JP2008256466A - Defect detection device, defect detection method, and defect detection program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To heighten accuracy and to heighten speed of defect detection based on three-dimensional measurement. <P>SOLUTION: A defect detection processing part 100 classifies observation point group data 1011 into non-defective point group data 1012 and defective point group data 1013 based on a distance to a corresponding point of reference point group data 1014 on each measuring point of the observation point group data 1011, performs alignment with the reference point group data 1014 by using the observation point group data 1011 from which the defective point group data 1013 are removed, compares the observation point group data 1011 with the reference point group data 1014, thereby discriminates a corresponding point of the reference point group data 1014 on each measuring point of the observation point group data 1011, and detects the defective point group data 1013 from the observation point group data 1011 based on the distance to the corresponding point of the reference point group data 1014 on each measuring point of the observation point group data 1011. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a defect detection apparatus, a defect detection method, and a defect detection program, and more particularly, to a method for automatically detecting defects such as product scratches and foreign substances based on shape data such as a three-dimensional point cloud and determining whether or not pass / fail. It is suitable for application.

As a method for managing product quality and detecting defects such as product scratches and foreign objects in the inspection process of the manufacturing process, for example, a two-dimensional image of the product is obtained using a camera or a microscope, and the two-dimensional image is obtained. There is a method of performing image processing by a computer.
There is also a method of detecting defects such as product scratches and foreign objects by performing three-dimensional measurement of the product using a laser displacement meter, a confocal microscope, or the like.

In addition, there is known a method for determining the quality of a product by classifying the detected defect into an allowable defect and an abnormal defect using information such as allowable dimensions (Patent Document 1).
Furthermore, when three-dimensional measurement point group data is given for a workpiece to be measured, a method is known in which the measurement point group data is compared with reference point group data to match the posture and position of the measurement point group data. (Patent Document 2 and Non-Patent Documents 1 to 3).

  For example, by marking the workpiece to be measured and comparing the position of the mark with the mark position of the reference point cloud data, the workpiece can be aligned or shared with the workpiece to be measured and the reference point cloud data A method for detecting a geometric shape is known (Patent Document 1). Further, a method is known in which the closest point between the reference data and the observation data is regarded as a corresponding point from the point cloud data, and coordinate transformation is performed from the position of the point cloud (Patent Document 2 and non-patent document). Literature 1, 3).

  Furthermore, when a correspondence relationship between reference data and observation data is given, a method of aligning a workpiece to be measured by calculating a three-dimensional coordinate transformation matrix between the reference data and observation data is known ( Patent Document 2 and Non-Patent Documents 1 to 3). In the ICP (Iterative Closest Point) method shown in Non-Patent Documents 1 and 3, points corresponding to the number of points corresponding to a specified ratio from points having a large distance between corresponding points are removed. A method of performing alignment is known.

JP 2005-201876 A JP 2001-82951 A PROCEEDINGS OF THE THIRD INTERNATIONAL CONFERENCE ON 3-D DIGITAL IMAGEING AND MODELING, IEEE, 2001, p. 145 (Efficient Variants of the ICP Algorithm) IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 9, NO. 5, IEEE, 1987, p. 698 (Least-Squares Fitting of Two 3-D Point Sets) IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 14, NO. 2, IEEE, 1992, p. 239 (A Method for Registration of 3-D Shapes)

  However, the conventional product quality determination method based on a two-dimensional image has a problem that an oversight or a false detection may occur depending on the type of defect. For example, in a method of detecting a portion of a grayscale image or color image that has a brightness or color that is significantly different from the surrounding portion as a defect, whether the brightness or color characteristics in the image are based on a three-dimensional flaw or a foreign object, or a simple discoloration of the surface Cannot be distinguished.

Further, in the conventional product quality determination method based on three-dimensional measurement, since the reference data and the observation data both have three-dimensional information, it is possible to grasp the mutual positional relationship between the reference data and the observation data. There was a problem that I could not.
Furthermore, in the methods disclosed in Non-Patent Documents 1 to 3, when the reference data and the observation data are aligned based on the three-dimensional point cloud data, the reference data and the observation data are different positions of the same object. There is a requirement that it be seen from

However, due to reasons such as individual differences occurring in products derived from the manufacturing process, noise resulting from measurement, defects such as scratches and foreign materials may be included in the observation data, etc. In some cases, the reference data and the observation data do not satisfy the condition that the same object is viewed from different positions.
For this reason, when the methods disclosed in Non-Patent Documents 1 to 3 are applied to the inspection process as they are, for example, the alignment between the observation data containing the defects and the reference data is performed, so that the alignment accuracy deteriorates. However, there is a problem that the detection accuracy of defects is lowered.

Further, the methods disclosed in Patent Documents 1 and 2 have a problem in that it is necessary to manually specify corresponding points of an object in advance by a pointing device or the like, and it is not possible to perform defect inspection at a large amount and at high speed.
Therefore, an object of the present invention is to provide a defect detection apparatus, a defect detection method, and a defect detection program capable of increasing the accuracy and speed of defect detection based on three-dimensional measurement.

  In order to solve the above-described problem, according to the defect detection apparatus according to claim 1, by comparing observation point cloud data obtained by observing the inspection object with reference point cloud data for the inspection object, Observation of the observation point group data based on the distance between corresponding point determination means for determining the corresponding point of the reference point group data at each observation point of the observation point group data and the corresponding point corresponding to the observation point Defect point determination means for determining a point as a defect point, defect point removal means for removing an observation point determined as a defect by the defect point determination means from the observation point group data, and the defect point removed Alignment means for performing alignment with each reference point of the reference point cloud data based on observation point cloud data, and a defect for performing defect detection on the inspection object based on the alignment result by the alignment means Characterized in that it comprises left and means.

According to the defect detection apparatus of claim 2, the corresponding point determination unit calculates a distance between each observation point of the observation point group data and each reference point of the reference point group data, and the observation point For each observation point of the group data, the nearest point having the shortest distance from the reference point is set as a corresponding point.
According to the defect detection apparatus of the third aspect, the defect point determination means sets the observation point whose distance to the nearest point exceeds a threshold as the defect point.
According to the defect detection apparatus of claim 4, the defect point determination unit generates a histogram of distances between the observation points of the observation point group data and the closest point, and the value of the histogram is A distance between the observation point corresponding to a certain ratio of a large higher rank and the closest point is set as the threshold value.

  According to the defect detection method of claim 5, each of the observation point group data is compared by comparing the observation point group data obtained by observing the inspection object with the reference point group data for the inspection object. Determining a corresponding point of the reference point group data at the observation point, and determining an observation point of the observation point group data as a defect point based on a distance between the corresponding point corresponding to the observation point; Removing the observation point determined to be defective by the defect point determination means from the observation point group data, and each reference point of the reference point group data based on the observation point group data from which the defect point has been removed And a step of detecting a defect with respect to the inspection object based on the alignment result.

According to the defect detection method of claim 6, a step of registering a plurality of non-defective product point group data as the reference point group data, and each reference of the observation point group data from the plurality of non-defective product point group data The method further comprises a step of selecting good product point cloud data having a good product point cloud having the highest coincidence with a point.
The defect detection apparatus according to claim 7 further includes a step of determining pass / fail of the inspection object based on a defect detection result for the inspection object.

  According to the defect detection program of claim 8, each observation point group data is obtained by comparing observation point group data obtained by observing the inspection object with reference point group data for the inspection object. Determining a corresponding point of the reference point group data at the observation point, and determining an observation point of the observation point group data as a defect point based on a distance between the corresponding point corresponding to the observation point; Removing the observation point determined to be defective by the defect point determination means from the observation point group data, and each reference point of the reference point group data based on the observation point group data from which the defect point has been removed And a step of causing the computer to execute a step of performing defect detection on the inspection object based on the alignment result.

  As described above, according to the present invention, by calculating the distance between each observation point and the reference point while taking correspondence between each observation point and the reference point, each observation point is a defect point. It is possible to determine whether there is a non-defective point. For this reason, even when a defect point such as a scratch or a foreign object is included in the observation point, the position of each point of the observation point cloud data and each point of the reference point cloud data is removed while removing such a defect point. It is possible to perform alignment, and it is possible to improve the alignment accuracy between each point of the observation point cloud data and each point of the reference point cloud data. As a result, it is not necessary to manually specify the corresponding points of the object with a pointing device or the like in advance, and the position of the defect can be specified with high accuracy, thereby improving the accuracy and speed of the defect detection based on the three-dimensional measurement. It becomes possible to plan.

Hereinafter, a defect detection apparatus according to an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing a schematic configuration of a defect detection apparatus according to the first embodiment of the present invention.
In FIG. 1, the defect detection apparatus 1020 detects defects from the inspection target while aligning the observation point group data 1011 obtained by observing the inspection target and the reference point group data 1014 for the inspection target. A pass / fail determination processing unit 1001 is provided for determining pass / fail of the inspection object based on the detection result of the defect by the detection processing unit 100 and the defect detection processing unit 100.

  Here, the defect detection processing unit 100 converts the observation point group data 1011 to the non-defective point group data 1012 based on the distance between the corresponding point of the reference point group data 1014 at each observation point of the observation point group data 1011. It is possible to detect defects from the inspection target while classifying them into the defect point group data 1013 and aligning the observation point group data 1011 from which the defect point group data 1013 has been removed and the reference point group data 1014.

  Then, the defect detection apparatus 1020 includes a CAD 1022 that generates three-dimensional observation point group data 1011 and reference point group data 1014, a three-dimensional data acquisition unit 1021 that acquires three-dimensional observation point group data 1011, and an inspection target. It is connected to a display device 1023 that displays a defect detected from the workpiece and a determination result of whether the workpiece to be inspected is a non-defective product or a defective product. The reference point group data 1014 may be generated from the non-defective observation point group data 1011 or may be generated from geometric information such as the CAD 1022 and drawings. Further, the defect detection apparatus 1020 may store the point cloud data that is correct information generated by the CAD 1022 such as non-defective products and models as reference point cloud data 1014.

  As the three-dimensional data acquisition unit 1021, a three-dimensional laser scanner, a laser displacement meter, a confocal microscope, or an omnifocal microscope may be used. In addition, an atomic force microscope or other three-dimensional measuring device may be used. In order to reduce noise, noise reduction means such as using an averaging filter related to the observation point group for the observation point group data 1011 may be provided.

  The reference point group data 1014 generated by the CAD 1022 and the observation point group data 1011 acquired by the three-dimensional data acquisition unit 1021 are input to the defect detection apparatus 1020. When the reference point group data 1014 and the observation point group data 1011 are input to the defect detection device 1020, the defect detection processing unit 100 uses the observation point group data 1011 including the defect point group data 1013 as it is. After alignment with the reference point group data 1014, the corresponding point of the reference point group data 1014 at each observation point of the observation point group data 1011 is determined by comparing the observation point group data 1011 with the reference point group data 1014. Determine.

When the corresponding point of the reference point group data 1014 at each observation point of the observation point group data 1011 is determined, the distance between each observation point of the observation point group data 1011 and each reference point of the reference point group data 1014 is calculated. The closest point having the shortest distance from the reference point for each observation point in the observation point group data 1011 can be set as the corresponding point.
Then, the defect detection processing unit 100 converts the observation point group data 1011 to the non-defective point group data 1012 and the defect based on the distance between the corresponding point of the reference point group data 1014 at each observation point of the observation point group data 1011. The defect point group data 1013 is removed from the observation point group data 1011.

  In addition, when classifying the observation point group data 1011 into the non-defective point group data 1012 and the defect point group data 1013, a histogram of the distance between the nearest point in each observation point of the observation point group data 1011 is generated. The distance between the observation points corresponding to a certain upper ratio with a large histogram value is set as the threshold value θ, and the distance between the closest points in the observation point group data 1011 where the distance between the observation points exceeds the threshold value θ. The observation point can be a defect point in the defect point group data 1013.

Then, the defect detection processing unit 100 performs alignment with the reference point group data 1014 by using the observation point group data 1011 from which the defect point group data 1013 has been removed, and then refers to the observation point group data 1011. By comparing with the point cloud data 1014, the corresponding point of the reference point cloud data 1014 at each observation point of the observation cloud data 1011 is determined.
Then, the defect detection processing unit 100 converts the observation point group data 1011 to the non-defective point group data 1012 and the defect based on the distance between the corresponding point of the reference point group data 1014 at each observation point of the observation point group data 1011. Further classification into point cloud data 1013 and defect point cloud data 1013 can be detected from the observation cloud data 1011.

Then, the defect detection processing unit 100 sets the reference point group data 1014 for each observation point of the observation point group data 1011 from which the defect point group data 1013 detected in the N-th (N is a positive integer) alignment is removed. It is possible to perform (N + 1) -th alignment with each reference point, detect defect point group data 1013 based on the (N + 1) -th alignment result, and send it to the pass / fail judgment processing unit 1001. Then, when the defect determination processing unit 1001 receives the defect point group data 1013 from the defect detection processing unit 100, the defect determination group 1001 determines the quality of the inspection target based on the defect point group data 1013, and the defect point group data 1013 or the inspection target Is output to the display device 1023. The display device 1023 can display the defect point group data 1013 sent from the quality determination processing unit 1001 and the quality determination result of the inspection target.
In addition, as the pass / fail judgment criteria in the pass / fail judgment processing unit 1001, the number of detected defect point group data 1013, the ratio of the defect point group in the observation point group, or the volume of the solid formed by the defect point group may be used. May be used.

  Thus, by calculating the distance between each observation point and the reference point while taking correspondence between each observation point of the observation point group data 1011 and the reference point of the reference point group data 1014, each observation point is It is possible to determine whether the point is a defect point or a non-defect point. For this reason, even when a defect point such as a flaw or a foreign object is included in the observation point group data 1011, each observation point of the observation point group data 1011 is removed while removing such a defect point from the observation point group data 1011. Can be aligned with each reference point in the reference point group data 1014, and the alignment accuracy between each observation point in the observation point group data 1011 and each reference point in the reference point group data 1014 can be improved. It becomes possible. As a result, it is not necessary to manually specify the corresponding points of the object with a pointing device or the like in advance, and the position of the defect can be specified with high accuracy, thereby improving the accuracy and speed of the defect detection based on the three-dimensional measurement. It becomes possible to plan.

FIG. 2 is a diagram for explaining a distance calculation method for classifying observation point group data into non-defective point group data and defective point group data according to an embodiment of the present invention.
In FIG. 2A, it is assumed that reference points P1 to P5 are given as reference point group data 1014, and observation points Q1 to Q3 are given as observation point group data 1011.
Then, as shown in FIG. 2 (b), for the observation point Q1, the distance from each reference point P1 to P5 is calculated, and when the distance d1 to the reference point P2 is the shortest, the reference point for the observation point Q1. P2 can be a corresponding point.

Further, as shown in FIG. 2 (c), for the observation point Q2, the distances from the reference points P1 to P5 are calculated, and the distance d2 from the reference point P3 is the shortest. P3 can be a corresponding point.
Further, as shown in FIG. 2 (d), for the observation point Q3, the distances from the respective reference points P1 to P5 are calculated, and when the distance d3 from the reference point P5 is the shortest, the reference point for the observation point Q3. P5 can be a corresponding point.

Then, a histogram of the distances d1 to d3 between the corresponding observation points Q1 to Q3 is generated, and the corresponding points at the observation points Q1 to Q3 corresponding to a certain upper ratio with a large value of the histogram are generated. The distance between them is set as the threshold value θ, and the observation points Q1 to Q3 whose distances d1 to d3 between the corresponding points exceed the threshold value θ can be defined as defect points.
For example, if the distance d1 between the observation point Q1 and the corresponding point is set as the threshold θ, the distance d3 between the corresponding point for each observation point Q3 exceeds the threshold θ, and therefore the observation point Q3 Can be determined as a defect point, and the observation point Q3 can be removed from the observation point group data 1011. Then, by using the observation point group data 1011 including the observation points Q1 and Q2 from which the observation point Q3 is removed, alignment with each point of the reference point group data 1014 including the reference points P1 to P5 can be performed. .

FIG. 3 is a diagram for explaining a method of classifying observation point group data into non-defect point group data and defect point group data according to an embodiment of the present invention.
Assuming that an observation point group 1211 including a defect point group 1221 and a reference point group 1201 are given in FIG. 3A, each point and reference point in the observation point group 1211 are given as shown in FIG. By calculating the distance to each point in the group 1201, a vector 1232 between each point in the observation point group 1211 and each corresponding point in the reference point group 1201 is calculated, and each point in the defect point group 1221 A vector 1232 between each corresponding point in the reference point group 1201 is calculated. The distance between each point in the defect point group 1221 and each corresponding point in the reference point group 1201 is longer than the distance between each point in the observation point group 1211 and each corresponding point in the reference point group 1201. Then, as shown in FIG. 3C, the defect point group 1221 is removed from the observation point group 1211. As shown in FIG. 3D, the observation point group 1211 from which the defect point group 1221 is removed and the reference point. Registration with the group 1201 can be performed.

FIG. 4 is a flowchart showing a defect detection process in the defect detection apparatus of FIG.
In FIG. 4, in defect detection processing S100, threshold setting processing S101 for setting a threshold θ for classifying observation point cloud data 1011 into non-defective point cloud data 1012 and defect point cloud data 1013, and observation point cloud data 1011 Point group classification processing S102 for classifying into non-defect point group data 1012 and defect point group data 1013, and geometric transformation processing for converting the coordinates of the observation point group from which the defect point group has been removed to the coordinates of the corresponding point group of the reference point group S103 is sequentially performed, and the threshold value setting process S101, the point group classification process S102, and the geometric transformation process S103 are sequentially repeated until the end determination condition in the end determination process S111 is satisfied. Then, when the end determination condition in the end determination process S111 is satisfied, a defect display process S104 for displaying a defect point group is performed.

Note that the end determination condition in the end determination process S111 may be a case where the amount of conversion in the geometric conversion process S103 is sufficiently small, or a case where it is executed a predetermined number of times.
In order to improve the alignment accuracy while reducing the amount of calculation, it is preferable to set the imaging state of the observation point group and the imaging state of the reference point group to be substantially the same. In order to make the shooting situation of the observation point group and the shooting situation of the reference point group substantially the same, the work position and the shooting position may be finely adjusted by visual observation, and the shooting position can be controlled. A position fixing device such as an XY stage may be used.

Hereinafter, the defect detection process of FIG. 4 will be described in more detail.
FIG. 5 is a flowchart showing the threshold setting process of FIG.
In FIG. 5, when the process is executed for the first time in the threshold setting process S101 (step S211), the threshold θ is set to infinity (step S204). When the process is executed for the second time and thereafter, if {Xo (j)} is an observation point group, a histogram of d (NP (j), j) is obtained for the observation point Xo (j) of the observation point group data 1011. Generate (step S201). However, NP (j) is the nearest point of the reference point group data 1014 corresponding to each observation point j, and d (NP (j), j) is the nearest point NP (j at each observation point Xo (j). ).
Then, by using the ratio indicated by the set value α set in step S221, the observation point Xo (k) corresponding to the upper fixed ratio with a large value of the histogram of d (NP (j), j) is obtained. The distance d (NP (k), k) from the nearest point NP (k) at the observation point Xo (k) is set as the threshold θ (step S203).

FIG. 6 is a flowchart showing the point group classification process of FIG.
In FIG. 6, if {Xr (i)} is a reference point group and {Xo (j)} is an observation point group (step S301), combinations of all reference points Xr (i) and observation points Xo (j) (Steps S302, S303, S305, S309, S311, and S313), by calculating the distance between the reference point Xr (i) and the observation point Xo (j) (Step S304), each observation point Xo (j ) To find the nearest point NP (j) (step S306).
Then, it is determined whether or not the distance d (NP (j), j) between each observation point Xo (j) and the nearest point NP (j) exceeds the threshold θ (step S312), and the nearest point NP (j ), The observation point Xo (j) whose distance d (NP (j), j) exceeds the threshold θ is classified as a defect point group (step S307), and then the observation point group {Xo (j)} Delete (step S308).

FIG. 7 is a flowchart showing the geometric transformation process of FIG.
In FIG. 7, if {Xr (i)} is a reference point group and {Xo (j)} is an observation point group (step S401), all observation points Xo (j) are referred to as observation points Xo (j). It is assumed that the nearest point Xr (NP (j)) of the point Xr (i) has a relationship to be transformed by affine transformation represented by the following equation (1) (step S402).
Xr (NP (j)) = A · Xo (j) + t + ε (j) (1)
Here, A is a matrix representing a three-dimensional rotation, t is a vector representing a three-dimensional translation, and ε (j) is a vector representing an error.

Then, the matrix A and the vector t are estimated by the parameter estimation process (step S403), and the estimated matrix A and the vector t are used for all the observation points Xo (j) (steps S404, S406, and S411). The observation point Xo (j) is affine transformed (step S405).
In the parameter estimation process S403, for example, the matrix A and the vector t can be estimated by a singular value decomposition method.

FIG. 8 is a flowchart showing the parameter estimation process of FIG.
In FIG. 8, {Xr (NP (j))} is a reference point group, and {Xo (j)} is an observation point group (step S701). Then, as shown in the following equations (2) and (3), the center of gravity Pr of the reference point group and the center of gravity Po of the observation point group are calculated (step S702).
Pr: = (Xr (NP (1)) + Xr (NP (2)) +... + Xr (NP (M)) / M
... (2)
Po: = (Xo (1)) + (Xo (2)) +... + Xo (M)) / M (3)
Where M is the number of observation points Xo (j)

Next, as shown in the following equations (4) and (5), a relative position Qr (j) from the center of gravity Pr of the reference point group and a relative distance Qo (j) from the center of gravity Po of the observation point group are calculated. (Step S703).
Qr (j): = Xr (NP (j))-Pr (4)
Qo (j): = Xo (j) −Po (5)
However, j is an integer value which takes the value of 1-M.

Next, as shown in the following equation (6), a product-sum matrix H of Qr (j) and Qo (j) is calculated (step S704).
H: = Qo (1) Qr (1) ^T + Qo (2) Qr (2) ^T +.
+ Qo (M) Qr (M) ^T (6)
Next, as shown in the following equation (7), singular value decomposition of the product-sum matrix H is performed (step S705).
H = UΛV ^T (7)
However, U and V are orthogonal matrices, and Λ is a diagonal matrix.

Next, as shown in the following equation (8), an estimated value of the matrix A is calculated from the singular value decomposition result of the product-sum matrix H (step S706).
^{A: = VU T ··· (8} )
Next, as shown in the following equation (9), the estimated value of the vector t is calculated by using the estimated value of the matrix A, the centroid Pr of the reference point group, and the centroid Po of the observation point group (step S707). .
t: = Pr-A · Po (9)

FIG. 9 is a flowchart showing the defect display processing of FIG.
In FIG. 9, {Xr (i)} is a reference point group, {Xo (j)} is an observation point group, and {Xe (j)} is a defect point group (step S501).
Then, the reference point group {Xr (i)} is displayed (step S502), and the set color set in step S521 is used so that the defect point group {Xe (j)} can be easily recognized. Thus, the display color is set (step S503).

Next, the coordinates of the defect point group {Xe (j)} are converted into the coordinates of the reference point group {Xr (i)} (step S504), and the defect point group {Xe (j)} is displayed (step S505). .
In step S504, in order to convert the coordinates of the defect point group {Xe (j)} into the coordinates of the reference point group {Xr (i)}, each point of the defect point group {Xe (j)} is regarded as a defect. The affine transformation process S405 in the geometric transformation process S103 performed from the time of classification to the end of the defect detection process S100 in FIG. 4 may be executed for all defect points. Alternatively, the geometric transformation process S103 for the observation point group {Xo (j)} in FIG. 7 may be performed simultaneously on the defect point group {Xe (j)}.

  Thereby, while aligning each point of the observation point group data 1011 and each point of the reference point group data 1014 based on the three-dimensional measurement data, the influence of defects such as scratches and foreign matters is stepwise by repeated calculation. The position of each point of the observation point group data 1011 and each point of the reference point group data 1014 is reduced by matching each point of the observation point group data 1011 having no defect with each point of the reference point group data 1014. It is possible to detect a defect from an inspection object while improving the alignment accuracy.

Further, by sequentially removing the defect portion from the observation point group data 1011, it becomes possible to accurately align each point of the observation point group data 1011 with the defect and each point of the reference point group data 1014. Precise defect detection using three-dimensional measurement data can be realized.
Further, it is possible to simultaneously perform alignment and defect detection between each point of the observation point group data 1011 and each point of the reference point group data 1014, and to detect defects from the observation point group data 1011 at high speed. .

In addition, the size of the area determined as a defect can be estimated from the number of defect points that are close to each other, and the size of the defect can be used as a criterion for distinguishing between a non-defective product and a defective product.
In addition, it is possible to determine a defect using point cloud data, and it is possible to apply it to an industrial product having a complicated shape that is difficult to express with a simple geometric shape.
The threshold setting process S101, the point group classification process S102, and the geometric transformation process S103 can be realized by causing a computer to execute a program in which an instruction for performing these processes is described.

If this program is stored in a storage medium such as a CD-ROM, the threshold setting process S101, the point group classification process S102, and the geometric transformation are performed by installing the storage medium in the computer and installing the program in the computer. Processing S103 can be realized.
Further, when the computer executes a program in which instructions for performing the threshold setting process S101, the point group classification process S102, and the geometric transformation process S103 are described, the program may be executed by a stand-alone computer or connected to a network. A plurality of computers may be distributed.

FIG. 10 is a front view showing a schematic configuration of a defect measurement apparatus to which the defect detection apparatus of FIG. 1 is applied.
10, the defect measuring apparatus is provided with a microscope 1121 that acquires three-dimensional observation point group data 1011 from an inspection target and a computer 1122 that executes the defect detection process of FIG. Here, the defect detection device 1020 shown in FIG.

  When the observation point group data 1011 measured by the microscope 1121 is input to the computer 1122, the computer 1122 executes the defect detection process of FIG. 4 to refer to each point of the observation point group data 1011. While aligning with each point of the point cloud data 1014, the observation point cloud data 1011 can be classified into the non-defective point cloud data 1012 and the defective dot cloud data 1013, and a defect can be detected from the inspection target. The computer 1122 can display the defect detected from the inspection target on the display device.

FIG. 11 is a block diagram showing a schematic configuration of a defect detection apparatus according to the second embodiment of the present invention.
11, a plurality of non-defective product point group data 1311 to 1313 are registered as reference point group data 1014 in the defect detection apparatus 1020 of FIG. A non-defective product selection processing unit 1301 for selecting non-defective product point group data 1311 to 1313 having a non-defective product point group having the highest coincidence with each point of the data 1011 is provided.

Then, when the observation point group data 1011 is input to the defect detection apparatus 1020, the non-defective product selection processing unit 1301 has the highest coincidence with the observation point group data 1011 among the plurality of non-defective product point group data 1311 to 1313, for example. The non-defective product point group data 1311 is selected as the data and is output to the defect detection processing unit 100. Then, the defect detection processing unit 100 uses the observation point cloud data 1011 and the good product point cloud data 1311 to align each point of the observation point cloud data 1011 with each point of the good product point cloud data 1311. The observation point group data 1011 is classified into the non-defect point group data 1012 and the defect point group data 1013, and the defect can be detected from the inspection target.
Thereby, the defect detection process can be executed using the good product point cloud data 1311 having the good product point cloud having the highest coincidence with each point of the observation point cloud data 1011, and the influence of the manufacturing unevenness of each product is reduced. Or a plurality of types of products can be inspected on the same line.

FIG. 12 is a perspective view showing an example of a schematic configuration of a non-defective product sorting apparatus according to the third embodiment of the present invention.
In FIG. 12, the non-defective product sorting apparatus is provided with a three-dimensional data acquisition unit 1021 for acquiring three-dimensional observation point cloud data 1011 from an inspection target 1431 and a computer 1122 for executing the defect detection process of FIG. 1431, transport systems 1422 and 1423 that transport 1431, a transport system controller 1421 that controls transport systems 1422 and 1423 based on the defect detection results by the computer 1122, a non-defective product stocker 1432 that stores an inspection target 1431 that is determined to be non-defective A defective product stocker 1433 for storing the inspection object 1431 determined to be provided is provided.

  The inspection target 1431 is transported by the transport system 1422, and the observation point group data 1011 is measured from the inspection target 1431 by the three-dimensional data acquisition unit 1021. When the observation point group data 1011 measured by the three-dimensional data acquisition unit 1021 is input to the computer 1122, the computer 1122 executes each defect of the observation point group data 1011 by executing the defect detection process of FIG. The observation point group data 1011 is classified into the non-defect point group data 1012 and the defect point group data 1013 while aligning the points with the respective points of the reference point group data 1014, and defects are detected from the inspection object 1431. Can do. Then, the computer 1122 can determine pass / fail of the inspection target 1431 based on the defect detection result from the inspection target 1431 and send the pass / fail determination result to the transport system controller 1421.

Then, the inspection target 1431 measured by the three-dimensional data acquisition unit 1021 is moved from the transport system 1422 to the transport system 1423, and the transport system control device 1421 determines that the inspection target 1431 is a non-defective product. When the inspection object 1431 is conveyed to the non-defective product stocker 1432 and the inspection object 1431 is determined to be defective, the conveyance system 1423 can convey the inspection object 1431 to the defective product stocker 1433.
Accordingly, by causing the computer 1122 to execute the defect detection process of FIG. 4, the inspection target 1431 is determined as a non-defective product and a defective product while performing the pass / fail determination of the inspection target 1431 based on the detection result of the defect from the inspection target 1431. This makes it possible to sort the inspection object 1431 in-line and improve the sorting accuracy.

1 is a block diagram showing a schematic configuration of a defect detection apparatus according to a first embodiment of the present invention. It is a figure explaining the calculation method of the distance for classifying the observation point group data which concern on one Embodiment of this invention into non-defect point group data and defect point group data. It is a figure explaining the method of classifying the observation point cloud data which concern on one Embodiment of this invention into non-defect point cloud data and defect point cloud data. It is a flowchart which shows the defect detection process in the defect detection apparatus of FIG. It is a flowchart which shows the threshold value setting process of FIG. It is a flowchart which shows the point group classification process of FIG. It is a flowchart which shows the geometric transformation process of FIG. It is a flowchart which shows the parameter estimation process of FIG. It is a flowchart which shows the defect display process of FIG. It is a front view which shows schematic structure of the defect measuring apparatus with which the defect detection apparatus of FIG. 1 is applied. It is a block diagram which shows schematic structure of the defect detection apparatus which concerns on 2nd Embodiment of this invention. It is a perspective view which shows schematic structure of the non-defective product selection apparatus which concerns on 3rd Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Defect detection process part 1001 Pass / fail judgment process part 1011 Observation point group data 1012 Non-defect point group data 1013 Defect point group data 1014 Reference point group data 1020 Defect detection apparatus 1021 Three-dimensional data acquisition means 1022 CAD
1023 Display device 1121 Microscope 1122 Computer P1 to P5 Reference point Q1 to Q3 Observation point 1201 Reference point group 1211 Observation point group 1221 Defective point group 1231, 1232 Vector 1014 Observation point group data 1311 to 1313 Good product point group data 1301 Good product selection processing Portion 1421 Conveyance system controller 1422, 1423 Conveyance system 1431 Inspection target 1432 Non-defective stocker 1433 Defective product stocker

Claims (8)

Correspondence to discriminate corresponding points of reference point cloud data at each observation point of the observation point group data by comparing observation point group data obtained by observing the inspection subject with reference point cloud data for the inspection subject Point discriminating means;
Defect point determination means for determining an observation point of the observation point group data as a defect point based on a distance between corresponding points corresponding to the observation point;
Defect point removal means for removing observation points determined as defects by the defect point determination means from the observation point group data;
Alignment means for performing alignment with each reference point of the reference point group data based on the observation point group data from which the defect point has been removed;
A defect detection apparatus comprising: defect detection means for performing defect detection on the inspection object based on an alignment result obtained by the alignment means.   The corresponding point determination means calculates a distance between each observation point of the observation point group data and each reference point of the reference point group data, and for each observation point of the observation point group data, between the reference points 2. The defect detection apparatus according to claim 1, wherein the closest point having the shortest distance is set as a corresponding point.   3. The defect detection apparatus according to claim 2, wherein the defect point determination means sets an observation point whose distance to the nearest point exceeds a threshold as the defect point.   The defect point determination means generates a histogram of the distance to the nearest point at each observation point of the observation point group data, and the histogram point at the observation point corresponding to a certain high-order ratio with a large value of the histogram The defect detection apparatus according to claim 3, wherein a distance from the closest point is set as the threshold value. Determining corresponding points of the reference point group data at each observation point of the observation point group data by comparing the observation point group data obtained by observing the inspection target with the reference point group data for the inspection target When,
Determining an observation point of the observation point group data as a defect point based on a distance between corresponding points corresponding to the observation point;
Removing the observation point determined to be defective by the defect point determination means from the observation point group data;
Performing alignment with each reference point of the reference point group data based on the observation point group data from which the defect point has been removed;
And a step of detecting a defect with respect to the inspection object based on the alignment result. Registering a plurality of non-defective point cloud data as the reference point cloud data;
6. The step of selecting non-defective product point cloud data having a good product point cloud having the highest coincidence with each reference point of the observation point cloud data from the plurality of good product point cloud data. Defect detection method.   The defect detection method according to claim 5, further comprising a step of determining pass / fail of the inspection object based on a defect detection result for the inspection object. Determining corresponding points of the reference point group data at each observation point of the observation point group data by comparing the observation point group data obtained by observing the inspection target with the reference point group data for the inspection target When,
Determining an observation point of the observation point group data as a defect point based on a distance between corresponding points corresponding to the observation point;
Removing observation points determined as defects by the defect point determination means from the observation point group data;
Performing alignment with each reference point of the reference point group data based on the observation point group data from which the defect point has been removed;
A defect detection program for causing a computer to execute a step of performing defect detection on the inspection object based on the alignment result.

Images