the defect classification of the second system performed by the defect number calculating unit 10 based on the approximate expression. FIG. 8 is an explanatory diagram of the derivation method of the third example.

First, similarly to the derivation method of the first example described with reference to FIG. 4, in step S1, defect information of each defect existing on the wafer 3 which is a given reference sample is acquired, and step S2 is performed. In step S3, the defect information is classified by the defect classification unit 9 according to the first defect classification. In step S3, the defect information detected in step S1 is converted into the second information by predetermined observation means such as SEM observation. Classify according to defect classification.
The defect information classified by the defect classification unit 9 and the defect information classified by a predetermined observation unit are input to the approximate expression deriving unit 11.
These steps S1 to S3 and S21 are repeated for a plurality of reference samples.

Next, in step S21, the approximate expression deriving unit 11 classifies the approximate expression deriving unit 11 according to each defect classified according to the first defect classification in step S2 and according to the second defect classification in step S3. The respective defects are associated with each other based on the identifiers and position information of the defects included in the defect information of both.
Then, for each class i (i = 1... N) according to the first defect classification in step S2, out of the total number Ni of defects classified into each class i, the class i is determined by the first defect classification in step S2. Bij / Ni which is the ratio of the total number Bij occupied by the defects classified into each class j by the second defect classification in step S3 is calculated (j = 1... M).

  Next, the approximate expression deriving unit 11 calculates the ratio of the number of defects of each class i (i = 1... N) calculated according to the first defect classification calculated for each of the reference samples in step S22 Ni /. Based on the relationship between N and the ratio Bij / Ni calculated in step S21, an approximate expression hij that approximates the ratio Bij / Ni is derived for each class i and j from the ratio Ni / N. . This is shown in FIG.

FIG. 8A shows the ratio Ni / N of the number of defects of each class i classified according to the first defect classification calculated for each of the reference samples, and the ratio Bij / calculated in step S21. It is the figure which plotted the relationship with Ni. The approximate expression deriving unit 11 is based on a known method for deriving an approximate expression such as the least square method, and the approximate expression Bij / Ni = hij (Ni / N) as shown by the solid line in FIG. A correlation coefficient rij between the ratios Ni / N and Bij / Ni is calculated.
Thereafter, the approximate expression deriving unit 11 derives the approximate expression hij as shown in the following expression (4).

  Then, the approximate expression deriving unit 11 converts the derived approximate expression (or constant Sm, parameters for determining each function hij, correlation coefficient rij (i = 1 to n, j = 1 to m)) into the approximate expression storage unit 12. To remember.

  Then, similarly to the derivation method of the first example described with reference to FIG. 4, in step S5, each defect present on the wafer 3 as the inspection sample is detected, and defect information of each defect is acquired. . In step S6, the defect classification unit 9 classifies the defect information according to the first defect classification, and sets the defect numbers N1 to Nn classified into the classes 1 to n according to the first defect classification as defects. It outputs to the number calculation part 10.

In step S23, the defect number calculation unit 10 obtains the ratio Ni / N of the number of defects of each class i (i = 1... N) classified according to the first defect classification for the wafer 3 as the inspection sample. Also, the approximate expression (4) (or constant Sm, parameters for determining each function hij, correlation coefficient rij (i = 1 to n, j = 1 to m)) stored in the approximate expression storage unit 12 is read out.
And according to the above formula (4) based on the ratio Ni / N, the number of defects A1 to An classified into each class 1 to m according to the second defect classification is calculated.

  INDUSTRIAL APPLICABILITY The present invention can be used for an appearance inspection apparatus and an appearance inspection method for classifying defects generated on an inspection sample from a captured image obtained by imaging the surface of the inspection sample. In particular, it can be suitably used for a semiconductor circuit pattern formed on a semiconductor wafer in a semiconductor manufacturing process, and an appearance inspection apparatus and an appearance inspection method for classifying defects of a liquid crystal display panel.

It is a block diagram which shows schematic structure of the conventional external appearance inspection apparatus. It is a figure which shows the arrangement | sequence of the die | dye on a semiconductor wafer. It is a whole block diagram of the external appearance inspection apparatus for semiconductor patterns of the Example of this invention. It is a flowchart of the 1st example of the defect number calculation method by the visual inspection apparatus shown in FIG. It is a flowchart of the 2nd example of the defect number calculation method by the visual inspection apparatus shown in FIG. It is explanatory drawing of the defect number calculation method shown in FIG. It is a flowchart of the 3rd example of the defect number calculation method by the visual inspection apparatus shown in FIG. It is explanatory drawing of the defect number calculation method shown in FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Stage 2 Sample stand 3 Semiconductor wafer 3A Die 4 Imaging device 5 Signal storage part 8 Defect detection part 9 Defect classification part 10 Defect number calculation part

Claims (10)

An imaging means for imaging the inspection sample surface, a defect detection means for detecting a defect existing on the inspection sample surface from an image captured by the imaging means, and the defect detected by the defect detection means as a first system defect In the visual inspection apparatus comprising defect classification means for classifying according to classification,
The number of defects of each class according to the second class of defect classification according to a predetermined approximate expression based on the number of defects classified by the defect classification means according to the first class of defect classification by the defect classification means. A defect number calculating means for calculating
The predetermined approximate expression has a result of classification according to the defect classification of the first system by the defect classification means for a known defect detected in a given reference sample, and higher resolution than the imaging means. From the result of classification according to the defect classification of the second system by the predetermined observation means, and from the set of the number of defects of each type obtained in advance by the defect classification of the first system, An appearance inspection apparatus characterized by being an approximate expression for calculating a set of the number of defects of each class obtained by the defect classification of system 2 . The approximate expression classifies the reference sample according to the defect classification of the second system by the predetermined observation means in each class of defects classified by the defect classification means according to the defect classification of the first system. The number of defects of each class according to the defect classification of the second system is calculated from the number of defects of each class classified according to the defect classification of the first system, according to the proportion of each class of defects. The appearance inspection apparatus according to claim 1, wherein the appearance inspection apparatus is approximated. The approximate expression is
For each of the plurality of reference samples, the ratio of the number of defects of each class classified according to the defect classification of the first system by the predetermined defect classification means, and the second system by the predetermined observation means Derived based on the relationship between the percentage of the number of defects of each class classified according to the defect classification,
Approximating the ratio of the number of defects of each class classified according to the defect classification of the second system from the ratio of the number of defects of each class classified according to the defect classification of the first system,
The appearance inspection apparatus according to claim 1.   The class that constitutes the defect classification of the second system includes a class that indicates a defect type different from the class that constitutes the defect classification of the first system. Appearance inspection device.   The class that constitutes the defect classification of the second system includes a class that shows the same defect type as each class that constitutes the defect classification of the first system. Appearance inspection device. A defect existing on the surface of the inspection sample is detected from a captured image obtained by imaging the surface of the inspection sample, and the detected defect is further classified into a first system defect classification by a predetermined defect classification means. In the appearance inspection method to be performed,
A known defect detected in a given reference sample is detected by a predetermined observation means having a higher resolution than the predetermined defect classification means and the imaging means , respectively, and the first system defect classification and the second system, respectively. According to the defect classification of
Based on each of the classified results according defect classification of the first and second systems, a set of the number of defects each class which is classified according to defect classification of the first system by the given defect classification means from derives an approximation formula a group of the number of defects in each class are classified according to defect classification of the second system by the predetermined observation means, calculates,
Based on the number of types of defects classified into the first type of defect classification by the predetermined defect classification means for the inspection sample, according to the approximate expression, each type of defect according to the second type of defect classification Calculate the number,
An appearance inspection method characterized by that. The approximate expression is the defect classification of the second system by the predetermined observation means for each type of defect classified by the predetermined defect classification means according to the defect classification of the first system with respect to the reference sample. Each type of defect according to the second type of defect classification from the number of defects classified according to the first type of defect classification according to the proportion of each type of defect classified according to The appearance inspection method according to claim 6, wherein the number is approximated. The approximate expression is
For each of the plurality of reference samples, the ratio of the number of defects of each class classified according to the defect classification of the first system by the predetermined defect classification means, and the second system by the predetermined observation means Derived based on the relationship between the percentage of the number of defects of each class classified according to the defect classification,
Approximating the ratio of the number of defects of each class classified according to the defect classification of the second system from the ratio of the number of defects of each class classified according to the defect classification of the first system,
The appearance inspection method according to claim 6.   9. Each class that constitutes the defect classification of the second system includes a class that indicates a defect type different from each class that constitutes the defect classification of the first system. Appearance inspection method.   The class that forms the defect classification of the second system includes a class that shows the same defect type as each class that forms the defect classification of the first system. Appearance inspection method.

Images