JP2013088414A - Shape inspection method and device of the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a three-dimensional shape inspection method for securing high measurement accuracy regardless of a shape of a measuring object by complementarily combining a plurality of three-dimensional shape measurement methods and surface measurement methods, and its device.SOLUTION: A three-dimensional shape inspection device includes: a storage part for storing reference data which is shape data of a reference model of an inspection object; an area specifying part for specifying an area to acquire first shape data from the reference data; a first three-dimensional shape sensor for acquiring the first shape data for the area specified in the area specifying part; a second three-dimensional shape sensor for acquiring second shape data different from the first shape data of the inspection object for an area other than the area specified in the area specifying part; and a complementary integration part for integrating the first shape data and the second shape data.

Description

  The present invention relates to a shape inspection method and an inspection apparatus for a processed product and a processing tool, for example.

  For the purpose of ensuring the quality of products during processing and assembly in manufacturing, there is a need for processing tools that can be quantitatively fed back to the manufacturing process regardless of the skill of visual inspectors, and for product shape and surface condition inspection technology. .

  In Patent Document 1, in a three-dimensional shape measurement capable of performing a wide range of three-dimensional shape measurement by laser light scanning, the laser light amount is set so that the reflected light amount is constant even when a color or a shadow is added to the three-dimensional shape. There has been proposed a method for performing three-dimensional shape measurement with good accuracy by adjustment.

  In Patent Document 2, the measurement point data representing the shape of the measurement object measured by the optical cutting method and the reference point data are aligned based on sequential convergence processing, and the measured measurement point data and the reference are aligned. An object shape evaluation apparatus that evaluates the shape of a measurement object based on point data is described. In the alignment process, the distance weight between adjacent points is determined based on the distance between adjacent points between adjacent measurement points or the distance between adjacent points between adjacent reference points. This is used when obtaining the sequential convergence evaluation value in the sequential convergence processing.

JP2009-204425A JP 2010-107300 A

  However, even if three-dimensional shape measurement is performed by a single three-dimensional shape measurement method as in Patent Documents 1 and 2, for example, when the shape of the measurement target includes an acute angle or a steep surface, the measurement accuracy is high. It is difficult to secure.

  Therefore, in view of the above problems, the present invention provides a three-dimensional shape inspection method that ensures high measurement accuracy regardless of the shape of the measurement target by combining a plurality of three-dimensional shape measurement methods and surface measurement methods in a complementary manner. An object is to provide such a device.

  In order to solve the above problems, the present invention includes a storage unit that stores reference data that is shape data of a reference model to be inspected, a region specifying unit that specifies a region from which the first shape data is acquired from the reference data, The first three-dimensional shape sensor that acquires the first shape data for the region specified by the region specifying unit, and the first shape data of the inspection target for a region other than the region specified by the region specifying unit 3, comprising: a second three-dimensional shape sensor that obtains second shape data different from the first shape data; and a complementary integration unit that integrates the first shape data and the second shape data. A dimension shape inspection apparatus is provided.

  According to another aspect of the present invention, reference data that is shape data of a reference model to be inspected is read, the first shape data is acquired for the specified region, and the inspection target is obtained for a region other than the specified region. A second shape data different from the first shape data is acquired, and the first shape data and the second shape data are integrated, and a three-dimensional shape inspection method is provided.

  According to the present invention, it is possible to provide a three-dimensional shape inspection method and apparatus that ensure high measurement accuracy regardless of the shape of a measurement object by complementarily combining a plurality of three-dimensional shape measurement methods and surface measurement methods. Can do.

It is a block diagram which shows the structure of the three-dimensional shape inspection apparatus which concerns on Example 1 of this invention. It is a flowchart which shows the procedure of the complementary integration which concerns on Example 1 of this invention. It is a figure which shows the direction and the sample which concern on Example 1 of this invention. It is a flowchart which shows the integration procedure of the point cloud data and image data which concern on Example 1 of this invention. It is a figure which shows the registration in the feature point of the point cloud data and image data which concern on Example 1 of this invention. It is a flowchart which shows the procedure of the complementary integration which concerns on Example 1 of this invention. It is a flowchart which shows the procedure of the complementary integration which concerns on Example 1 of this invention. It is a figure which shows the similarity of the sample which concerns on Example 1 of this invention. The flowchart which shows shape quantification by the self-reference shape data which concerns on Example 1 of this invention is shown. It is a figure which shows the derivation | leading-out method of the self-reference shape data which concerns on Example 1 of this invention. The flowchart which shows shape quantification by the average shape data which concerns on Example 1 of this invention is shown. It is a figure which shows the derivation method of the average shape data which concerns on Example 1 of this invention. It is a flowchart which shows the procedure of shape defect quantification using the self-reference shape data and CAD data which concern on Example 1 of this invention. It is a flowchart which shows the procedure of shape defect quantification using the self-reference shape data which concerns on Example 1 of this invention, and average shape data. It is a figure which shows the detection of the defect from the difference of the feature-value of the normal part and defect part of the sample surface imaged with the two-dimensional camera image which concerns on Example 1 of this invention. It is a flowchart which shows the procedure which detects a defect from the difference in the feature-value of the normal part and defect part of the sample surface which were imaged with the two-dimensional camera image which concerns on Example 1 of this invention. It is a flowchart which shows the procedure which derives | leads-out the influence degree of the drilling tool which concerns on Example 1 of this invention. It is a figure which shows the inspection parameter of the drilling tool which concerns on Example 1 of this invention. It is a figure which shows GUI which displays the test result which concerns on Example 1 of this invention. It is a block diagram which shows the structure of the three-dimensional shape inspection apparatus which concerns on Example 2 of this invention.

  Inspecting a three-dimensional shape requires measuring the three-dimensional shape, comparing it with a reference model, quantifying the shape defect, and further estimating the degree of influence from the defect quantitative value. Here, the influence degree is obtained by quantifying the influence given to the index representing the performance of the product in the case of a product, or to the index representing the machining performance in the case of a machining tool.

  The three-dimensional measuring devices in Patent Documents 1 and 2 each have problems with respect to the above technique. There is a tendency for the accuracy of edges and sharp corners to be insufficient at the time of measurement, CAD data is often required when comparing measured data, inspection is not possible when CAD data is not at hand, and shape defect values Therefore, it may not have the function of estimating the degree of influence.

  Embodiments of the present invention made in view of the above problems will be described below with reference to the drawings.

  A first embodiment of the present invention will be described with reference to FIGS.

  FIG. 1 shows the configuration of the three-dimensional measuring apparatus. Sample 1 is held by holding mechanisms 101 and 102. Here, the entire sample 1 and the holding mechanisms 101 and 102 are connected to the servo motor 103 and have a rotation mechanism centered on the y-axis on the xz plane. Here, the holding mechanisms 101 and 102 have an appropriate holding force that does not cause a deviation between the rotation amount of the servo motor 103 and the rotation amount of the sample 1. The relative position between the sample 1 and the image capturing unit 120 and the point group measuring unit 130 is set by the rotation of the servo motor 103. Here, the relative position between the sample 1 and the image capturing unit 120 and the point cloud measuring unit 130 is such that the inspection coverage ratio (= inspectable area / total area sample) is increased, that is, the inspection region is larger. It arrange | positions to the holding mechanisms 101 and 102 so that it may become large. Sample 1 is a processed product that requires quality assurance by three-dimensional shape measurement or a processing tool that requires shape measurement for processing accuracy management.

  The sample 1, the holding mechanisms 101 and 102, and the servo motor 103 are all held by a base 105, and the base 105 is mounted on an x stage 106, a y stage 107, and a θ stage 108. The rotation direction of the θ stage 108 is in the xy plane, and the θ axis is orthogonal to the xy plane. The x stage 106, the y stage 107, the θ stage 108, and the base 105 are mounted on an anti-vibration surface plate 110. The servo motor 103 is controlled by a control PC 140 through a motor controller 104, and the x stage 106, y stage 107, and θ stage 108 through a three-axis stage controller 109.

  In the three-dimensional measuring apparatus shown in FIG. 1, the surface state and shape of the sample 1 are measured by the image capturing unit 120 and the point group measuring unit 130. In the image capturing unit 120, the illumination unit 121 illuminates the sample 1 from an arbitrary direction, and the reflected light, scattered light, diffracted light, and diffused light are captured by the two-dimensional camera 123 using the lens 122, and the three-dimensional shape is captured. Are acquired as two-dimensional image data. A lamp, LED (Light Emitting Diode), or the like can be used for the illumination unit 121, and FIG. 1 shows illumination from a single direction, but the illumination direction may be a plurality of directions, or ring illumination may be used. . In addition, even when illuminating from a single direction, it has a mechanism that allows the illumination direction to be freely set, and irradiates illumination light from a direction that reveals surface irregularities and shapes according to the surface state and shape of the sample 1. Can do. For the two-dimensional camera 123, a CCD (Charge Coupled Device) image sensor, a CMOS (Complementary Metal Oxide Semiconductor) image sensor, or the like can be used. Here, the pixel pitch of the two-dimensional camera 123 is set to be finer than the resolution determined from the magnification and aperture ratio of the lens 122. The two-dimensional camera 123 is controlled by the control PC 140 via the camera controller 124, and the measurement result is output to the monitor 141.

  The point cloud measurement unit 130 includes a point cloud measurement sensor 131 and a sensor controller 132, is controlled by the control PC 140, and the measurement result is output to the monitor 141. The point group measurement sensor 130 includes a non-contact optical type, a contact type probe type, and the like, and is a device that measures the shape of the object surface and outputs three-dimensional coordinates of a large number of points as a point group. Many methods have been proposed for the optical point group measurement sensor, and any method can be applied to the present invention. As a method that can be used in the present invention, a light cutting method based on triangulation, a TOF (Time Of Flight) method in which light is applied to an object and the distance is measured according to the time when the light returns, and the focal point of the lens is moved. There are a lens focus method for measuring a distance by focusing, a moire method for projecting a lattice pattern onto an object and measuring the shape from a pattern pattern deformed according to the shape of the object surface, and an interference method using white interference. In recent years, a point cloud measurement method using an optical frequency comb having a large number of optical frequency modes arranged at equal intervals in a frequency space and a point cloud measurement method using a frequency feedback laser have been proposed.

  Although details will be described later, the apparatus of FIG. 1 performs complementary measurement of the surface state and shape of the sample 1, and the complementary integration unit 1401 that integrates the image of the two-dimensional camera 123 and the data of the point cloud measurement sensor 130. A storage unit in which CAD (Computer Aided Design) data 142 representing a 3D shape connected to the control PC 140 is stored, a region specifying unit that specifies a region in which shape data is acquired by the point cloud measurement sensor 131 from the CAD data 142, A defect quantification unit 1402 and a defect quantification unit 1402 for quantifying the shape defect value by comparing the CAD data 142 or the self-reference shape data derived from the similarity of the sample 1 with the data integrated by the complementary integration unit 1401 Database based on past experimental results and simulation results (DB) 143 refers to the, the doneness of the product OK, with a decision unit that performs up to NG judgment or extent of judgment.

  In the following, by combining the point cloud measurement method proposed in the present invention and the advantages of images, complementary integration for more accurately measuring 3D shapes and surface irregularities, and 3D information on the point of interest of the sample restored by complementary integration A defect quantification method for quantifying the shape defect and a method for estimating the influence of the shape defect on the final product quality and the processing accuracy of the machining tool will be described in detail below.

(Complementary integration)
The point cloud measurement method is suitable for grasping the global shape, but the measurement accuracy is reduced for local changes and minute irregularities. In the contact type using a probe, the shape measurement accuracy tends to be lowered at an edge or an acute angle portion where it is difficult to scan the probe. Even in a non-contact type using laser or other optics, a shape having an edge, an acute angle portion, and a steep inclination is greatly different from a shape having a flat light reflection angle, and the measurement accuracy tends to be lowered. Furthermore, the point cloud measurement method is not good at measuring minute irregularities such as the surface roughness of the sample.

  Here, when a shape having an edge portion, an acute angle portion, and a steep inclination is acquired under illumination that is a two-dimensional camera, the brightness difference from the surroundings in the acquired image increases or the brightness change increases. have. Therefore, it is possible to restore from the image information on the edge portion, the acute angle portion, and the shape having a steep inclination that the point cloud measurement method is not good at. Further, the shape of the surface state and minute unevenness can be derived from the shadow in the image. However, the global shape that the point cloud measurement method is good at is an image with a flat intensity distribution, and there are few characteristic parts, so that it may be difficult to restore the shape.

  In the present embodiment, 3 points which are more accurate than the case where only the point group measurement or only the two-dimensional camera image is used are compensated by using the image picked up by the two-dimensional camera for the portion which is not good for the point group measurement method. This paper proposes complementary integration to realize dimensional shape measurement. Hereinafter, a specific example of the complementary integration method will be shown.

  FIG. 2 shows a complementary integration procedure in shape measurement. The CAD data representing the three-dimensional shape of the sample includes information representing the direction of the surface constituting the shape. The point cloud measurement method has a direction and an inclination of a surface whose measurement accuracy is guaranteed by the method. An area in which the measurement accuracy is guaranteed is an area that can be measured by the point cloud measurement sensor. The CAD data is read (S100), the measurement area of the sample 1 is set (S101), the arrangement when holding the sample 1 shown in FIG. 1 and the relative position of the rotation axis and the point cloud measurement sensor are calculated (S102). The slope β between the sample surface and the point group measurement sensor shown in FIG. 3 is determined to be equal to or less than a threshold value determined by the characteristics of the point group measurement sensor (S103), and the region of the surface where the point group measurement is possible is specified. Specify by part. For the area below the threshold, a point cloud representing the shape is measured by the point cloud measurement sensor (S104). Since the accuracy of the point cloud measurement sensor is not guaranteed for the area above the threshold value, it is imaged with a two-dimensional camera (S105), and shape data is estimated from the captured image (S106). The point cloud of the point cloud measurement sensor and the position information of the shape data estimated from the image are collated (S107), and desired shape data is acquired (S108).

  There are a plurality of methods for estimating shape data from an image captured by a two-dimensional camera (S105). In the apparatus configuration of FIG. 1, the sample 1 can be imaged by the two-dimensional camera 123 from many viewpoints by controlling the rotation amount of the servo motor 103. Using these images from a plurality of viewpoints, shape data such as an edge and an acute angle portion that the point cloud measurement sensor is not good at is estimated. For the estimation of shape data, the stereo method, which is a depth measurement method based on triangulation, the inverse rendering method that estimates the tilt of the measurement surface from the shadow of the image, and the view volume intersection that estimates the contour of the sample from the silhouette of multiple images Use the law. The stereo method uses two images obtained by capturing the same visual field from different directions. Images from different directions are acquired by rotating the servo motor 103 of FIG. 1 by a certain amount and controlling the relative position with the two-dimensional camera 123. For example, after acquiring the first image, the sample 1 is rotated 5 degrees by the servo motor 103 to acquire the second image. The depth is calculated from these two images using the principle of triangulation. By rotating the servo motor 360 degrees in increments of 5 degrees, the shape data of the surface parallel to the rotation axis of the sample 1 can be acquired. In the inverse rendering method, the illumination state and the object reflectance are known, and a geometric shape is estimated from an image obtained by a two-dimensional camera using a rendering equation. It is possible to acquire object shape data from a single image. Similar to the stereo method, by controlling the relative position with the two-dimensional camera 123 by the rotation of the servo motor 103, the shape data of the surface parallel to the rotation axis of the sample 1 can be acquired. The view volume intersection method is a technique for deriving the shape of the target object by deriving a view volume that is a cone having a viewpoint as a vertex and a silhouette as a cross section, and obtaining a common part of the view volumes at all viewpoints. The servo motor 103 is rotated to acquire an image of a plane parallel to the rotation axis of the sample 1 by 360 degrees with the two-dimensional camera 123, and the viewing volume is derived from the silhouette of each image. Get shape data from the part. To obtain a silhouette from an image, calculate the difference between the input image containing the target object and the background image prepared in advance to extract the silhouette, and statistical values for the common part of multiple images with different viewpoints. There is a Davis method for extracting silhouettes using the.

FIG. 4 shows an example of position collation (S107). The acquired point cloud data (S200) and image data (S202) have different spatial resolutions determined at the time of designing the point cloud measurement unit and the image capturing unit. there,
As shown in FIG. 5, the spatial resolution of the point cloud data is adjusted to be the same scale as the image data acquired by the image capturing unit (S201). Here, the adjustment of the scale may be performed on the image processing unit as long as the point cloud data and the image data have the same spatial resolution, or may be applied to both the point cloud data and the image data. If the point cloud data and the image data are designed to have the same spatial resolution, this scale adjustment (S201) is not necessary. After the scale adjustment (S201), the point cloud data and the image data are roughly adjusted based on the difference in the geometric arrangement between the point cloud measurement unit and the image capturing unit (S203). After the coarse adjustment (S203), detailed position matching is performed at the common feature points in both the data 161 of the point 161 where the displacement changes sharply in the point cloud data of FIG. 5 and the location 162 where the intensity change is large in the image. (S204). When there are a plurality of feature points, position matching is performed so that the sum of the distances between the feature points is minimized. Alternatively, the position matching may be performed so that the distance between the feature points is statistically an intermediate value.

FIG. 2 shows a complementary integration flow when acquiring shape data. For measurement of minute irregularities below the accuracy of the point cloud measurement method such as surface roughness which is not good for the point cloud measurement method, FIG. The resolution determined by the pixel size of the lens 122 and the two-dimensional camera 123 is set to a height that can sufficiently capture the minute unevenness to which attention is focused, and an image is acquired.
FIG. 2 shows a method for calculating a region for point cloud measurement using CAD data, but the region for using a two-dimensional image is determined using the comparison result between the acquired point cloud data and CAD data shown in FIG. In this way, complementary integration with a two-dimensional image is possible. An area for measuring the shape of the sample is set (S300), and point cloud measurement is performed (S301). The CAD data (S302) of the sample is input, and the distance between corresponding points is calculated by aligning with the point group measured in (S301) (S303). For this collation, an ICP (Irative Closest Point) method widely used for collation of point clouds is used. In the ICP method, the distance between the measured point group and each corresponding point in the CAD data is averaged, and the average value is compared with a certain threshold value (S304). Since the accuracy of the point cloud measurement sensor is not guaranteed for the area above the threshold value, it is imaged with a two-dimensional camera (S305), and shape data is estimated from the captured image (S306). The point cloud of the point cloud measurement sensor and the position information of the shape data estimated from the image are collated (S307), and the desired shape data is acquired (S308). Compare actual measurement results with CAD to determine whether or not to use point cloud measurement results, so outliers in point cloud measurement that are not possible with the flow shown in FIG. 2 can be removed, improving measurement accuracy To do. Note that the threshold value to be compared with the average of the distances between corresponding points is set to a value larger than the shape defect or surface unevenness to be inspected.

  In the methods shown in FIGS. 2 and 6 described so far, either point cloud data or image data is acquired to restore the shape data. However, both of the data are acquired and an edge or an acute angle portion is measured in the shape measurement. May use image data, otherwise point cloud data may be used. Further, image data may be used for the inspection of surface irregularities. FIG. 7 shows a measurement flow in which both point cloud data and image data are acquired before CAD data comparison. An area for measuring the shape of the sample is set (S400), point cloud measurement is performed on the set area (S401), and an image is also taken (S402). After the measurement, the CAD data is read (S403), and it is determined whether the inclination β between the sample surface and the point group measurement sensor is equal to or less than the threshold value determined by the characteristics of the point group measurement sensor (S404). Clarify the surface area. The shape data measured by the point cloud measurement sensor is adopted for the area below the threshold (S406), and the shape data is calculated from the shadow of the image in the area above the threshold (S405). Then, the point cloud measurement data and the shape data calculated from the image are aligned (S407), and the shape data is output (S408). When both point cloud data and image data are acquired before CAD data comparison, an overlapping portion is required when aligning the point cloud data and image data, and both data are acquired over the entire area. Will increase, but there will be no shortage of overlap. Therefore, by acquiring both the point cloud data and the image data before the CAD data comparison, it is possible to use highly accurate data out of the point cloud data and the image data, and further improve the measurement accuracy. it can.

(Defect quantification)
In order to quantify the shape defect, the measured shape is quantitatively compared with the reference shape. As a reference shape, either CAD data that is an ideal shape, average self-reference shape data calculated using the similarity of the sample itself, average shape data of a plurality of samples of the same shape, or a plurality of types Is used. The method for deriving each shape data, the method for quantifying the defect, and the combination will be described below.

  When CAD data is used, the difference between the measured shape data and CAD data is taken and the difference value is output as a shape defect value.

  A method for deriving self-reference shape data using the similarity of the sample itself and shape defect quantification using the self-reference shape data will be described. FIG. 8 shows a six-blade impeller 170 as an example of a sample to be measured. The six wings 170a to 170f have the same shape. 1 is mounted so that the impeller central axis 1701 and the y axis coincide with each other, and the point cloud measuring sensor 131 and the two-dimensional camera 123 are rotated by the servo motor 103 about the y axis by the method described above. Acquire shape and surface data. FIG. 9 shows a flow for deriving self-reference shape data. The shape / surface data is acquired by the inspection apparatus of FIG. 1 (S501), and the similarity is recognized from the shape / surface data (S502). A waveform representing the shape of the impeller 170 is shown in FIG. In FIG. 10A, the waveform profiles 171a to 171f of the wings 170a to 170f are identical in design. Recognition of the similarity between the profiles 171a to 171f may be performed manually by the apparatus operator, or may be automatically recognized based on a pattern recognition technique. If there are three or more parts having the same shape, the self-reference shape data 171 is calculated from their statistical average value (S503). The statistical average value is calculated by the median or the average value is calculated by omitting the maximum value and the minimum value in order to remove the influence of the shape defect and the surface irregularity defect from the self-reference shape data. Here, it is assumed that the probability of occurrence of a shape defect or surface irregularity is low. The self-reference shape data 171 and the profiles 171a to 171f are aligned (S504), the difference between them is calculated as shown in FIG. 10B (S505), and the difference value is output as a shape defect value. (S506). Since the similarity of the sample itself is used, shape defects can be quantified without CAD data.

  Next, a method for deriving the average shape data of a plurality of samples having the same shape and quantification of the shape defect using the average shape data will be described. The derivation flow of average shape data is shown in FIG. The shape / surface data is acquired by the inspection apparatus of FIG. 1 (S601). A waveform representing the shape of the impeller 170 is shown in FIG. FIG. 12A shows the entire profiles 172a to 170n when the impeller is measured n times. The average shape data 172 is calculated from the statistical average value of these profiles (S602). The statistical average value is calculated by the median or the average value is calculated by omitting the maximum value and the minimum value in order to remove the influence of the shape defect and the surface irregularity defect from the average shape data. Here, it is assumed that the probability of occurrence of a shape defect or surface irregularity is low. The average shape data 172 and the profiles 172a to 171n are aligned (S603), the difference between them is calculated as shown in FIG. 12B (S604), and the difference value is output as a shape defect value ( S605). Self-referenced shape data cannot be used unless the sample itself is similar. On the other hand, since the average shape data is derived from a plurality of samples, comparative inspection is possible even for samples having no CAD data and no similarity.

  Next, a description will be given of shape defect quantification combining three standard shape data: CAD data, self-reference shape data, and average shape data. When the number of samples is insufficient to calculate the average shape data, the shape defect is quantified using self-reference shape data and CAD data, as shown in FIG. The shape / surface unevenness data is measured by the apparatus of FIG. 1 (S701), and the presence or absence of similarity is determined (S702). As described above, similarity recognition may be performed manually by the apparatus operator, or may be automatically recognized based on a pattern recognition technique. If there is no similarity, the difference between the measured shape data and CAD data is taken and the difference value is output as a shape defect value (S703). If there is similarity, the similar part calculates a difference value from the self-reference shape data according to the flow shown in FIG. 9 (S704), and the dissimilar part calculates a difference from the CAD data (S705). A defective value is output (S706).

  When the average data can be calculated, as shown in FIG. 14, the shape defect is quantified using the self-reference shape data and the average shape data. The shape / surface unevenness data is measured by the apparatus of FIG. 1 (S801), and the presence or absence of similarity is determined (S802). As described above, similarity recognition may be performed manually by the apparatus operator, or may be automatically recognized based on a pattern recognition technique. If there is no similarity, the difference between the measured shape data and the average shape data is taken according to the flow shown in FIG. 11, and the difference value is output as a shape defect value (S803). If there is similarity, the similar part calculates a difference value from the self-reference shape data according to the flow shown in FIG. 9 (S804), and the dissimilar part calculates a difference from the average shape data (S805) A defective value is output (S806).

  When detecting defects such as irregularities on the sample surface, the defect may be detected from the difference in the feature quantity between the normal portion and the defective portion of the sample surface captured by the two-dimensional camera image. The feature amount represents the feature of the image, and includes a higher-order local autocorrelation (HLAC) feature amount. In the present invention, the position of the two-dimensional camera, the position of illumination, and the tilt of the sample are also used as the feature amount. The position of the two-dimensional camera 123 shown in FIG. 15 is the unit vector I301, the position of the illumination 121 is the unit vector S302, and the tilt of the sample in the Cartesian coordinate system is measured using a point cloud measurement sensor. The normal vector 304 of the triangular patch 303 formed by three adjacent points is used as a feature amount representing the inclination of the sample. When CAD data is present and only surface irregularities are inspected, CAD data may be used to derive the normal vector 304. The image obtained by the two-dimensional camera 123 depends on the position of the illumination 121, the position of the two-dimensional camera 123, and the tilt of the sample. The part can be determined more accurately. Further, the shielding portion 306 caused by the shape of the sample can be calculated, and the defect portion 305 and the normal portion can be more accurately discriminated in consideration of a change in image characteristics due to a decrease in luminance due to a shadow.

  FIG. 16 shows a flowchart. An image of the sample is measured with a two-dimensional camera (S1001). 25-dimensional HLAC features, illumination positions, 2-dimensional camera positions, and sample normal vector values for each pixel (or a value obtained by performing smoothing processing or the like from a plurality of peripheral pixels) acquired with a 2-dimensional camera Is calculated as a feature amount (S1002). Next, the normal part and the defective part are taught, and the boundary for discriminating between the normal part and the defective part in the feature amount space is determined (S1003). To determine the boundary, Support Vector Machine (SVM) or the like is used. Finally, a defective part is extracted using the boundary calculated in (S1003) (S1004). Note that the boundary determined in (S1002) may be determined using a sample before inspection.

(Product processing accuracy judgment)
In the three-dimensional shape inspection apparatus of FIG. 1, after quantifying the shape defect, if the shape defect value is a processed product, the performance is indicated by the product performance as an index, and the sample is a processing tool. Outputs a pass / fail judgment result using the machining accuracy as an index.

  An example of a drilling tool is shown in FIGS. The inspection procedure will be described with reference to the flow of FIG. A parameter representing the shape of the tool is designated (S901). In this example, the wear area 180 at the tip of the drilling tool in FIG. 18 and the angle 181 of the tip of the tip are used as evaluation parameters. Next, the shape of the drilling tool as a sample is measured by the above-described method, and compared with any of CAD data, self-reference shape data, or average shape data, and the shape defect value is quantified (S902). Then, the wear area and the angle of the blade edge are calculated from the shape defect value indicating the difference in shape by distance (S903). The influence of the wear area and the blade edge angle on the machining result is estimated from past experimental results and machining simulation results. Prepare a database that correlates the relationship between the depth and roundness of the hole that determines the performance of the drilling tool, the wear area 180 and the angle 181 of the cutting edge with past experimental results and machining simulation results. Depth and roundness are estimated (S904). Further, an increase in the wear area due to the number of uses and a change in the angle of the blade edge are also estimated, and the remaining number of uses is calculated. Further, in order to perform OK / NG determination of the drilling tool, a threshold is set in advance for the shape abnormality represented by the depth and roundness of the hole, and the size of the allowable value is determined (S905). If the allowable value is exceeded, an NG determination is output (s906). If it is within the allowable value, the number of times of use estimated from the OK judgment and the database is output (S907). In the case of a processing tool, it is possible to estimate the state of the processing tool from the shape of chips generated accompanying the processing, and the method of FIG. 17 can be applied as it is. FIG. 19 shows a GUI for inputting inspection conditions and outputting results. The GUI is displayed on a monitor 141 mounted on the inspection apparatus. The allowable values of the depth and roundness are input as the inspection area and the degree of influence on the machining, and the estimated values of the cutting edge inclination, the wear area, and the degree of influence on the machining derived from the measurement results are output. Further, the final determination value is output as the inspection result 191.

  According to the present embodiment, the point cloud measurement data and the two-dimensional camera image are used in a complementary manner, thereby realizing three-dimensional shape measurement with higher accuracy than when only the point cloud measurement data and the two-dimensional camera image are used. can do.

  Further, by using self-reference shape data generated by utilizing the similarity of the sample itself, it is possible to quantify the shape defect without design data.

  In addition, the quality and yield of products can be predicted by determining the quality based on the performance when using processed products and processing tools, and grasping the degree of quality. It is possible to manage the life prediction and the like considering the deterioration / wearing state of the machining tool and the reduction of machining accuracy.

  A second embodiment of the present invention will be described with reference to FIG. The apparatus configuration is shown in FIG. The basic configuration is the same as that of the first embodiment. The polarization camera 200 that can analyze the polarization state of the reflected light is used instead of the two-dimensional camera 123 of FIG. 1, and the ring illumination system 201 is used instead of the illumination unit 121. ing. In the polarization camera, a minute polarizer is attached in front of each pixel, and four pixels form one group, and the four pixels in the group have different polarizer orientations. It is known that the polarization state of the light reflected from the sample 1 varies depending on the direction of the surface, and the direction of the surface can be detected by integrating different polarization information obtained by the four pixels. By using the polarization information, the direction of the surface can be determined with high accuracy even if there is a color change such as a spot that is difficult to identify in a normal image. In addition, the polarization information is highly responsive to minute irregularities such as scratches, and surface irregularities can be effectively revealed.

  The embodiments described so far are merely examples of implementation in carrying out the present invention, and the technical scope of the present invention is not limitedly interpreted by these. That is, the present invention can be implemented in various forms without departing from the technical idea or the main features thereof. Further, the present invention may be implemented by combining the first and second embodiments.

DESCRIPTION OF SYMBOLS 1 ... Sample 101, 102 ... Holding mechanism 103 ... Servo motor 104 ... Motor controller 105 ... Base 106 ... x stage 107 ... y stage 108 ... θ stage 109- ..Stage controller 110 ... Anti-vibration surface plate 120 ... Image capturing unit 121 ... Illumination unit 122 ... Lens 123 ... Two-dimensional camera 124 ... Camera controller 130 ... Point group measurement Part 131 ... Point cloud measuring sensor 132 ... Sensor controller 140 ... Control PC
141 ... Monitor 142 ... CAD data 143 ... Database 144 ... Input device 160 ... One point 161 on point cloud data ... Point cloud data feature point 162 ... Image data feature point 170 ... impeller 1701 ... impeller central axis 170a, 170b, 170c, 170d, 170e, 170f ... wing 170 ... self-reference shape data 171a, 171b, 171c, 171d, 171e, 171f Profile 180 ... Wear area 181 ... Blade angle 190 ... GUI
191 ... Inspection result 200 ... Polarized camera 201 ... Ring illumination system 301 ... Two-dimensional camera direction unit vector 302 ... Illumination direction unit vector 303 ... Triangular mesh 304 ... Method Line vector 305 ... Defect 306 ... Shielding part 1401 ... Complementary integration part 1402 ... Defect quantification part 1403 ... Determination part

Claims (14)

A storage unit that stores reference data that is shape data of a reference model to be inspected;
An area specifying unit for specifying an area for acquiring the first shape data from the reference data;
A first three-dimensional shape sensor that acquires the first shape data for the region specified by the region specifying unit;
A second three-dimensional shape sensor that acquires second shape data different from the first shape data of the inspection object for a region other than the region specified by the region specifying unit;
A three-dimensional shape inspection apparatus comprising: a complementary integration unit that integrates the first shape data and the second shape data.   The three-dimensional shape inspection apparatus according to claim 1, further comprising a defect quantification unit that quantitatively evaluates the integrated data integrated by the complementary integration unit and the reference data.   The first three-dimensional shape sensor is a point group measurement sensor that measures a point group of a three-dimensional shape, and the second three-dimensional shape sensor is a two-dimensional camera that acquires a three-dimensional shape as two-dimensional image data. The three-dimensional shape inspection apparatus according to claim 1, wherein the three-dimensional shape inspection apparatus is provided.   The first three-dimensional shape sensor is a point group measurement sensor that measures a point group of a three-dimensional shape, and the second three-dimensional shape sensor is a polarization camera that measures a three-dimensional shape from a plurality of different polarization information. The three-dimensional shape inspection apparatus according to claim 1.   The three-dimensional shape inspection apparatus according to claim 1, wherein the storage unit stores self-reference shape data derived from the similarity of the inspection object itself as reference data.   The three-dimensional shape inspection according to claim 2, further comprising: a determination unit configured to determine a shape abnormality of the inspection target from a shape defect value obtained by the defect quantification unit and an experimental value or a simulation value. apparatus.   The determination unit performs a shape abnormality determination using any one of a position of an illumination unit that irradiates the inspection target, a position of the two-dimensional camera, and a tilt of the inspection target as a feature amount. 6. The three-dimensional shape inspection apparatus according to 6. Read the reference data that is the shape data of the reference model to be inspected,
Specify the area from which the first shape data is obtained from the reference data,
Obtaining the first shape data for the identified region;
Obtaining second shape data different from the first shape data of the inspection object for an area other than the specified area;
3. A three-dimensional shape inspection method, wherein the first shape data and the second shape data are integrated.   The three-dimensional shape inspection method according to claim 8, wherein the integrated data integrated by the complementary integration unit and the reference data are quantitatively evaluated.   9. The three-dimensional shape inspection method according to claim 8, wherein the first shape data is a point group of a three-dimensional shape, and the second shape data is an image captured by a two-dimensional camera.   9. The three-dimensional shape inspection method according to claim 8, wherein the first shape data is a point group of a three-dimensional shape, and the second shape data is an image captured by a polarization camera.   9. The three-dimensional shape inspection method according to claim 8, wherein the reference data is self-reference shape data derived from the similarity of the inspection object itself.   Determining the shape abnormality of the inspection object from the shape defect value obtained by quantitatively evaluating the integrated data integrated by the complementary integration unit and the reference data, and an experimental value or a simulation value The three-dimensional shape inspection method according to claim 9.   The three-dimensional shape inspection method according to claim 13, wherein a shape abnormality is determined using any one of a position of an illumination unit that irradiates the inspection target, a position of the two-dimensional camera, and a tilt of the inspection target as a feature amount.

Images