JP6420180B2 - Surface inspection apparatus and surface inspection method - Google Patents

Description

  The present invention relates to a surface inspection apparatus and a surface inspection method for inspecting surface irregularities of an object to be detected.

  Necessary metal parts such as rolling, cutting, and extruding are performed on metal materials at various production sites to produce the required metal parts, and injection molding, blow molding, extrusion molding, etc. are appropriately performed on resin materials. The required resin parts are manufactured by performing the resin molding process. Such metal parts and resin parts have irregularities such as dents, chips, and scratches due to contact with multiple members in the manufacturing process and parts transportation, etc. Contamination may occur on the surface of the component due to the mixing of etc. In addition, in resin parts, irregularities are also generated on the surface of the parts due to dents (so-called sink marks) due to shrinkage during molding of the resin material. Such irregularities on the surface of the component are not only undesirable in appearance, but may affect the performance of the component. For this reason, the surface condition of a part is inspected during the manufacturing process to maintain a certain quality. In many cases, such inspection of the surface of a product is performed by visual inspection of an operator, but in recent years, a method for inspecting surface irregularities by irradiating light onto an inspection surface has also been proposed. (For example, patent document 1).

JP 2000-28339 A

  However, in the inspection by the worker, there is a concern that the unevenness determination result may vary, or the inspection may be omitted due to the fatigue of the worker or a decrease in concentration. In addition, although the quality required for the parts is increased and it is required to inspect a slight surface unevenness, the size of the unevenness that can be detected is naturally limited when the inspection is performed by an operator. On the other hand, Patent Document 1 is an inspection apparatus that inspects surface irregularities using laser slit light, and detects irregularities by photographing scattered light scattered at the boundary between an irregular portion and a portion having no irregularity with a camera. Though conceivable, there is a problem that it is necessary to irradiate the entire surface with laser slit light, and it takes a lot of inspection time to inspect the entire surface.

  That is, the present invention has been proposed in view of the above-mentioned problems inherent in the prior art, and provides a surface inspection apparatus and a surface inspection method capable of accurately detecting unevenness of an object to be inspected. For the purpose.

In order to solve the above-mentioned problems and achieve the intended purpose, a surface inspection apparatus according to claim 1 of the present invention includes:
A surface inspection device for inspecting unevenness of an object to be inspected,
A light irradiation unit for irradiating light to a measurement region of the object to be inspected;
Imaging means for capturing an image of the inspected object in a state where light is irradiated by the light irradiation unit and acquiring image data of a measurement region;
Surface normal vector calculation means for calculating a surface normal vector of a measurement point in the measurement region based on image data captured by the imaging means;
Index value calculating means for calculating an index value by inner producting the surface normal vectors calculated by the surface normal vector calculating means;
Coordinate specifying means for specifying an uneven position in the measurement region based on the index value calculated by the index value calculating means,
An index value setting filter that sets a calculation point indicating a position where an inner product of the surface normal vector and an index value calculated based on the calculation point are set is set,
By sequentially changing the measurement points that match the calculation points of the index value setting filter at the measurement points in the measurement area, the corresponding reference points are obtained by inner producting the surface normal vectors of the decision points that match the calculation points for each change. By calculating the index value by the index value calculation means, each time the index value is calculated, the coordinates of the measurement region where the reference point is located and the index value of the coordinates are specified.
The gist of the invention is that the coordinate specifying means specifies coordinates of the measurement area corresponding to an index value out of a reference range among the index values calculated by the index value calculating means.
In this way, the surface normal vector corresponding to each measurement point in the measurement area of the object to be inspected is calculated, and the surface normal vector of the measurement point located at the calculation point of the template is inner-producted, so that the reference in the template An index value indicating the degree of unevenness of the position of the point can be calculated, and further, the measurement point to be matched with the calculation point is changed in order at the measurement point interval in the measurement area, so that it corresponds to each coordinate position of the measurement area. An index value can be obtained. Therefore, by specifying the coordinates corresponding to the index value out of the reference range, it is possible to accurately detect the uneven position of the object to be inspected. In addition, since the surface normal vector corresponding to each measurement point in the measurement area can be calculated based on image data captured in a state where light is irradiated over a wide area of the measurement area, the process of calculating the uneven position It is possible to increase the speed and specify the uneven position present on the object to be inspected in a short time.

The surface inspection apparatus according to claim 2 comprises:
In the index value setting filter, the calculation points are arranged in a ring centered on the reference point,
The index value calculation means calculates the index value of the reference point by inner product of the surface normal vectors of the measurement points that coincide with the calculation points located so as to sandwich the reference point in the index value setting filter. This is the summary.
In this way, by calculating the index value by inner product of the surface normal vectors of the measurement points that coincide with the calculation point located across the reference point, the index value when the reference point is at the uneven position and the reference point The difference from the index value in the case where is not at the uneven position can be made significant, and the coordinates of the uneven position can be specified with high accuracy.

The surface inspection apparatus according to claim 3 is:
In the index value setting filter, a plurality of sets of calculation points are arranged in a ring centered on the reference point,
The index value calculation means is configured to calculate an index value of a reference point by calculating an average value of inner products obtained by inner product of surface normal vectors for each measurement point located at each set of calculation points. Is the gist.
In this way, by using the average value of the inner product values calculated by inner product of multiple sets of surface normal vectors as the index value of the reference point, the surrounding surface state where the reference point is located at the time of calculating the index value Can be appropriately reflected in the index value, and the unevenness position can be specified with higher accuracy using the index value.

The surface inspection apparatus according to claim 4 is:
An arbitrary index value calculated by the index value calculation means is used as a reference index value, and it is determined whether or not the reference index value is a minimum value among index values within a predetermined comparison area including the reference index value. And an effective index value setting means for setting the reference index value as an effective index value in the case of a minimum value,
The effective index value setting means is configured to determine whether each index value calculated by the index value calculation means is an effective index value as a reference index value in order,
The coordinate specifying unit is configured to specify the coordinates of the measurement region corresponding to the index value out of the reference range among the index values determined as the effective index value by the effective index value setting unit. And
As described above, by specifying the concave / convex position based on the effective reference value that is the minimum value in the predetermined comparison region, the center of the concave / convex portion generated on the surface of the object to be inspected can be accurately specified. Moreover, the coordinate detected as an uneven | corrugated position about one uneven | corrugated | grooved part can be suppressed.

In order to solve the above problems and achieve the intended purpose, a surface inspection method according to claim 5 of the present invention includes:
A surface inspection method for inspecting unevenness of an object to be inspected,
Calculate the surface normal vector of the measurement points in the measurement area of the object to be inspected based on the image data captured by the imaging means,
A measurement point that matches the calculation point by matching the calculation point of the index value setting filter that defines the calculation point indicating the inner product of the surface normal vector and the reference point corresponding to the calculation point to the measurement point in the measurement area In addition to calculating the index value corresponding to the reference point by inner product of the surface normal vector of
Each time the index value is calculated, the index value corresponding to each coordinate of the measurement area is calculated by changing the measurement point that matches the calculation point of the index value setting filter and calculating the index value corresponding to the reference point for each change. Identify
The gist is to detect the concave / convex position of the detected object by specifying the coordinates corresponding to the index value out of the reference range in the measurement region.
In this way, the surface normal vector corresponding to each measurement point in the measurement area of the object to be inspected is calculated, and the surface normal vector of the measurement point located at the calculation point of the template is inner-producted, so that the reference in the template By calculating an index value that indicates the degree of unevenness of the position of the point, and sequentially changing the measurement point that matches the calculated point at the interval of the measurement point in the measurement area, the index value corresponding to each coordinate position of the measurement area Can be sought. Therefore, by specifying the coordinates corresponding to the index value that is out of the reference range, the uneven position of the object to be inspected can be automatically detected with high accuracy. In addition, since the surface normal vector corresponding to each measurement point in the measurement area can be calculated based on image data captured in a state where light is irradiated over a wide area of the measurement area, the process of calculating the uneven position It is possible to increase the speed and specify the uneven position present on the object to be inspected in a short time.

The surface inspection method according to claim 6 comprises:
Based on an index value setting filter in which the calculation points are arranged in a ring centered on the reference point, a reference is obtained by inner producting the surface normal vectors of the measurement points that coincide with the calculation points located so as to sandwich the reference point. The gist is to calculate an index value corresponding to a point.
In this way, by calculating the index value by inner product of the surface normal vectors of the measurement points that coincide with the calculation point located across the reference point, the index value when the reference point is at the uneven position and the reference point The difference from the index value in the case where is not at the uneven position can be made significant, and the coordinates of the uneven position can be specified with high accuracy.

The surface inspection method according to claim 7 comprises:
Based on an index value setting filter in which a plurality of sets of calculation points are arranged in a ring centered on the reference point, an average of inner product values obtained by inner product of surface normal vectors for each measurement point that matches the calculation points of each group The gist is to calculate the value as an index value corresponding to the reference point.
In this way, by using the average value of the inner product values calculated by inner product of multiple sets of surface normal vectors as the index value of the reference point, the surrounding surface state where the reference point is located at the time of calculating the index value Can be appropriately reflected in the index value, and the unevenness position can be specified with higher accuracy using the index value.

The surface inspection method according to claim 8 comprises:
When each index value is set as a reference index value in order, a reference index value that is a minimum value within a predetermined comparison area including the reference index value is set as an effective index value, and the effective index value is out of the reference range in the measurement area. The gist is to detect the uneven position of the detected object by specifying the corresponding coordinates.
As described above, by specifying the concave / convex position based on the effective reference value that is the minimum value in the predetermined comparison region, the center of the concave / convex portion generated on the surface of the object to be inspected can be accurately specified. Moreover, the coordinate detected as an uneven | corrugated position about one uneven | corrugated | grooved part can be suppressed.

  The surface inspection apparatus and the surface inspection method according to the present invention can accurately detect the unevenness of the object to be inspected.

It is the schematic which shows the whole structure of the surface inspection apparatus which concerns on an Example. (a) shows the state of the surface normal vector in the plane area, (b) shows the state of the surface normal vector in the convex part, (c) is a schematic diagram showing the state of the surface normal vector in the concave part . Note that (c) also shows the relationship between the uneven portion and the index value setting filter. (a) shows the relationship between a calculation point and a reference point that form a pair in an annular index value setting filter with a diameter of 7 pixels as an example, and (b) shows a rectangular index value setting filter with a side of 6 pixels as an example. The relationship between the calculation point and the reference point that are paired in is shown. (a) is a conceptual diagram showing a state in which the index value is calculated by applying the filter while setting the index value setting filter with respect to the surface normal vector map in which the surface normal vector at each measurement point has been calculated, (b) (C) is a conceptual diagram showing the correspondence between index values of each coordinate calculated based on the index value setting filter, and (c) is a comparison area filter applied based on an index value map in which index values are associated with each coordinate. It is a conceptual diagram which shows the state which determines an effective index value for the first time. It is a conceptual diagram showing a state of determining an effective index value based on the reference value of the index value map, (a) shows the index value map and the effective index value map before the determination of the effective index value, (b) is The state of determining whether or not the index value of the index value map coordinates (RA2, LA2) is an effective index value is shown. (C) is whether the index value of the index value map coordinates (RA2, LA3) is an effective index value. (D) shows a state in which the effective index value determining means is created by determining whether or not the index value of each coordinate in the index value map is an effective index value. It is the graph which displayed the distribution of the index value in an index value map three-dimensionally with the vertical axis as the index value and the horizontal axis as the coordinate position. It is a table | surface which shows the samples (a)-(j) which concern on an experiment example. It is an image displayed by color according to the direction of the surface normal vector calculated for samples (a) to (j) according to the experimental example, and “x” is attached to the center of the uneven position. The images are color-coded according to the index values calculated for the samples (a) to (j) according to the experimental example, and “x” is attached to the center of the uneven position. It is a flowchart which shows the surface inspection process performed with the surface inspection apparatus of an Example.

  Next, the surface inspection apparatus and the surface inspection method according to the present invention will be described in detail below with reference to the accompanying drawings by giving preferred embodiments. The present invention calculates the surface normal vector of the measurement point T included in the measurement region L of the device under test 10 based on the image data of the device under test 10 captured by the camera 20, and uses the calculated surface normal vector. By calculating the index value indicating the degree of unevenness by the inner product in association with the coordinates of the measurement region L, and specifying the coordinates corresponding to the index value outside the reference range in the measurement region L, The uneven position is detected. Specifically, the calculation point IP indicating the position of the measurement point T for inner product of the surface normal vector and the calculation point IP of the index value setting filter FR defining the reference point M1 corresponding to the calculation point IP are set in the measurement region L. The index value corresponding to the reference point M1 is calculated by inner product of the surface normal vectors of the measurement point T that coincide with the calculation point IP in accordance with the measurement point T, and the index value is set every time the index value is calculated. By changing the measurement point T that matches the calculation point IP of the filter FR in the measurement region L and calculating the index value corresponding to the reference point M1 for each change, the index value corresponding to each coordinate of the measurement region L is calculated. Based on this, the coordinates of the measurement area L that are index values that are out of the reference range are specified, thereby detecting the uneven position of the object 100 to be inspected. Therefore, by describing a surface inspection apparatus using the surface inspection method of the present invention, a specific method of the surface inspection method will be described together.

  FIG. 1 is a schematic diagram illustrating an overall configuration of a surface inspection apparatus 10 according to an embodiment. The surface inspection apparatus 10 according to the embodiment includes a light irradiation unit 12, 14, 16 that irradiates light to the measurement region L of the object 100 to be inspected and a state in which light is irradiated by the light irradiation unit 12, 14, 16. A camera (imaging means) 22 that captures an image of the body 100 and acquires image data, and a control device 30 that identifies the uneven position in the measurement region L based on the image data acquired by the camera 20 are provided. Here, the surface inspection apparatus 10 according to the embodiment includes three light irradiation units 12, 14, and 16 that irradiate light from different directions with respect to the measurement region L of the inspection object 100. By capturing the measurement region L irradiated with light by each of 12, 14, and 16 with the camera 20, three types of image data with different light source directions capturing the same measurement region L of the device under test 100 are acquired. It is configured to Then, on the basis of the image data corresponding to the light irradiators 12, 14, and 16 of the inspection object 100 captured by the camera 20, each pixel (measurement point T) corresponding to the measurement region L of the inspection object 100 is measured. ) And the light source direction information of each image data, the control device 30 calculates the surface normal vector of each measurement point T, and the uneven region with respect to the planar region in the measurement region L based on the surface normal vector. By indexing the degree of unevenness of the (uneven portion 104), the position (uneven position) of the uneven portion 104 is specified.

(About light irradiation unit and camera)
Here, for each of the light irradiators 12, 14, and 16, a light source that irradiates light of the same wavelength may be used, or a light source that irradiates light of a different wavelength may be employed. It is also possible to employ light with a broad wavelength (white light) or light with a specific wavelength (red light or blue light visible light, or invisible light such as infrared light). In the embodiment, LED lamps that emit light of the same color are employed as the light sources of the light irradiators 12, 14, and 16. The camera 20 may be a CCD image sensor using a charge coupled device (CCD) as an imaging device or a CMOS image sensor using a complementary metal oxide semiconductor (CMOS). What is necessary is just to be able to sense the light irradiated from 12, 14, and 16 and reflected by the object under test 100. The image data picked up by the camera 20 is raster data in which values obtained by quantizing at least the brightness (luminance) of each pixel included in the predetermined measurement region L in the inspection object 100 are arranged in a two-dimensional array format. Is stored in the data storage means 42 included in the control device 30. The data storage means 42 is a temporary storage device such as a RAM. Further, the surface inspection apparatus 10 is provided with a display device 50 such as a liquid crystal display so as to be connected to the control device 30, so that the image of the inspected object 100 captured by the camera 20 can be displayed on the display device 50. It has become.

(About the control unit)
As shown in FIG. 1, the control device 30 includes a surface normal vector calculation unit 32 that calculates a surface normal vector at each measurement point T in the measurement region L based on image data captured by the camera 20, and a surface method. An index value calculation unit 34 that calculates an index value based on the surface normal vector calculated by the line vector calculation unit 32, and an unevenness portion 104 in the measurement region L based on the index value calculated by the index value calculation unit 34. Coordinate specifying means 36 for specifying a position (uneven position) is provided. Here, the index value calculation means 34 is configured to calculate an index value for each coordinate in the measurement region L as will be described later, and coordinates whether the index value is included in a predetermined reference range. The position of the concavo-convex portion 104 is specified by determination by the specifying means 36.

(About surface normal vector calculation means)
The surface normal vector calculation means 32 is configured to calculate the surface normal vector data of the measurement point T by the illuminance difference stereo method based on the image data having different light source directions input from the camera 20. . Here, the illuminance difference stereo method is an image process for calculating a surface normal vector of the measurement surface based on the luminance (brightness) of the reflected light of the light irradiated on the measurement surface (the surface of the inspection object 100). Is the method. The principle of this illuminance difference stereo method will be briefly described below. In the illuminance difference stereo method, it is assumed that the measurement surface is a perfect diffuse reflection surface (Lambert reflection surface), and the brightness of the measurement surface to be observed depends on the orientation with respect to the light source (light irradiation units 12, 14, 16). This is a technique for estimating the surface normal vector of the measurement surface by utilizing the change.

As shown in FIG. 1, when each image data is acquired by alternately emitting a plurality of (three) light sources from different directions with respect to the surface normal direction of the measurement surface and illuminating the measurement surface, If the brightness is L, the reflectance of the measurement surface is ρ, and the brightness of the measurement surface (measurement point T) imaged by the camera is V, the following equations (1) and (2) hold: Thus, Equation (3) of Equation 2 is obtained. The vector n ^T represents a surface normal vector at an arbitrary measurement point T on the measurement surface, and the vector s represents a light source direction vector with respect to the measurement point T. In each equation, a vector is represented by an arrow symbol above the symbol. That is, by capturing at least three pieces of image data having different light source directions, the surface of each measurement point T is calculated from the brightness of each pixel (measurement point T) and the information on the light source direction of each image data based on Equation (3). A normal vector can be obtained. Then, the surface normal vector calculation means 32 calculates the surface normal vector for each measurement point T included in the measurement region L, so that the surface of the measurement region L corresponding to each measurement point T corresponds to the surface normal vector. A normal vector map MP1 (see FIG. 4) is created and stored in the data storage means. In FIG. 4, the positions corresponding to the respective measurement points T are represented by rectangular cells.

(About the index value calculation means 34)
As shown in FIG. 1, the index value calculating unit 34 calculates an index value representing the degree of unevenness for each coordinate of the measurement region L based on the surface normal vector calculated by the surface normal vector calculating unit 32. Configured to calculate. Here, as shown in FIG. 2A, in the flat region (planar region) in the measurement region L of the device under test 100, the surface normal vectors of the respective measurement points T are basically directed in the same direction. There is almost no difference in angle between the surface normal vectors. On the other hand, the surface normal vector at the measurement point T does not point in the same direction in the convex or concave portion (that is, the concave and convex portion 104) in the measurement region L of the inspection object 100, and the angle difference between the surface normal vectors. (See FIGS. 2B and 2C). That is, the inner product of the surface normal vectors of the measurement points T around the coordinates in the plane region in the measurement region L and the surface method of the measurement points T around the coordinates in the uneven portion 104 in the measurement region L. The value is different from the value obtained by inner product of line vectors. Therefore, the index value calculation means 34 calculates the index value by inner producting the surface normal vectors located around arbitrary coordinates of the measurement area L, so that the coordinate position is a plane area or an uneven part. 104 can be distinguished.

  Specifically, assuming that one surface normal vector to be inner product is “a”, the other surface normal vector is “b”, and the intersection angle of both surface normal vectors is “θ”, the following equation (3) 4) holds. If the surface normal vector calculated by the surface normal vector calculation means 32 is normalized to a unit vector, | a || b | = 1. That is, assuming that the component of the surface normal vector is vector a = (Nax, Nay, Naz) and vector b = (Nbx, Nby, Nbz), the index value calculation means 34 calculates the inner product based on the following equation (5). It is set to calculate a value. That is, in the inspected object 100, the index value obtained by inner product of the surface normal vectors of the measurement points T in the planar region is a value close to “1”, while the measurement points T in the concavo-convex portion 104 are measured. The index value obtained by the inner product of the surface normal vectors is a value lower than “1”.

(Specification of measurement point T for inner product of surface normal vector)
Here, the control device 30 includes a calculation point IP indicating a combination position where the index value calculation means 34 performs the inner product of the surface normal vectors, and a reference point M1 corresponding to the index value calculated based on the calculation point IP. Is set (see FIG. 3). Here, in the index value setting filter FR, each calculation point IP is arranged in accordance with the interval of the measurement points T in the surface normal vector map MP1, and each calculation point IP set in the filter FR is used as a surface method. It is set so that it can coincide with the position of the measurement point T in the line vector map MP1. That is, when calculating the index value, the index value calculation means 34 fits the index value setting filter FR to the surface normal vector map MP1, so that the inner product of the surface normal vectors is calculated based on the calculation point IP of the filter FR. The measurement point T to be measured is specified, and the coordinates where the reference point M1 is located in the measurement region L are specified. In other words, when the index value setting filter FR is applied to the surface normal vector map MP1, the index value corresponding to the coordinates of the measurement region L where the reference point M1 is located becomes the calculation point IP that forms a pair in the filter FR. It is calculated by inner producting the surface normal vectors of the corresponding measurement points T. In FIG. 3, the calculation point IP of the index value setting filter FR is represented by a rectangular cell, and the same symbol is attached to the cell of the calculation point IP that is the combination position.

  In addition, the index value calculation means 34 sequentially changes the measurement points T that match the calculation points IP of the index value setting filter FR at the measurement points T in the measurement region L (surface normal vector map MP1). It is set to calculate the index value of the corresponding reference point M1 by inner product of the surface normal vectors of the determination points T that coincide with the calculation point IP (see FIG. 4). That is, by moving the index value setting filter FR in accordance with the set interval of the measurement points T in the measurement region L (surface normal vector map MP1) and calculating the index value, the index value calculation means 34 has the index The coordinates of the measurement region L where the reference point M1 of the index value setting filter FR is located are determined at the timing of calculating the value. The index value calculated by the index value calculation unit 34 is stored in the data storage unit 42 as an index value map MP2 (see FIG. 4B) for specifying the coordinates of the measurement region L and the index value of the coordinates. The In FIG. 4B, the coordinate position corresponding to each reference point M1 is represented by a rectangular cell, and the index value stored corresponding to each coordinate is indicated by a triangle.

  Here, as shown in FIG. 3, the index value setting filter FR has a plurality of calculation points IP arranged in an annular shape centered on the reference point M1, and is located between the filter FR and the reference point M1. The calculation point IP to be set is set to be a combination position indicating the measurement point T for inner product of the surface normal vectors. That is, by calculating the index value by inner product of the surface normal vectors located at the calculation point IP located across the reference point M1, the degree of unevenness at the position of the reference point M1 is expressed by the index value. be able to. The index value setting filter FR has a plurality of calculation points IP indicating combination positions, and the index value calculation means 34 calculates the surface normal vector of the measurement point T that matches the calculation point IP of each combination. In addition to the inner product, the index value calculating means 34 is set to calculate the average value of the calculated inner product values as the index value. 3A shows an example of an annular index value setting filter FR having a diameter of 7 pixels in which calculation points IP are arranged in a circle, and FIG. 3B shows calculation points IP in a square shape. An example of a rectangular annular index value setting filter FR having 6 pixels on each side is shown.

  The interval from the reference point M1 to the calculation point IP in the index value setting filter FR is set according to the size (diameter) of the uneven portion 104 to be detected. That is, as shown in FIGS. 2 (b) and 2 (c), when the object 100 has an uneven portion 104, the concave portion or the most convex portion (hereinafter referred to as the maximum deformation) in the uneven portion 104. The angle difference between the surface normal vectors at the boundary with the plane region located across the point Z) increases. On the other hand, the amount of deformation decreases as the concavo-convex portion 104 approaches the vicinity of the maximum deformation point Z, and the direction of the surface normal vector approaches the direction of the surface normal vector in the plane region in the vicinity of the maximum deformation point Z. The angle difference between the surface normal vectors is reduced. Therefore, when the reference point M1 is located at the maximum deformation point Z of the concavo-convex portion 104, the calculation point IP and the index value setting filter FR are set such that the calculation point IP is located at the boundary portion (outer edge) of the concavo-convex portion 104. It is preferable to set the setting position of the reference point M1 according to the size of the uneven portion 104 to be detected (see FIG. 2C).

  Further, it is preferable that the calculation points IP arranged annularly in the index value setting filter FR are set so as to be aligned with the shape of the boundary portion (outer edge) of the uneven portion 104 to be detected. For example, when detecting the uneven portion 104 whose boundary portion (outer edge) with the planar region is nearly circular, arbitrary measurement points T arranged in a two-dimensional array in the measurement region L (surface normal vector map MP1). A plurality of calculation points IP are located in an annular shape centering on the reference point M1 so that the measurement point T located at an approximately equal distance from the reference point M1 is the calculation point IP. It is set as follows. That is, as shown in FIG. 2 (c), the maximum deformation of the concavo-convex portion 104 is obtained by using an annular index value setting filter FR in which the calculation points IP are arranged in a circle centered on the reference point M1. When the point Z is set as the reference point M1, each calculation point IP can be positioned at the boundary portion of the uneven portion 104, and the detection accuracy of the uneven portion 104 can be improved. In particular, when a minute uneven portion 104 is formed on the surface 102 of the object to be inspected 100, since the ratio of the boundary portion (outer edge) of the uneven portion 104 being close to a circle increases, the index value setting filter FR is set. The detection accuracy of the concavo-convex portion 104 can be increased by using an annular shape. As described above, since the measurement points T are arranged in a two-dimensional array in the surface normal vector map MP1, the measurement points on or adjacent to the circumference around the reference point M1. The index value setting filter FR is set so that the calculation point IP indicates T.

  Further, in the embodiment, the index value setting filter FR is set so that the index value calculation means 34 performs the inner product of the surface normal vector of the measurement point T that coincides with the calculation point IP located symmetrically with respect to the reference point M1. Is preferred. In this way, the reference point M1 is positioned at the maximum deformed portion Z by instructing the combination of the measurement points T on which the surface normal vector is inner producted by the calculation point IP located symmetrically with respect to the reference point M1. In this case, it is possible to increase the angle difference between the surface normal vectors that are inner productd to make the calculated index value remarkable.

  Here, the index value setting filter FR has an odd number of pixels (in the vertical and horizontal directions) with respect to an arbitrary measurement point T (pixel) arranged in a two-dimensional array in the measurement region L (surface normal vector map MP1). For example, the reference point M1 and the calculation point IP of the index value setting filter FR are set on the surface normal vector map MP1 by setting the ring shape having the calculation point IP of 3 × 3 pixels, 5 × 5 pixels, or the like. It can be made to coincide with the position of the measurement point T. That is, in the index value setting filter FR in which the calculation points IP for odd-numbered pixels are set in the vertical and horizontal directions, the index value calculation means 34 adjusts the position of each measurement point T included in the measurement region L to the reference point M1. A value can be calculated. In this case, as shown in FIGS. 4 (a) and 4 (b), the coordinates of the index value map MP2 generated by the index value calculation means 34 are used as the measurement points T in the measurement region L (surface normal vector map MP1). The coordinates of the measurement region L can be easily specified based on the index value map MP2. On the other hand, with respect to an arbitrary measurement point T (pixel) arranged in a two-dimensional array in the measurement region L (surface normal vector map MP1), even pixels (for example, 2 × 2 pixels, 4 When the calculation point IP is set to coincide with the measurement point T on the surface normal vector map MP1, the measurement in the surface normal vector map MP1 is performed. The reference point M1 can be positioned between the points T. That is, in the index value setting filter FR in which the calculation points IP for even pixels are set in the vertical and horizontal directions, the index values of the coordinates located between the measurement points T in the measurement region L can be calculated.

(Coordinate specifying means)
The coordinate specifying unit 36 determines whether or not the index value calculated by the index value calculating unit 34 is outside a predetermined reference range, and coordinates of the measurement region L corresponding to the index value outside the reference range. By specifying, the position of the concavo-convex portion 104 in the measurement region L is specified. Here, when the measurement point T for the inner product of the surface normal vectors is in an ideal plane region without distortion, the index value calculated based on the equation (5) is “1”, while When the measurement point T is on the uneven portion 104, the index value calculated based on the equation (5) is a value lower than “1”. Therefore, the reference value is “1”, and the index value of each coordinate in the measurement region L calculated by the index value calculation means 34 (the index value of each coordinate in the index value map MP2) is lower than the reference value (reference range). The coordinate specifying means 36 specifies the coordinates corresponding to the index value outside the reference range, and the position (coordinates) of the concavo-convex portion 104 in the measurement region L is detected. It is possible.

  Then, when the coordinate specifying unit 36 specifies the coordinates of the measurement region L corresponding to the index value out of the reference range, the control device 30 sets the coordinate position in the image displayed on the display device 50. An instruction mark is displayed. Thus, by displaying the indication mark indicating the uneven position on the image displayed on the display means 52, the inspection operator can visually grasp the position of the uneven portion 104 in the measurement region L of the inspection object 100. It has come to be able to do. Here, the control device 30 includes a display control unit 44 that controls an image to be displayed on the display device 50 based on the data stored in the data storage unit 42, and controls the display control unit 44. Based on the surface normal vector distribution image (see FIG. 8) color-coded according to the direction of the surface normal vector of the surface normal vector map MP1, or according to the index value of the index value map MP2 The index value distribution image (see FIG. 9) can be displayed, and the surface normal vector distribution image and the indication mark indicating the uneven position can be displayed on the index value distribution image.

  Here, the distribution of the index values in the measurement area L calculated by the index value calculation means 34 is apparent from FIG. 6 in which the vertical axis is the index value value and the horizontal axis is the coordinate position of the measurement area L. As described above, in reality, the inspected object 100 is not necessarily flat even if it is a planar region, or is slightly inclined, or the surface normal vector calculating unit 32 calculates based on image data captured by the camera 20. Since the index value obtained by inner product of the surface normal vectors of the measurement points T in the plane region is not always “1” with high accuracy due to the error of the surface normal vector and the like. When the value is “1”, coordinates other than the uneven portion 104 are detected as the uneven portion 104, and noise increases. For this reason, it is possible to accurately detect the uneven portion 104 by setting a reference value that can sufficiently detect the uneven portion 104 and does not detect the noise portion with respect to the index value calculated by the index value calculating unit 34. Is possible. Specifically, it is preferable to set a value smaller than the minimum value of the index value obtained by inner product of the surface normal vector of the measurement point T in the plane area as the reference value. Here, since the surface state of the object 100 to be inspected changes depending on the material of the object 100 to be detected and the like, a reference value corresponding to the object 100 to be detected is set in advance. Detection is performed by creating an index value map MP2 as described above using a reference object having no irregularities 104 in the measurement region L, and obtaining an index value that is the minimum value in the index value map MP2 in advance. A reference value suitable for the object 100 to be inspected can be set. By making this reference value close to “1”, a slight change in the index value due to the concavo-convex portion 104 can be identified by the coordinate specifying means 36, and by making the reference value a value away from “1”, The coordinate specifying means 36 can identify a change in the index value due to the uneven portion 104 having a certain depth (height).

(Effective index value judgment means)
Further, as shown in FIG. 1, the control device 30 includes an index value (referred to as an effective index value) that is compared with a reference value by the coordinate specifying unit 36 among the index values calculated by the index value calculating unit 34. The effective index value determining means 38 for determining Here, in the concavo-convex portion 104, not only the index value of the coordinate corresponding to the maximum deformed portion Z is a value deviating from the reference value, but also the index value of the coordinates located around the maximum deformed portion Z is also from the reference value. It is out of value. For this reason, when the coordinate specifying means 36 determines the index values of all the coordinates included in the index value map MP2 by comparing with the reference value, the number of coordinates that become an index value deviating from the reference value increases, and the coordinate specifying means This is a factor that increases the coordinates detected as the uneven position by 36. Therefore, the effective index value determination unit 38 narrows down the effective index values that are compared with the reference value by the coordinate specifying unit 36 among the index values included in the index value map MP2, thereby accurately detecting the uneven portion 104. However, the coordinates detected as the concave / convex position for one concave / convex portion 104 are suppressed.

  As shown in FIGS. 4 (c) and 5, the effective index value determination unit 38 determines the index value when the index value at an arbitrary coordinate calculated by the index value calculation unit 34 is used as a reference index value. The map MP2 is configured to determine whether or not the reference index value is a minimum value among the index values in the comparison region of the predetermined pixel range including the reference index value. When the reference index value is the minimum value, It is set to execute a non-minimum point suppression process using the reference index value as an effective index value. Specifically, the control device 30 includes a reference point M2 indicating a position to be used as a reference index value by the effective index value determining unit 38, and a position (comparison region) for comparing the reference index value and the index value of the reference point M2. A comparison area filter HF that defines a comparison point H indicating the above is set (see FIGS. 4 and 5). In other words, the effective index value determination unit 38 matches an arbitrary coordinate in the index value map MP2 with the reference point M2, and thereby an index value (reference index value) positioned at the reference point M2 and an index value positioned at the comparison point H. Are compared to determine whether or not an effective index value is used. Here, in order to specify the center of the concavo-convex portion 104, it is preferable to determine the comparison point H so as to surround the reference point M2 in the comparison region filter HF.

  Further, the effective index value determining means 38 sequentially changes the coordinates for matching the reference point M2 of the comparison area filter in the measurement area L (index value map MP2), and the index value that matches the reference point M2 for each change is obtained. It is set to compare with the index value in the comparison area. In other words, the effective index value determining unit 38 is configured to sequentially determine all the index values calculated by the index value calculating unit 34 included in the index value map MP2 as reference index values, and the coordinates of the measurement region L The data storage means 42 is configured to store an effective index value map MP3 (see FIG. 5) for specifying an effective index value. That is, the coordinate specifying means 36 accurately detects the position of the concavo-convex portion 104 by specifying the coordinates of the measurement region L corresponding to the index value out of the reference range based on the effective index value map MP3. Can do. Each coordinate of the effective index value map MP3 is set to coincide with the coordinates of the index value map MP2. That is, the comparison area is a detection unit of the uneven portion 104, and the coordinates detected as the uneven portion 104 can be reduced by widening the setting range of the comparison area, and conversely, the setting of the comparison area is set. Coordinates detected as the uneven portion 104 can be increased by setting the range narrow. Here, it is preferable to set the comparison area of the comparison area filter HF to be smaller than the index value setting filter FR in which the calculation points IP are arranged in a ring shape. In FIG. 4C, each coordinate in the effective index value map MP3 is represented by a rectangular cell.

  Here, referring to FIG. 5, the comparison area filter HF in which the reference point M2 is set at the center is set for the data in the range of the coordinates (RA1 to RA5, LA1 to LA5) of the index value map MP2 (row, column). = (3 × 3 pixels) where (the number of rows of coordinates of the reference point M2−1 to the number of rows of coordinates of the reference point M2 + 1, the number of columns of coordinates of the reference point M2−1 to the number of rows of coordinates of the reference point M2 + 1) A state where the effective index value determination means 38 generates data in the range of the coordinates (RB1 to RB5, LB1 to LB5) of the effective index value map MP3 when the area is set as the comparison point H (comparison area) will be described. . When the data of the coordinates (RA2, LA2) of the index value map MP2 is set as the reference index value (0.8), the coordinate range of (RA1 to RA3, LA1 to LA3) in the index value map MP2 is used as a comparison area. The effective index value determining means 38 determines whether or not the reference index value is a minimum value in the comparison region. In this case, since the index value (0.7) lower than the reference index value (0.8) is included in the coordinates (RA2, LA3) of the index value map MP2, the reference index value is used as the effective index. The effective index value determining means 38 determines that the value is not to be a value, and the corresponding coordinates (RB2, LB2) of the effective index value map MP3 indicate a value indicating a non-concave portion (that is, a value larger than the reference value. In this example, “ 1 ”) is set.

  Further, when the data of the coordinates (RA2, LA3) of the index value map MP2 is set as the reference index value (0.7), the coordinate range of (RA1 to RA3, LA2 to LA4) in the index value map MP2 is compared with the comparison area. The effective index value determining means 38 determines whether or not the reference index value is a minimum value in the comparison area. In this case, since there is no value smaller than the reference index value (0.7) in this comparison area, the effective index value determining means 38 determines that the reference index value is to be the effective index value, and the effective index A reference index value is set as an effective index value at the corresponding coordinates (RB2, LB3) in the value map MP3. Similarly, the effective index value determination unit 38 determines whether or not the reference index value is the minimum value in the comparison region using the data of each coordinate of the index value map MP2 as the reference index value. Either an effective index value or a value indicating a non-concave portion is set to each coordinate of the effective index value map MP3 corresponding to each coordinate. In the example of FIG. 5, when the data of the coordinates (RA4, LA4) in addition to the coordinates (RA2, LA3) is set as the reference index value (0.6) in the index value map MP2, the corresponding comparison areas (RA3 to RA5) , LA3 to LA5), the reference index value is a local minimum value. Therefore, the reference index value is an effective index for each of the corresponding coordinates (RA2, LA3) and coordinates (RB4, LB4) in the effective index value map MP3. A value indicating a non-recessed portion is set in other coordinates.

(Operation of Example)
Next, the operation of the above-described surface inspection apparatus and surface inspection method will be described below. Each of the light irradiation units 12, 14, and 16 independently irradiates the inspection object 100 with light, images the inspection object 100 with the camera 20, and applies light to the same measurement region L of the inspection object 100. A plurality of (three or more types) image data having different directions are acquired. Then, as shown in FIG. 10, for each pixel (measurement point T) at the position corresponding to the measurement region L in each acquired image data, the surface normal vector of each measurement point T is obtained from the information on the brightness and the light source direction. Calculate (Step S1), create a surface normal vector map MP1 in which each measurement point T of the measurement region L and the surface normal vector correspond to each other, and measure the inner surface of the calculated surface normal vector. An index value indicating the degree of unevenness of each coordinate in the region L is calculated. In this case, a surface normal vector corresponding to each measurement point T in the measurement region L is calculated based on image data captured in a state where light is irradiated on the entire measurement region L of the inspection object 100. Therefore, the processing speed for calculating the concave / convex position is increased and the concave / convex position existing on the object to be inspected can be identified in a short time compared to the configuration in which the measurement area L is irradiated and scanned with the slit light. It becomes possible.

  Here, when calculating the index value, the index value setting filter FR defining the calculation point IP indicating the position of the measurement point T where the surface normal vector is inner product and the reference point M1 corresponding to the calculation point IP, In the measurement region L, the calculation point IP is moved so as to match the interval of the measurement points T (that is, one pixel interval) and applied to the surface normal vector map MP1, and the filter FR is positioned at the calculation point IP at each application position. The index value corresponding to the reference point M1 is calculated by inner product of the surface normal vectors of the measurement point T (step S2). Since the surface normal vector map MP1 indicates the value of the surface normal vector at each measurement point T in the measurement region L, the surface method is used so that the calculation point IP determined in the index value setting filter FR is aligned with the measurement point T. By applying to the line vector map MP1, the coordinates of the measurement region L where the reference point M1 is located at this point are specified. That is, by applying the index value setting filter FR so as to scan the measurement area L based on the surface normal vector map MP1, the measurement area L where the reference point M1 is located at each application position of the index value setting filter FR. It is possible to create an index value map MP2 in which the coordinates of and the index value corresponding to the coordinate position are associated with each other.

  Since the index value is a value obtained by inner product of the surface normal vectors, the index value calculated by inner product of the surface normal vectors of the measurement points T in the plane area is a value close to “1”. On the other hand, the index value calculated by inner product of the surface normal vectors of the measurement points T in the concavo-convex portion 104 is a value lower than “1”. That is, an index value indicating the degree of unevenness of coordinates corresponding to the reference point M1 is calculated, and a coordinate corresponding to an index value (value lower than the reference value) outside the reference range in the measurement region L is specified (step S4) The position of the concavo-convex portion 104 can be detected with high accuracy in the measurement region L, and the surface state of the object 100 can be inspected.

  Further, when calculating the index value (when creating the index value map MP2), an index value setting filter FR in which a plurality of calculation points IP are arranged in a circle centering on the reference point M1 is applied, and the reference point The value of the index value calculated by the inner product of the surface normal vectors of the measurement points T in the plane region by inner product of the surface normal vectors of the measurement points T that coincide with the calculation point IP located across M1 The difference from the index value calculated by the inner product of the surface normal vectors of the measurement points T in the concavo-convex portion 104 can be made significant. That is, the uneven portion 104 formed on the inspected object 100 has a surface where the maximum deformation portion Z is sandwiched as shown in FIGS. 2B and 2C because the maximum deformation portion Z becomes an inflection point. The portion faces the opposite side, and the angle difference between the surface normal vectors at the surface portion increases. Therefore, in a state where the reference point M1 of the index value setting filter FR coincides with the maximum deformed portion Z of the uneven portion 104, the measurement point T located across the reference point M1 (maximum deformed portion Z) is used as the calculation point IP. The index value calculated by inner product of the surface normal vector is smaller than the index value calculated when the reference point M1 is located at a position other than the maximum deformed portion Z of the uneven portion 104 (minimum value). It can be. Accordingly, it is possible to accurately detect the coordinates of the uneven position corresponding to the index value (value lower than the reference value) outside the reference range in the measurement region L.

  By the way, the actual inspected object 100 rarely changes the surface state uniformly, and completely eliminates the error of the surface normal vector calculated based on the image data captured by the camera 20. It is difficult. For this reason, for example, when calculating the index value by inner product of the surface normal vectors of one set of measurement points T, the reference point M1 is positioned by the surface normal vectors of only two measurement points T in the measurement region L. The index value of the coordinate to be determined is determined, and the surface state may not be accurately reflected in the index value. Therefore, a plurality of calculation points IP indicating the positions of a pair of measurement points T at which the inner product of the surface normal vectors is set in one index value setting filter FR, and the filter FR is applied in the surface normal vector map MP1. At the same time, the index value is calculated by calculating the average value of the inner product values of the inner products of the surface normal vectors for each measurement point T corresponding to the calculation point IP of each group as the index value corresponding to the reference point M1. Thus, the surface state (surface normal vector) around the reference point M1 can be appropriately reflected in the index value, and the position of the concavo-convex portion 104 can be detected with higher accuracy using the index value. It becomes.

  In particular, a plurality of calculation points IP are arranged in an annular shape centering on the reference point M1, and inner products of surface normal vectors of a plurality of sets of measurement points T that coincide with the calculation point IP located across the reference point M1. By using the average value of the inner product values as the index value, the surface state of the region surrounded by the plurality of calculation points IP (the region surrounded by the index value setting filter FR in the measurement region L) is changed to the index value at the center position. It can be.

  In addition, as shown in step 3 of FIG. 10, the surface normal vector is inner productd as a pre-stage for specifying the coordinates corresponding to the index value (value lower than the reference value) out of the reference range in the measurement region L. Each index value calculated in step (each index value of each coordinate of the index value map MP2) is sequentially used as a reference index value, and it is determined whether or not the reference index value is a minimum value among the index values in the comparison area. By specifying a coordinate corresponding to an effective index value (a value lower than the reference value) that is out of the reference range in the measurement region L, using the minimum reference index value as an effective index value, one uneven portion 104 is set as an uneven position. The specified coordinates can be suppressed. Further, as described above, the index value calculated in a state where the reference point M1 is in the maximum deformed portion Z of the concavo-convex portion 104 (that is, the center of the concavo-convex portion 104) is a minimum value, and thus becomes a minimum value in the comparison region. By selecting the index value to be an effective index value, the center of the uneven portion 104 can be detected. Then, by matching the size of the concavo-convex portion 104 with the comparison area as a detection target, one effective index value can be included for one concavo-convex portion 104, and specific coordinates are set according to the concavo-convex portion 104. Can be detected. In addition, when the comparison region is set to be smaller than the uneven portion 104 to be detected, a plurality of effective index values are included for one uneven portion 104, so that a plurality of coordinates can be detected as the uneven portion position. The concavo-convex portion 104 of the device under test 100 can be detected.

(Experimental example)
Next, a surface inspection experiment of the inspected object 100 was performed using the surface inspection apparatus 10 and the surface inspection method according to the present invention. In this experimental example, each light irradiation unit 12, 14, 16 is disposed at a position 5 m away from the inspection object 100 disposed at a fixed position of 90 cm from the floor surface, and the camera 20 is disposed 90 cm from the floor surface. At a height position of 1 m from the object 100 to be inspected. As the camera 20, an 18-bit high-sensitivity camera (model 18G-01C-S) using a CMOS image sensor manufactured by ViewPLUS was used. The resolution of the camera 20 is (horizontal) 1280 pixels × (vertical) 1024 pixels. Further, as the object to be inspected 100, the same as the five types of samples in which a recess having a depth shown in FIGS. 7A to 7E is formed at the center of a (vertical) 50 mm × (horizontal) 50 mm metal plate. Five types of samples in which convexities having depths shown in (f) to (j) of Table 1 were formed were prepared.

  Then, the surface normal vector is calculated with the entire surface of each sample as the measurement region L, and the index value map of each coordinate in the measurement region L is calculated using the index value setting filter FR having a diameter of 9 pixels. MP2 is created, and an effective index value is calculated using a 5 × 5 pixel area as a comparison area, and a value smaller than the reference value (0.99998) among the effective index values is used as the coordinates of the uneven position. It was. FIG. 8 is an image of each sample displayed in different colors according to the direction of the calculated surface normal vector. FIG. 9 is specified as the coordinates of the uneven position in the index value map MP2 corresponding to the measurement region L. It is surrounded by a circle (instruction mark) centered on the position. 8 and 9 indicates the center of the uneven position. That is, it can be seen that an uneven portion formed at the center is detected in each sample. Moreover, it turns out that the very slight unevenness | corrugation of 0.06 mm is detectable like a sample (e). Note that in the samples (a), (b), (f), and (d) in which relatively large irregularities are formed, a plurality of coordinates are specified for one irregularity, all of which are included in the irregularities. You can see that That is, the actual indentation state can be easily confirmed by the inspector confirming the object to be inspected according to the detected coordinate position.

(Change example)
(1) In the embodiment, the LED is used as the light source of the light irradiation unit, but is not necessarily limited to the LED. That is, other light sources such as a halogen lamp, a dye laser, a light emitting diode (LD), and electroluminescence (EL) can be employed as the light source of the light irradiation unit. However, LEDs are particularly preferably used from the viewpoint of cost and heat generation.
(2) In the embodiment, the object to be inspected is irradiated with light by three light irradiating units. However, the object to be inspected is provided with four or more light irradiating units and irradiated with light by the light irradiating unit. The image data of the surface of the object to be inspected corresponding to each light irradiation unit can be acquired.
(3) In the embodiment, the plurality of light irradiation units are configured to sequentially irradiate the object to be inspected and acquire image data corresponding to each light irradiation unit. However, the present invention is not limited to this. In other words, any configuration may be used as long as at least three types of image data having different light source directions can be acquired so that the surface normal vector can be calculated by the illuminance difference stereo method. Specifically, a light source that emits light having different wavelengths may be employed for a plurality of light irradiation units, and an imaging unit such as a 3CCD camera that can acquire spectral image data by separating each color may be employed.
(4) In the embodiment, as a circular index value setting filter in which a calculation point indicating the position of a measurement point to which the surface normal vector is inner product and a reference point corresponding to an index value calculated based on the calculation point is determined, An annular index value setting filter in which calculation points are arranged in a circle or a rectangular ring arranged in a square shape is shown. However, the present invention is not limited to this, and the calculation points may be arranged in a triangular or other polygonal shape. Alternatively, the calculation points may be arranged in an elliptical shape. In addition, when adopting the index value setting filter in which the calculation points are arranged in an annular shape, it is preferable to adopt an annular shape that matches the contour shape of the uneven portion to be inspected, and the contour of the uneven portion to be inspected When the shape is unknown, it is preferable to use an annular one.
(5) Further, the index value setting filter may define at least two calculation points indicating the positions of measurement points at which the surface normal vector is inner product and a reference point corresponding to the calculated index value. That's fine.
(6) The object to be inspected is not limited to metal parts, but can be made of various synthetic resin parts and films made of polypropylene or polyethylene, and papers can also be inspected. Is possible.
(7) In the embodiment, the surface normal vector of the measurement point in the measurement region of the object to be inspected is calculated using the illuminance difference stereo method, but the present invention is not limited to this, and the surface normal vector is calculated. It is possible to adopt various conventionally known methods capable of

12, 14, 16 Light irradiation unit, 20 camera (imaging means), 32 plane normal vector calculation means 34 index value calculation means, 36 coordinate identification means, 38 effective index value determination means,
100 test object, FR index value setting filter, IP calculation point, M1 reference point L measurement area, T measurement point

Claims (8)

A surface inspection device for inspecting unevenness of an object to be inspected,
A light irradiation unit for irradiating light to a measurement region of the object to be inspected;
Imaging means for capturing an image of the inspected object in a state where light is irradiated by the light irradiation unit and acquiring image data of a measurement region;
Surface normal vector calculation means for calculating a surface normal vector of a measurement point in the measurement region based on image data captured by the imaging means;
Index value calculating means for calculating an index value by inner producting the surface normal vectors calculated by the surface normal vector calculating means;
Coordinate specifying means for specifying an uneven position in the measurement region based on the index value calculated by the index value calculating means,
An index value setting filter that sets a calculation point indicating a position where an inner product of the surface normal vector and an index value calculated based on the calculation point are set is set,
By sequentially changing the measurement points that match the calculation points of the index value setting filter at the measurement points in the measurement area, the corresponding reference points are obtained by inner producting the surface normal vectors of the decision points that match the calculation points for each change. By calculating the index value by the index value calculation means, each time the index value is calculated, the coordinates of the measurement region where the reference point is located and the index value of the coordinates are specified.
A surface inspection apparatus configured to identify the coordinates of the measurement region corresponding to an index value out of a reference range among index values calculated by the index value calculation unit. . In the index value setting filter, the calculation points are arranged in a ring centered on the reference point,
The index value calculation means calculates the index value of the reference point by inner product of the surface normal vectors of the measurement points that coincide with the calculation points located so as to sandwich the reference point in the index value setting filter. The surface inspection apparatus according to claim 1. In the index value setting filter, a plurality of sets of calculation points are arranged in a ring centered on the reference point,
The index value calculation means is configured to calculate an index value of a reference point by calculating an average value of inner product values obtained by inner producting surface normal vectors for each measurement point located at each set of calculation points. Item 1. The surface inspection apparatus according to Item 1. An arbitrary index value calculated by the index value calculation means is used as a reference index value, and it is determined whether or not the reference index value is a minimum value among index values within a predetermined comparison area including the reference index value. And an effective index value setting means for setting the reference index value as an effective index value in the case of a minimum value,
The effective index value setting means is configured to determine whether each index value calculated by the index value calculation means is an effective index value as a reference index value in order,
2. The coordinate specifying unit is configured to specify coordinates of the measurement region corresponding to an index value out of a reference range among index values determined to be effective index values by the effective index value setting unit. The surface inspection apparatus as described in any one of -3. A surface inspection method for inspecting unevenness of an object to be inspected,
Calculate the surface normal vector of the measurement points in the measurement area of the object to be inspected based on the image data captured by the imaging means,
A measurement point that matches the calculation point by matching the calculation point of the index value setting filter that defines the calculation point indicating the inner product of the surface normal vector and the reference point corresponding to the calculation point to the measurement point in the measurement area In addition to calculating the index value corresponding to the reference point by inner product of the surface normal vector of
Each time the index value is calculated, the index value corresponding to each coordinate of the measurement area is calculated by changing the measurement point that matches the calculation point of the index value setting filter and calculating the index value corresponding to the reference point for each change. Identify
A surface inspection method for detecting an uneven position of an object to be detected by specifying coordinates corresponding to an index value deviating from a reference range in the measurement region.   Based on an index value setting filter in which the calculation points are arranged in a ring centered on the reference point, a reference is obtained by inner producting the surface normal vectors of the measurement points that coincide with the calculation points located so as to sandwich the reference point. The surface inspection method according to claim 5, wherein an index value corresponding to the point is calculated.   Based on an index value setting filter in which a plurality of sets of calculation points are arranged in a ring centered on the reference point, an average of inner product values obtained by inner product of surface normal vectors for each measurement point that matches the calculation points of each group The surface inspection method according to claim 5, wherein the value is calculated as an index value corresponding to the reference point.   When each index value is set as a reference index value in order, a reference index value that is a minimum value within a predetermined comparison area including the reference index value is set as an effective index value, and the effective index value is out of the reference range in the measurement area. The surface inspection method as described in any one of Claims 5-7 which detects the uneven | corrugated position of a to-be-detected body by specifying a corresponding coordinate.