JP4018091B2 - Apparatus and program for determining surface condition - Google Patents

Description

  The present invention relates to a technique for determining the surface state of an inspection object by image analysis.

  Inspection of scratches and other surface conditions on the surface of gears and other processed products is often performed by visual inspection. However, while the operation is simple, it requires concentration and is difficult. Therefore, there is a need for an apparatus that can inspect gears and other processed products without depending on visual observation.

  Japanese Patent Application Laid-Open No. 2004-228561 describes a method of determining the presence or absence of scratches such as dents on the gear tooth surface by photographing the gear tooth surface and evaluating the luminance of the obtained image. A gear tooth surface to be inspected and a gear tooth surface different from the tooth surface are photographed at a predetermined interval for one gear, and a plurality of images from substantially the same direction regarding the different tooth surfaces are created. A difference image representing a difference in luminance is generated. The difference image is added to create an added image, and the presence or absence of a dent is determined by comparing the luminance of the added image with a reference value.

Japanese Patent Application Laid-Open No. 2004-228561 creates a self-organizing map from an image of an object using a neural network, forms a cluster of neurons corresponding to the same image, and authenticates the image of the object using this cluster Is described.
Patent Application Publication Number Sho 63-201556 Patent Application Publication No. 2002-109541

  In the technique described in Patent Document 1, a region (hereinafter referred to as overdetection or overdetection region) that is different from the brightness of other image regions but is not a scratch such as uneven color or dirt on the surface of a gear or other processed product. Many of them are detected, and considerable visual work is required to detect scratches and other surface conditions.

  In the technique described in Patent Document 2, since the inspection object is recognized using one self-organizing map, there is a possibility that sufficient accuracy may not be obtained.

  Depending on the part of the inspection object, different features that can be easily identified may appear in the image of the object. For example, an image obtained by photographing the end portion of the inspection object is an image obtained by photographing a central portion of the object, while a part (for example, a background) different from the object is photographed. The object may be photographed over the entire image area. The image features clearly differ between an image including a background and an image not including a background regardless of the presence or absence of a scratch or other surface condition. If the same self-organizing map is applied to such images having different features that can be easily discriminated, the accuracy of determining the surface state of the inspection object may be reduced.

  In order to improve the accuracy of determining the surface state of the inspection object, it is conceivable to increase the number of neurons in the self-organizing map. However, an increase in the number of neurons causes an increase in the time required for the determination process.

  The present invention provides a technique that can more efficiently determine the surface state while reducing the visual determination effort.

  In one form (Claim 1), the surface state determination apparatus of the present invention includes, for each of a plurality of predetermined portions of the inspection object, a storage unit that stores a data map obtained by learning a predetermined surface state in the portion. The inspection area is set so as to include a determination area having a luminance difference exceeding a predetermined threshold with respect to other areas in the imaging means for imaging the inspection object and the image of the inspection object acquired from the imaging means. Means for extracting a feature vector from the examination area, means for judging a part to which the examination area belongs among a plurality of parts of the examination object, and selecting a data map corresponding to the part; Determination means for inputting a feature vector into the selected data map and determining a predetermined surface state in the inspection region.

  According to the present invention, for each part of the object, the surface state of the part is determined using a data map obtained by learning a predetermined surface state such as scratches, dents, and peeling, so that the time required for the determination is not increased. The determination for each part can be performed efficiently.

  In an embodiment of the present invention (Claim 2), the storage means stores a data map in which a predetermined surface state of a region where the inspection object is imaged over the entire inspection area and a part of the inspection area where the inspection object is present. A data map in which a predetermined surface state of a part to be imaged is learned is stored. The characteristics of the inspection region differ depending on whether the inspection object is imaged over the whole or not, regardless of the presence or absence of a predetermined surface state. By using a data map adapted to each of the inspection regions having different characteristics as described above, the capacity of each data map can be suppressed. Since the capacity of the data map is suppressed, an increase in time required for determination can be avoided.

  In an embodiment of the present invention (Claim 3), the inspection object is a gear, and the storage means stores a data map in which a predetermined surface state is learned in a central portion of an area where the gear is photographed in the image, and an image. And a data map in which a predetermined surface state is learned at the end portion of the region where the gear is photographed. An image obtained by imaging the end portion of the gear may be imaged different from the gear, and an image obtained by imaging the central portion of the gear may be imaged over the entire image. By using a data map adapted to each of the surface state of the central portion and the end portion of the photographing region of the gear, the capacity of each data map can be suppressed. Since the capacity of the data map is suppressed, an increase in time required for determination can be avoided.

  In one form of this invention (Claim 4), it has a means to produce | generate a data map about each of several site | part of a test target object. The means for generating the data map further includes: (a) means for defining a self-organizing map for each of a plurality of parts of the inspection object; and (b) a plurality of samples in the inspection region for the inspection object. Preparing, for each of the plurality of sample inspection areas, 1) extracting a feature vector from the inspection area; 2) determining a part to which the inspection area belongs among a plurality of parts of the inspection object; A self-organizing map corresponding to the region is selected, 3) means for inputting the extracted feature vector into the selected self-organizing map and learning, and (c) the same sample in the learned self-organizing map Means for grouping adjacent neurons into clusters, and (d) each of the clusters corresponding to a predetermined surface state and not Are classified into, and means for generating a data map, the.

  According to the present invention, since a data map in which a predetermined surface state is learned is generated for each part of the inspection object, the capacity of each data map is suppressed, and thus determination without increasing processing time is performed. Can be implemented. In particular, with respect to an object in which features that are clearly different for each part are imaged, the determination for each part can be performed efficiently while suppressing the capacity of each data map.

  Next, embodiments of the present invention will be described with reference to the drawings. FIG. 1 shows a configuration of an imaging apparatus when an object to be inspected is a gear 11. The gear 11 is attached to the rotary table 15 and rotated stepwise at a gear tooth pitch by a step motor (not shown) provided on the base 17. The gear 11 is irradiated by the lighting device 19. A camera 21 equipped with a CMOS image sensor photographs the tooth surface of the gear 11. Photographing is performed in synchronization with the stepwise rotation of the gear 11, and a still image is photographed for each tooth surface each time one tooth rotates. In one embodiment, the lighting device 19 uses blue light emitting diodes, but a lamp or other lighting device can also be used. Further, by using a wide dynamic range CMOS camera having a wide measurable luminance range as the camera 21, a good image suitable for scratch determination can be obtained. Depending on the object, a camera using a CCD image sensor can also be used.

  The positioning sensor 18 is generally called a proximity sensor, and generates a high-frequency magnetic field by a detection coil. When a metal detection object approaches this magnetic field, an induced current (eddy current) flows through the detection object due to electromagnetic induction. This current changes the impedance of the detection coil and stops the high-frequency oscillation, thereby detecting the approach of the detection object. In this embodiment, the proximity sensor detects the approach of the gear teeth crest. A signal from the positioning sensor 18 is sent to the computer 23 (FIG. 2), and the timing unit 24 sends a signal to the camera 21 in response to detection of the gear teeth. The camera 21 shoots the gear in response to this signal.

  FIG. 2 shows functional blocks of the computer 23 that processes an image photographed by the camera 21. The computer 23 may be a personal computer, a workstation or other computer. The computer 23 includes a processor (CPU) that executes arithmetic operations, a storage device that stores various data and programs, an input device such as a mouse and a keyboard that allows a user to input data for processing, and the processor. A display device (display) for displaying the processing result and an output device such as a printer for outputting the processing result can be provided. The storage device can include a memory such as a ROM and a RAM and an auxiliary storage device such as a disk device. The processing by each functional block in the figure is executed by the processor.

  The image acquisition unit 25 receives an image captured for each tooth surface pitch from the camera 21 and passes it to the difference image generation unit 27. The difference image generation unit 27 calculates a difference between the image acquired this time and the image acquired last time, that is, the image of the gear tooth surface one pitch before. Since the tooth surfaces of the gear without any scratches or other surface conditions are substantially the same, taking a difference results in an image with a low overall brightness. If the gear tooth surface has scratches such as dents and peeling, the brightness of the scratched part is different from that of the other parts, so that it appears as a part of high brightness in the difference image.

  When the difference between the two images is taken, there is a pixel having a negative luminance value, so that the intermediate luminance 128 is added to the luminance value of the difference to make the luminance values of all the pixels positive. A pixel whose luminance value exceeds 255 by adding 128 has a luminance value of 255, and a pixel that becomes negative even if 128 is added has a luminance value of 0. In this way, the luminance value range of the difference image is adjusted from 0 to 255.

  Referring to FIG. 3A, the inspection area setting unit 28 detects an area having a luminance greater than the threshold value from the difference image as a potential flaw area 31, and sets an area 33 surrounding the area as an inspection area. To do. The shape of the inspection region 33 is a quadrangle in this embodiment, but may be other polygons, or may be a circle or an ellipse. The size is set to a size sufficient to surround the wound area 31 according to the potential wound area 31.

  The inspection area setting unit 28 stores the set position information of the inspection area 33 in the storage device of the computer 23. Referring to FIG. 4, an example of an image 41 obtained by photographing a gear tooth surface is shown. Reference numeral 42 indicates a gear photographed in the image 41. An inspection area 33 is set in the image 41. The tooth width of the gear is W, and the position of the inspection region 33 in the tooth width direction is specified by w ′. Thus, the position information of the inspection region 33 includes the position of the gear in the tooth width direction, and this is stored in the storage device of the computer 23.

  The feature extraction unit 29 extracts the following six parameters.

(1) Area S _{A of} potential scratch region 31
The inspection area is displayed in coordinates with i as the vertical axis and j as the horizontal axis, and the area S _A is obtained by integration calculation using the coordinates of the boundary line of the scratch area 31.

(2) Long axis inclination θ _A with respect to the tooth streak when the wound region 31 is approximated to an ellipse
A dent often occurring in a gear can be approximated to an ellipse. In FIG. 3B, an approximate equation of an ellipse that approximates the flaw region 31 is obtained using a graphic approximation program. The approximate expression can be obtained using a known program. Typically, an approximate expression is obtained using a multivariate nonlinear least squares fitting program. An angle θ _A formed by the long axis 37 of the ellipse thus obtained and a straight line 35 parallel to the gear tooth stripe 13 (FIG. 1) is obtained.

(3) The luminance value entropy entB of the inspection area 33
The entropy of the luminance value of the inspection area 33 including the potential flaw area 31 is obtained by the following expression. The luminance value entropy represents the information amount of the luminance value distribution, and a large information amount represents that the degree of randomness of the luminance value distribution is large.

  In this embodiment, the randomness of the luminance value distribution from luminance values 0 to 255 is used as the luminance value entropy. In [Equation 1], p is the luminance value, and rel [p] is the frequency of the luminance value p.

  FIG. 5 shows an example of a luminance value histogram. In the case of a gear tooth surface having dents or other scratches, since the scratches appear as pixel areas having various brightness values in the difference image, the brightness value entropy is large. FIG. 5A is a luminance value histogram of an image in which pixels having various luminance values are present randomly. Brightness values are widely distributed around the intermediate brightness 128, and the entropy is large. FIG. 5B is a luminance value histogram of an image with small variations in luminance values, and the entropy is small.

(4) Luminance value anisotropy ansB of inspection area 33
The anisotropy of the luminance value of the inspection area 33 is expressed by the following formula. Anisotropy indicates the degree of symmetry of the luminance value distribution.

  In this expression, k is a luminance value with the lowest appearance frequency. That is, ansB is a value obtained by dividing the luminance value entropy from 0 to the minimum luminance value in the inspection area 33 by 0 to the entropy of all luminance values. When the luminance value distribution is symmetrical on the left and right of the intermediate luminance value 125, ansB = −0.5. The closer ansB is to -0.5, the higher the symmetry and the smaller the anisotropy. The closer ansB is to -0 or -1 from -0.5, the lower the symmetry and the greater the anisotropy.

  Referring to FIG. 6, the luminance value histogram as shown in FIG. 6A has a high luminance symmetry and a small anisotropy because the luminance values are distributed without any bias around the intermediate value of 128. In the luminance value histogram of (B), the luminance value distribution is different between the left side and the right side of the intermediate value 128, the symmetry is low, and the anisotropy is large. Assuming that the luminance value with the lowest appearance frequency is 230 in (B), ansB is close to -1.

(5) Circularity C _A of potential scratch area 31
Circularity C _A wound area 31 is expressed by the following equation.

In this equation, S _A is the area of the wound area 31, r _max is the maximum distance from the center of gravity of the wound area 31 to the end of the region 33.

(6) The average luminance edgB of the edge image in the inspection area 33
If represents the area of the inspection area 33 in S _B, the average luminance edgB edge images of the inspection area 33 is obtained by the following expression.

  Here, i and j indicate the coordinates of the pixel, and P (i, j) indicates the luminance value of the pixel at the coordinates i and j. ΔP (i, j) is the square of the difference between the luminance values of two pixels adjacent to the upper and lower sides (i-axis direction) of P (i, j) and the luminance of the two pixels adjacent to the left and right (j-axis direction) This is the square root of the sum of the difference squared. ΔP (i, j) represents the magnitude of the difference between the luminance values of adjacent pixels. Taking the difference between the luminance values of adjacent pixels corresponds to generating an image in which an edge portion having a large change in luminance value in the image is emphasized, that is, an edge image. edgB is obtained by dividing the sum of the magnitudes of the luminance values of adjacent pixels by the area of the region 33, and indicates the average luminance of the edge image.

  For each of the six parameters obtained in this way, normalization is performed according to the following equation using the maximum and minimum values for many samples in the inspection region.

Normalization = (obtained value-minimum value) / (maximum value-minimum value)
By this normalization, each parameter takes a value in the range of 0 to 1. Thus, the inspection region 33 including the potential flaw region 31 can be represented by a numerical vector of six feature parameters.

FIG. 7 shows the feature vector thus obtained. FIG. 7A shows a feature vector corresponding to a dent. FIG. 7B shows a feature vector of color unevenness (overdetection) of the gear tooth surface. There is no significant difference between the area parameter S _A of the potential flaw region 31, the long axis inclination θ _A when the ellipse is approximated, and the anisotropy ansB, but the luminance value entropy entB, circularity C _A , and edge It can be seen that there is a large difference in the average luminance edgB of the image.

  Returning to FIG. 2, the feature extraction unit 29 extracts the feature vector including the above-described six parameters from the inspection region 33 as described above. In an alternative embodiment, the feature extraction unit 29 extracts a feature vector composed of three parameters: the luminance value entropy of the inspection area, the circularity of the potential flaw area, and the average luminance of the edge image of the inspection area.

  The determination unit 30 acquires the position information of the inspection region 33 set by the inspection region setting unit 28 from the storage device. Since the pedestal 17 (FIG. 1) on which the camera 21 and the gear are arranged is fixed, the position at which the gear is photographed in the acquired image is determined in advance, and thus the position information of the inspection area is determined by the inspection area. Indicates which part of the gear belongs to. The determination unit 30 determines which part of the gear the inspection region 33 belongs to based on the position information, and selects a determination data map corresponding to the part to which the inspection region belongs. The determination unit 30 applies the feature vector extracted by the feature extraction unit 29 to the selected determination data map, and determines the surface condition of the inspection region 33 such as a scratch.

Classification of Inspection Area A method for determining which part of the gear the inspection area 33 belongs to according to one embodiment of the present invention will be described with reference to FIG. A gear is photographed in the acquired image 61, and an image area where the gear is photographed is represented by reference numeral 62. In the acquired image 61, a background is photographed in a portion other than the gear region 62. As shown in FIG. 8B, a central area A (enclosed by a thick black line) and an end area B (enclosed by a thick broken line) are set in the image. The center area A shows the center portion of the gear to be photographed. As indicated by reference numeral 63, when the inspection area 33 is set to the central area A, the gear is photographed over the entire inspection area 33. An end region B indicates an end portion (edge portion) of a gear to be photographed. If a flaw or other surface condition or overdetection is present at the end portion of the gear, the inspection region 33 is set to the end region B as indicated by reference numeral 64. In such an inspection area 33, the gear is photographed only in a part thereof, and the background is photographed in the other part.

  Since the inspection region 33 set in the end region B includes a gear portion and a background portion, and the inspection region 33 set in the central region A includes only the gear portion, whether or not there is a scratch or other surface condition exists. Regardless, these examination areas have distinctly different characteristics. In this embodiment, a first determination data map is prepared for the central region A and stored in the storage device, and a second determination data map is prepared for the end region B and stored in the storage device. If the determination unit 30 determines from the position information of the inspection region 33 that the inspection region 33 belongs to the central region A, the determination unit 30 determines a scratch or other surface state using the first determination data map, and the inspection region If it is determined that 33 belongs to the end region B, the scratch and other surface conditions are determined using the second determination data map.

  In this way, when different features appear in the examination region according to the gear part, the accuracy and efficiency of judging the gear surface state is improved by preparing a judgment data map that matches the feature of each part. be able to.

  Since the position in the image of the area 62 where the gear is photographed is determined in advance, the positions of the central area A and the end area B in the image can also be determined in advance. The width W2 of the end region B is typically determined depending on the width (length in the horizontal direction of the image) W1 of the inspection region 33. Referring to FIG. 8C, the portion where the right edge of the gear is photographed is shown enlarged, and two examples of the inspection region 33 set to surround the potential scratch region 31 are shown. ing. Assuming that the inspection area 33 is set so that the potential scratch area 31 is located substantially in the center, (c-1) is a case where the right end of the set inspection area 33 reaches the gear edge. 2) is a case where the approximate center of the set inspection region 33 is on the edge of the gear.

  When the inspection area 33 is shifted to the right side from the position shown in (c-1), the inspection area 33 includes the background. Therefore, it is preferable that the end region B is set to have at least the width of W1 on the gear side with respect to the edge and the width of (1/2 × W1) on the background side. When the position of the end region B is determined, the remaining portion of the gear being photographed is set as the central region A.

  When the inspection region 33 extends over the central region A and the end region B, for example, the inspection region 33 depends on whether the position of the potential scratch region 31 in the inspection region 33 belongs to the central region A or the end region B. The region (central region A or end region B) to which the symbol belongs may be determined. Alternatively, the region to which the inspection region 33 belongs may be determined according to the ratio of the portion occupied by the central region A and the portion occupied by the end region B in the inspection region 33.

  In this embodiment, two parts (a central area A and an end area B) are defined on the inspection object according to whether or not the background is photographed in the set inspection area. Alternatively, if different features appear due to structural differences, for example, between a certain part of the inspection object and another part, a determination data map is prepared for each of these parts, and the determination is made. It is preferable to do so.

  As described above, as a preferable mode, when different characteristics exist for each part of the inspection object, preparation of a determination data map for each part has been described. However, it should be noted that the present invention includes preparing a determination data map for each part of the inspection object and making a determination for each part even when different features do not appear for each part.

  A method of creating the first and second determination data maps and a method of determining the surface state using these maps will be described later.

Scratches flow of processing for determining other surface conditions Figure 9, for one gear tooth surfaces in the embodiment of the present invention, a flow chart showing a flow of a process for determining the damage or other surface conditions, all of the gears This process is continuously executed until the determination on the tooth surface is completed. The gear 11 is attached to the rotary table of FIG. In step S101, it is determined whether or not all gear tooth surfaces of the gear 11 have been inspected. If incomplete, the process proceeds to step S103. To determine whether all gear tooth surfaces have been inspected, the number of teeth of the gear 11 is set in the counter, and the counter is decremented each time processing for one tooth is completed, and the value of this counter is checked. By doing.

  In step S103, the gear 11 is rotated by one tooth. The positioning sensor 18 detects the gear teeth and sends a detection signal to the computer 23. The timing unit 24 of the computer 23 sends a drive signal to the camera 21 in accordance with a signal from the positioning sensor 18. In response to this, the camera 21 captures the gear 11 in step S105, and sends the image to the image processing personal computer 23. When the number of photographing is 1 (S107), that is, when the first tooth of the gear 11 has just been photographed, image processing using the difference image cannot be performed, so this image is saved and the next tooth is stored. In order to take a picture, the subsequent steps are skipped, the first process is completed, and the process starts again from the start.

  In step S109, as described above, a difference image is generated from two consecutive tooth images, and an examination region 33 surrounding the potential wound region 31 (FIG. 3) is set. In step S111, a feature vector is extracted from the difference image as described with reference to FIG. In step S113, a routine for selecting a determination data map is executed. In step S115, scratches and other surface conditions are determined using the selected determination data map.

  FIG. 10 shows a flowchart of the routine executed in step S113 of FIG. In step S121, the position information of the inspection area 33 is acquired. In step S122, based on the position information of the inspection region 33, it is determined whether the inspection region 33 belongs to the central region A or the end region B. If it belongs to the central area A, the first determination data map is selected in step S123. If it belongs to the end region B, the second determination data map is selected in step S124. Thus, when the inspection area 33 belongs to the central area A, the first determination data map is used to determine scratches and other surface conditions. When the inspection area 33 belongs to the end area B, the second determination data map is used. Use to determine the surface condition of the scratches.

Generation of Determination Data Map Next, a method for generating a determination data map by self-organization learning using a neural network and a method for determining scratches and other surface states using the determination data map will be described.

In this embodiment, 40 × 40 neurons are used to generate a self-organizing map. FIG. 11 shows three neurons N _j−1 , N _j , N _{j + 1} . ω _j = (ω _j0 , ω _j1 ,..., ω _jm ) is a coupling coefficient vector of the neuron N _j , and χ = (χ ₀ , χ ₁ ,..., χ _m ) is an input vector is there. The input vector in this embodiment is a feature vector extracted by the feature extraction unit 29 (FIG. 2) for a plurality of sample inspection areas 33.

  The self-organizing map is also called a self-organizing feature map, and is generated based on an unsupervised algorithm. This map automatically learns by extracting hidden features in the input data, and generates a map that selectively reacts to the input data. The self-organizing map was proposed by Kohonen and explained in, for example, “Neuro Fuzzy Genetic Algorithm” published in 1994 by Sangyo Tosho Co., Ltd.

  In this embodiment, the first determination data map is generated based on the first self-organization map, and the second determination data map is generated based on the second self-organization map. The input vector is different between the generation of the first self-organizing map and the generation of the second self-organizing map. If the sample inspection region 33 is set to the central region A, the feature vector of the inspection region 33 is input to the first self-organizing map. If the sample inspection region 33 is set to the end region B, the feature vector of the inspection region 33 is input to the second self-organizing map. However, the first and second self-organizing maps are generated in the same manner according to the steps shown below. Therefore, in the following description, “determination data map” indicates that either the first or second determination data map may be used.

1) Step 1: Network initialization:
The neuron connection vector ω _j = (ω _j0 , ω _j1 ,..., Ω _jm ) is determined for all neurons using random numbers.

2) Step 2: Input vector input:
An input vector χ = (χ ₀ , χ ₁ ,..., Χ _m ) is given to each neuron. The relationship between each neuron and the input vector is as shown in FIG.

3) Step 3: Calculate distance between neuron coupling coefficient vector and input vector:
The distance between the connection vector of each neuron and the input vector is calculated by the following equation.

4) Step 4: Determine winner neuron:
The neuron with the smallest distance d _j is selected. This neuron is called the winner neuron and is represented by j *.

5) Step 5: Learning coupling coefficient vectors:
The coupling coefficients (weights) of the winner neuron and nearby neurons are updated according to the following formula.

η is a positive constant and is 0.05 in this embodiment. h (j, j *) is called a neighborhood function and is represented by the following equation.

  σ (t) decreases as the learning progresses. Accordingly, the range of the neighborhood function is wide at the initial stage of learning, as indicated by a circle in FIG. 12, and becomes narrower as the learning progresses. In other words, the coarse adjustment changes to the fine adjustment as the learning progresses. Thus, the neighborhood function effectively generates a map. FIG. 12 shows the winner neuron as a small circle 91, and the neighborhood function surrounding it is represented as a large circle 93.

6) Step 6: t update, return to input vector input:
The learning count t is updated to t + 1, the input vector input in step 2 is returned, and steps 2 to 6 are repeated to repeatedly update the coupling coefficient vector.

  In the self-organizing map, the winner neuron and nearby neurons all approach the current input vector. At the beginning of learning, many neurons are considered to be close by the neighborhood function, and a rough map is formed. As the learning progresses, the number of neurons considered to be close to the winner neuron by the neighborhood function decreases. For this reason, local fine adjustment progresses and spatial resolution increases.

  Features obtained by learning by selecting all the feature vectors of the examination area 33 including scratches and other surface states or over-detection obtained for a plurality of samples in order or randomly and inputting them into a self-organizing map A self-organizing map reflecting the similarity of vectors can be generated. That is, a cluster having similar characteristics gathers on the self-organizing map.

  FIG. 13 shows the self-organizing map obtained in this way. In the embodiment, 40 × 40 neurons are used, but in FIG. 13, a 20 × 20 map is shown for simplicity. In FIG. 13, each grid corresponds to one neuron. Blocks (1), (2),... (28) separated by solid lines represent clusters determined as follows.

  An image of the examination region 33 having a feature vector having the smallest distance from the coupling coefficient vector is arranged at each neuron position of the self-organizing map. The examination area 33 is one of a plurality of samples used for generating the self-organizing map. Adjacent neurons with the same image are selected and grouped. A group of neurons formed in this way is called a cluster. The grouping here is the same as the grouping function used in the existing drawing program, and the grouped neurons can be selected as a whole group and the properties can be set.

  Next, the images arranged in the respective clusters are visually confirmed, and it is determined whether the inspection region 33 includes a surface condition such as a scratch or an overdetection such as surface unevenness. This determination result is set as a cluster property. Specifically, with “Overdetection” entered in the property fields of all clusters as the initial state, a self-organizing map is displayed on the computer display, and the clusters determined to be scratched are displayed with the mouse. Right-click and change and enter “Scratch”. Since neurons can be classified in cluster units in this way, working time can be shortened.

  FIG. 14 shows a map (determination data map) in which clusters are classified in this way. In the figure, hatched clusters (1), (2), (6), (10), (11), (12), (13), (15), (16), (22), (27 ), (28) are clusters classified as “scratches”, and clusters without hatching are “overdetected” clusters. This determination data map is registered in the storage device of the computer 23 (FIG. 2), and this is used by the determination unit 30 for determination of scratches.

In this embodiment, a first determination data map is generated based on the first self-organizing map and registered in the storage device of the computer 23, and the second determination data map is based on the second self-organizing map. Is generated and registered in the storage device of the computer 23. The first determination data map is a map adapted to the central region of the gear, and the second determination data map is a map adapted to the end region of the gear.

A specific method for determining scratches and other surface states performed by the determination / determination unit 30 for scratches and other surface states using the determination data map will be described. For both the first and second determination data maps, the determination is performed according to the same method. The determination unit 30 obtains the distance between the feature vector received from the feature extraction unit 29 and the coupling coefficient vector of the determination data map selected according to the position information of the inspection region 33. This distance is calculated according to [Equation 5] shown above. Neurons are selected in ascending order of the distance, a predetermined number, for example, 10 or less neurons are selected, and a circle or ellipse surrounding them is set as a neighborhood region. As is clear from the generation process of the decision data map, the neurons that approximate the coupling coefficient vector are grouped as a cluster, so if you select a neuron in ascending order of distance as described above, a cluster of adjacent neurons is formed. Selected. A circle or ellipse surrounding this cluster of neurons is defined as a neighborhood region. For example, referring to FIG. 15, a circle 57 defines the neighborhood region.

  A circle 57 obtains the center position of the cluster of these neurons from the array on the determination data map of the selected cluster of neurons, and a circle having a radius from this center to the farthest neuron of the cluster of neurons. Can be obtained as

  Based on the ratio (K) / (S) of the number of neurons (K) belonging to the wound cluster included in the neighboring area defined by the circle or ellipse thus selected and the total number of neurons (S) in the neighboring area Check for scratches. When (K) / (S) ≧ D, it is judged as a scratch, and when (K) / (S) <D, it is judged as overdetection. D is a preset threshold value, for example, 0.5. In this way, the scratch and other surface conditions and overdetection are discriminated. In the example of FIG. 15, since there are seven neurons belonging to the wound cluster (12) in the circle 57 and the total number of neurons in the circle 57 is 12, (K) / (S) = 8/12 = Therefore, it is determined that a flaw is included in the current examination region 33.

Update of Self-Organizing Map The tooth surface of the gear determined to be flawed by automatic determination by a computer using the above-described determination data map is visually inspected to confirm the flaw. Similarly, the tooth surface of the gear determined to be overdetected by automatic determination can be visually inspected and confirmed. Thus, the feature vector of the inspection region of the gear tooth surface finally determined to have a scratch or other surface condition is input to the self-organization map before clustering to be learned. At this time, the initial value of the coupling coefficient vector of the neurons of the self-organizing map uses the latest coupling coefficient vector generated by past learning.

  In this way, the self-organizing map is updated, and the above-described clustering process is executed again. Accordingly, a cluster corresponding to a new type of flaw or overdetection can be generated and a determination data map can be generated without reducing the detection accuracy of existing types of flaws. By using the updated determination data map, the inspection accuracy can be improved. In this update process, since the coupling coefficient vector obtained in the past is used, the determination data map can be updated in a short learning time.

  Although the present invention has been described with respect to specific embodiments, the present invention is not limited to such embodiments.

The block diagram which shows the whole structure of the imaging device of a gear tooth surface according to one Example of this invention. 1 is a functional block diagram showing an overall configuration of an image processing computer according to one embodiment of the present invention. The figure which shows the relationship between a potential flaw area | region and a test | inspection area | region according to one Example of this invention. The figure which shows the positional information on a test | inspection area | region according to one Example of this invention. The figure which shows the luminance value histogram from which entropy differs according to one Example of this invention. The figure which shows the luminance value histogram from which anisotropy differs according to one Example of this invention. The figure which shows the feature vector according to one Example of this invention. The figure which shows the center area | region and end area | region of a gear according to one Example of this invention. The flowchart which shows the flow of the process which determines the damage | wound of a gear tooth surface according to one Example of this invention. The flowchart which shows the flow of the process which selects the determination data map according to one Example of this invention. The figure which shows the partial structure of the neural network according to one Example of this invention. The figure which shows the production | generation process of the self-organization map according to one Example of this invention. The figure which shows the self-organization map divided into the clusters according to one Example of this invention. The figure which shows the determination data map produced | generated from the self-organization map according to one Example of this invention. The figure for demonstrating the damage | wound determination using the determination data map according to one Example of this invention.

Explanation of symbols

11 Gear 19 Lighting device (LED)
21 Camera 23 Computer

Claims (6)

An apparatus for determining scratches on an inspection object,
For the first and second sites predetermined for the object, storing a second judgment data map learned the first determination data map and wound at the site of the second learned scratches in said first region Storage means for
Imaging means for imaging the object;
Means for setting an inspection region to include a determination required region having a luminance difference exceeding a predetermined threshold with respect to another region in the image of the object acquired from the imaging unit;
Means for extracting a feature vector from the inspection region;
If it is determined that the examination region belongs to the first part, the first judgment data map is selected. If it is judged that the examination area belongs to the second part, the second judgment data map is selected. Means for selecting
Determination means for inputting the feature vector into the selected first or second determination data map and determining a flaw in the inspection region; and
The first part is a part in which the object is imaged over the entire examination area and does not include an image of the background of the object, and the second part is a part of the examination area. The object is imaged, and the remaining part includes the background image.
apparatus. The object is a gear;
The first portion is a central portion of the gear , and the second portion is an end portion of the gear .
The apparatus of claim 1. For the first and second sites predetermined of said object, said comprises a first and second judgment data map means for generating respectively a means for the generation further,
(A) means for defining a self-organizing map for each of the first and second parts of the object;
(B) For the object, preparing a plurality of samples in the inspection area, and for each of the inspection areas of the plurality of samples,
(B-1) extracting a feature vector from the inspection region;
(B-2) Of the predetermined first and second parts of the object, determine a part to which the examination region belongs, and select a self-organizing map corresponding to the part;
(b-3) means for inputting and learning the feature vector extracted in (b-1) to the selected self-organizing map;
(C) means for grouping adjacent neurons into clusters that correspond to each of the same samples in the learned self-organizing map;
(D) classifying the cluster into one corresponding to a flaw and one not corresponding to a flaw according to a determination result input by a user as to whether the cluster corresponds to a flaw or not, and Means for generating a second determination data map;
The apparatus of claim 1, comprising: A computer program that causes a computer to determine a scratch on an inspection object,
The computer, the first and second sites predetermined of said object, said first judgment data map and a second learned the wound at the site of the second learned wounds at the first site A storage device for storing the judgment data map ;
A function of setting an inspection region to include a determination required region having a luminance difference exceeding a predetermined threshold with respect to another region in the image of the object imaged by the imaging device;
A function of extracting a feature vector from the inspection region;
If it is determined that the examination region belongs to the first part, the first judgment data map is selected. If it is judged that the examination area belongs to the second part, the second judgment data map is selected. With the ability to select
The Enter the feature vector to the first or second judgment data map that is the selected, a function of determining flaw in the examination region, was implemented in the computer,
The first part is a part in which the object is imaged over the entire inspection region and does not include a background image of the object, and the second part is a part of the inspection region. The object is imaged, and the remaining part includes the background image.
Computer program. The inspection object is a gear, the first portion is a central portion of the gear, and the second portion is an end portion of the gear.
The computer program according to claim 4 . For the predetermined first and second parts of the object, the device further includes a function of generating the first and second data maps, respectively , and the function of generating further includes
(A) a function of defining a self-organizing map for each of the first and second parts of the object;
(B) For the object, preparing a plurality of samples in the inspection area, and for each of the inspection areas of the plurality of samples,
(B-1) extracting a feature vector from the inspection region;
(B-2) Of the first and second predetermined parts of the object, determine a part to which the examination region belongs, and select a self-organizing map corresponding to the part;
(b-3) a function of inputting the feature vector extracted in (b-1) into the selected self-organizing map and learning the function vector;
(C) In the learned self-organizing map, a function corresponding to each of the same samples and grouping adjacent neurons into a cluster;
(D) classifying the cluster into one corresponding to a flaw and one not corresponding to a flaw according to a determination result input by a user as to whether the cluster corresponds to a flaw or not, and A function of generating a second determination data map ,
The computer program according to claim 4 .

Images