JP4150390B2 - Appearance inspection method and appearance inspection apparatus - Google Patents

Description

  The present invention relates to an appearance inspection method and an appearance inspection apparatus for inspecting the appearance of an inspection object.

Conventionally, an imaging unit that images an electronic component and outputs image data, and an inspection target range on the electronic component based on the image data captured by the imaging unit, and each of the image data in the inspection target range Inspection area specifying means for assigning orthogonal coordinates to pixels, and average brightness calculating means for calculating the average brightness of a plurality of pixels connected in a direction parallel to one coordinate axis in the orthogonal coordinates within the inspection target range for each other coordinate In addition, there is known an electronic component appearance inspection apparatus that includes a luminance difference calculating unit that calculates a luminance difference between the luminance of each pixel and the average luminance corresponding to the pixel (see, for example, Patent Document 1).
JP-A-6-42935

  However, even when an inspection method using an appearance inspection apparatus as described in Patent Document 1 is used, when the inspection object has a curved surface, different polishing streaks or the like are formed for each inspection object. In such a case, it is difficult to detect the defect depending on the shape of the defect generated in the inspection object.

  The present invention provides an appearance inspection method and an appearance inspection apparatus capable of reliably detecting defects in an inspection object, regardless of the shape of the defect in the inspection object or different polishing streaks formed on each inspection object. The purpose is to provide.

  An appearance inspection method according to the present invention is an appearance inspection method for performing an appearance inspection of an inspection object, the step of imaging the inspection object and obtaining a captured image including the inspection object image, and the inspection object For each first pixel column composed of a series of pixels parallel to the first direction along the outer edge of the image, the average of the luminance values of the series of pixels included in the first pixel column is calculated, and the first direction Calculating the average of the luminance values of a series of pixels included in the second pixel column for each second pixel column composed of a series of pixels parallel to a second direction orthogonal to the first direction and A step of calculating an average luminance distribution of the inspection object image in the second direction, a step of generating a reference image serving as a reference for the inspection object image based on the average luminance distribution, and the captured image and the reference image A step of generating a first corrected image, and a Sobel filter. A step of calculating a change amount of a luminance value of each pixel in the first corrected image to generate a Sobel processed image, a step of generating a second corrected image from the Sobel processed image and the first corrected image, and a second corrected image And binarizing to identify a connected area, and comparing the number of pixels in the connected area with a threshold value.

  In the appearance inspection method according to the present invention, the inspection object is imaged, and the average luminance distribution is obtained for each inspection object in the first direction and the second direction along the outer edge of the inspection object image. A reference image serving as a reference for the inspection object image is generated based on each average luminance distribution, and a first corrected image is generated from the captured image and the reference image. For this reason, since the captured image is individually corrected according to the reference image generated for each inspection object, even if the polishing streaks formed on the inspection object are individually different The defect of each inspection object can be detected. Further, the second corrected image is generated by processing the first corrected image using the Sobel processed image generated by calculating the change amount of the luminance value of the first corrected image by the Sobel filter. It is possible to detect a defect in the inspection object with high accuracy, and in particular, it becomes easy to detect a crack in the inspection object and a foreign matter attached to the inspection object.

  Further, in the step of generating the first corrected image, the luminance value at each pixel of the reference image is subtracted from the luminance value at each pixel of the captured image for each pixel at a position corresponding to the captured image and the reference image. It is preferable to generate the first corrected image. In this way, the amount of change in luminance value in the area other than the defective portion of the first corrected image is extremely small (the luminance value is substantially uniform), while the amount of change in luminance value is extremely large in the region of the defective portion. . As a result, a range in which a threshold value for binarizing the second corrected image generated thereafter can be set, and the threshold value can be easily set.

  Further, in the step of generating the second corrected image, the luminance at each pixel of the Sobel processed image is calculated from the luminance value at each pixel of the first corrected image for each pixel at the corresponding position in the first corrected image and the Sobel processed image. It is preferable to generate the second corrected image by subtracting the value. In this way, in the second corrected image, it is possible to widen the difference between the luminance value in the defective portion area and the luminance value other than the defective portion. Can be detected more easily.

  A step of calculating a change amount of a luminance value of each pixel in the first correction image by a differential filter to generate a differential processing image; a step of generating a third correction image from the differential processing image and the first correction image; It is preferable that the method further includes a step of binarizing the third corrected image to identify a connected area and comparing the number of pixels in the connected area with a threshold value. In this way, it becomes easy to detect a defect due to a chip, in which the amount of change in luminance value can vary from a small value to a large value in the defective portion region.

  Further, in the step of generating the third corrected image, for each pixel at a position corresponding to the differentiated image and the first corrected image, the luminance value in each pixel of the differentiated image and the luminance in each pixel of the first corrected image. It is preferable to generate the third corrected image by adding the value. In this way, in the third corrected image, it is possible to widen the difference between the magnitude of the luminance value in the defective area and the magnitude of the luminance value in the area other than the defective area, so that it becomes easier to detect defects due to chipping. .

  Moreover, it is preferable to further include a step of binarizing the captured image to identify a connected area and comparing the number of pixels in the connected area with a threshold value. In this way, it is possible to detect a defect due to peeling having a large area, which is corrected in the first correction image.

  On the other hand, an appearance inspection apparatus according to the present invention is an appearance inspection apparatus for inspecting the appearance of an inspection object, and illuminates the inspection object and images the inspection object illuminated by the illumination means. Included in the first pixel column for each first pixel column consisting of an imaging unit and a series of pixels parallel to the first direction along the outer edge of the inspection object image included in the captured image captured by the imaging unit A series of pixels included in the second pixel column for each second pixel column composed of a series of pixels parallel to a second direction orthogonal to the first direction. Means for calculating an average luminance distribution of the inspection object image in the first direction and the second direction by calculating an average of the luminance values of the pixels, and a reference image serving as a reference for the inspection object image Means for generating based on the average luminance distribution; Means for generating a first corrected image from an image image and a reference image, means for calculating a change amount of a luminance value of each pixel in the first corrected image by a Sobel filter and generating a Sobel processed image, and a Sobel processed image And means for generating a second correction image from the first correction image, means for binarizing the second correction image to identify a connected area, and comparing the number of pixels in the connected area with a threshold value. Features.

  In the appearance inspection apparatus according to the present invention, the inspection object is imaged by the imaging means, and the average luminance distribution is obtained for each inspection object in the first direction and the second direction along the outer edge of the inspection object image. . Then, a reference image corresponding to the inspection object imaged based on each average luminance distribution is generated, and a first correction image is generated from the captured image and the reference image. For this reason, since the captured images are individually corrected according to the reference image generated for each inspection object, each inspection object can be obtained even when the polishing streaks formed on the inspection object are individually different. Defects can be detected. Further, the second corrected image is generated by processing the first corrected image using the Sobel processed image generated by calculating the change amount of the luminance value of the first corrected image by the Sobel filter. It is possible to detect a defect in the inspection object with high accuracy, and in particular, it becomes easy to detect a crack in the inspection object and a foreign matter attached to the inspection object.

  Further, the means for generating the first corrected image subtracts the luminance value at each pixel of the reference image from the luminance value at each pixel of the captured image for each pixel at a position corresponding to the captured image and the reference image. It is preferable to generate the first corrected image. In this way, the amount of change in luminance value in the area other than the defective portion of the first corrected image is extremely small (the luminance value is substantially uniform), while the amount of change in luminance value is extremely large in the region of the defective portion. . As a result, a range in which a threshold value for binarizing the second corrected image generated thereafter can be set, and the threshold value can be easily set.

  Further, the means for generating the second corrected image is configured such that, for each pixel at a position corresponding to the first corrected image and the Sobel processed image, the luminance at each pixel of the Sobel processed image from the luminance value at each pixel of the first corrected image. It is preferable to generate the second corrected image by subtracting the value. In this way, by reflecting the amount of change in the luminance value of each pixel in the first corrected image obtained by the Sobel filter (the luminance value of each pixel in the Sobel processed image) in the first corrected image, Since the difference between the magnitude of the brightness value and the brightness value other than the defective portion widens, by binarizing such a second corrected image, defects such as cracks and foreign matter attached to the inspection object can be detected more. It becomes easy.

  A means for calculating a change amount of a luminance value of each pixel in the first correction image by a differential filter to generate a differential processing image; a means for generating a third correction image from the differential processing image and the first correction image; It is preferable to further include means for binarizing the third corrected image to identify a connected area and comparing the number of pixels in the connected area with a threshold value. In this way, it becomes easy to detect a defect due to a chip, in which the amount of change in luminance value can vary from a small value to a large value in the defective portion region.

  Further, the means for generating the third correction image is configured to generate a luminance value for each pixel of the differential processing image and a luminance for each pixel of the first correction image for each pixel at a position corresponding to the differential processing image and the first correction image. It is preferable to generate the third corrected image by adding the value. In this way, the luminance value of the defective portion is reflected by reflecting, in the first corrected image, the amount of change in the luminance value of each pixel in the first corrected image obtained by the differential filter (the luminance value of each pixel in the differentiated image). Since the difference between the magnitude of the value and the magnitude of the brightness value other than the defective portion is widened, binarization of such a second corrected image makes it easier to detect defects due to chipping.

  Moreover, it is preferable to further include a step of binarizing the first corrected image to identify a connected area and comparing the number of pixels in the connected area with a threshold value. In this way, it is possible to detect a defect due to a somewhat large peeling that is corrected in the first correction image.

  According to the present invention, it is possible to reliably detect the defect of the inspection object regardless of the shape of the defect in the inspection object or the different polishing streaks formed on each inspection object.

  Preferred embodiments of the present invention will be described with reference to the drawings. In the description, the same reference numerals are used for the same elements or elements having the same function, and a duplicate description is omitted.

(Configuration of appearance inspection equipment)
With reference to FIG. 1, the structure of the external appearance inspection apparatus 10 which concerns on this embodiment is demonstrated. FIG. 1A is a perspective view schematically showing a configuration of an appearance inspection apparatus, and FIG. 1B is a perspective view of an inspection object.

  The appearance inspection apparatus 10 is an apparatus for inspecting the appearance of an inspection object 22 to be described later. For this purpose, the appearance inspection apparatus 10 includes a camera 12, an image processing unit 14, a display 16, and LED illuminators 18a and 18b.

  The camera 12 images the outer surface of the inspection object 22 illuminated by the LED illuminators 18a and 18b. For example, a CCD (Charge Coupled Device) camera can be used. When the camera 12 acquires the captured images P1 and P13 (see FIGS. 7A and 10A) of the inspection object 22, the camera 12 outputs data of the captured image P1 to the image processing unit 14. The camera 12 may be a CMOS (Complementary Metal Oxide Semiconductor) camera, an infrared camera, a black and white camera, a color camera, or the like, as long as it can obtain at least luminance information.

  The image processing unit 14 includes an unillustrated ECU (Electronic Control Unit) including a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like, and a captured image output by the camera 12. The image processing is performed. When the image processing unit 14 performs image processing on the captured images P1, P13, and the like, the image processing unit 14 outputs an image signal to the display 16 so that the processed image is displayed on the display 16.

  The display 16 displays various images such as captured images P1 and P13, a first corrected image P3 described later, and the presence or absence of defects. Therefore, the appearance of the inspection object 22 can be inspected by human eyes by observing various images displayed on the display 16 by the operator.

  The LED illuminators 18a and 18b are ring-shaped lighting fixtures that can illuminate the inspection object 22 from the direction of 360 ° around the LED illuminators 18a and 18b, and the inspection object 22 is placed above the inspection object 22. Each is provided so as to face the plate 20. The LED illuminator 18a is disposed on the mounting plate 20 side, has a substantially cylindrical shape, and a plurality of LED light sources are arranged on the inner surface thereof. For this reason, the LED illuminator 18a illuminates from the inner side surface toward the center of the LED illuminator 18a. On the other hand, the LED illuminator 18b is disposed on the camera 12 side, and has an inner surface of a substantially conical shape, and a plurality of LED light sources are arranged on the inner surface. For this reason, the LED illuminator 18b illuminates downward from the inner surface with a predetermined angle. By using these LED illuminators 18a and 18b, the inspection object 22 is uniformly illuminated.

  Here, in this embodiment, the inspection object 22 has an outer peripheral surface 22a and an inner peripheral surface 22b that are formed into a cylindrical surface, and the side surface 22c has an arc shape. The outer peripheral surface 22a and the inner peripheral surface 22b of the inspection object 22 are in the axial direction of the outer peripheral surface 22a and the inner peripheral surface 22b that form a cylindrical surface in the process of manufacturing the inspection object 22, and each inspection object Different polishing streaks are formed for each object 22. The inspection object 22 is used as a magnet for a motor such as a cooler, for example.

  Examples of defects that can occur in the inspection object 22 include cracks, chips, and peeling. A crack has a relatively small size (area) compared to other defects. Since the crack has a deep defect depth (defect depth) from the surface of the inspection object 22, the amount of change in the luminance value in the defect portion of the captured image tends to be a large value. The chipping can occur from a small to a large defect size (area). Since the defect has a defect depth from a shallow one to a deep one, the amount of change in the luminance value in the defective portion of the captured image can take a small value to a large value. In peeling, the size (area) of defects is relatively large. In peeling, since the defect depth is relatively shallow, the amount of change in the luminance value in the defect portion of the captured image tends to be a small value.

(Appearance inspection method using appearance inspection equipment)
Next, an appearance inspection processing method for the inspection object 22 using the appearance inspection apparatus 10 having the above configuration will be described with reference to FIGS. In addition, especially the case where the external appearance test | inspection of the outer peripheral surface 22a or the internal peripheral surface 22b used as the cylindrical surface shape among the test objects 22 is performed is demonstrated.

  FIG. 2 is a flowchart showing a procedure from the start to the end of the appearance inspection process. FIG. 3 is a flowchart showing a procedure from the start to the end of the corrected image creation process. FIG. 4 is a flowchart showing a procedure from the start to the end of the first defect detection process. FIG. 5 is a flowchart showing a procedure from the start to the end of the second defect detection process. FIG. 6 is a flowchart showing a procedure from the start to the end of the third defect detection process. FIG. 7 shows each image in the defect detection first process, (a) shows a captured image, (b) shows a master image, and (c) shows a first corrected image. (D) is a diagram showing a Sobel processed image, (e) is a diagram showing a second corrected image, (f) is a diagram showing a binarized image, and (g) is a defect detection. It is a figure which shows a result image. 8A is a diagram showing an inspection object image for explaining a method of calculating the average luminance value, FIG. 8B is a diagram showing a distribution of average luminance values in the axis X direction, and FIG. ) Is a diagram showing a distribution of average luminance values in the axis Y direction. FIG. 9 shows each image in the defect detection second process, (a) shows the first corrected image, (b) shows the differential processed image, and (c) shows the third corrected image. (D) is a figure which shows a binarized image, (e) is a figure which shows a defect detection result image. FIG. 10 shows each image in the third defect detection process, (a) shows a captured image, and (b) shows a binarized image. In the flowcharts shown in FIGS. 2 to 6, the step is abbreviated as S. 7, 9, and 10, the defect detection first process, the defect detection second process, and the defect detection third process are described using different inspection objects 22.

  When the appearance inspection process is started by the appearance inspection apparatus 10, the process proceeds to Step 1 shown in FIG. 2, and the inspection object 22 is positioned by the operator of the appearance inspection apparatus 20. Specifically, by pressing the inspection object 22 against a pin (not shown) provided on the mounting plate 20 with a spring, the polishing streaks of the inspection object 22 become short sides (in the inspection object image P1a). The inspection object 22 is fixed to the mounting plate 20 along the axis Y direction shown in FIG. In this way, the outer edges of the captured image P1 and the image (inspection object image) P1a indicating the inspection object 22 in the captured image P1 are parallel, and a first corrected image P3 described later can be easily obtained from the captured image P1. Become.

  Subsequently, when proceeding to Step 2, the inspection object 22 illuminated by the LED illuminators 18a and 18b is imaged by the camera 12 (see FIG. 7A). When the inspection object 22 is imaged by the camera 12, data of the captured image P <b> 1 is output to the image processing unit 14. In the first corrected image generation process, the defect detection first process, the defect detection second process, and the defect detection third process described below, the defect of the inspection target 22 is detected by the image processing unit 14 based on the data of the captured image P1. Various processes are performed for this purpose.

(First corrected image generation process)
In the subsequent step 3, a first corrected image generation process is performed. When the first corrected image generation process is started, the process proceeds to step 11 shown in FIG. 3, and after obtaining a region of continuous pixels whose luminance value is not 0 in the captured image P <b> 1, a rectangle circumscribing the region of the pixel Is obtained to determine the outer shape area A1 of the inspection object image P1a in the captured image P1. At this time, a process of calculating the determined area (number of pixels) of the outer region A1 and the coordinates where the outer region A1 is located is also performed. In step 12, it is determined whether or not the area of the outer shape area A1 determined in step 11 is equal to or larger than a predetermined threshold value. That is, since the area (number of pixels) of the inspection object image P1a that is a target for detecting a defect is substantially the same in each inspection object 22, imaging is performed if the area (number of pixels) of the outer region A1 is equal to or greater than the threshold value. If it is determined that the object to be inspected is the inspection object 22 and the process proceeds to step 13, otherwise, it is determined that the imaged object is not the inspection object 22 (for example, a fragment of the inspection object 22). The process proceeds to step 8 shown in FIG. 2 to end the appearance inspection process.

In step 13, an area (processing area) A2 to be processed in the captured image P1 is determined based on the outer area A1. The processing for determining the processing area A2 will be specifically described below with reference to FIG. In addition, in the outline area A1 shown in FIG. 8A, each rectangular area surrounded by a white line grid corresponds to a pixel (for the sake of illustration, the size is exaggerated from the actual size). The coordinates of the pixel along the axis X direction parallel to the long side are 1 to x _max , and the coordinates of the pixel along the axis Y direction parallel to the short side and orthogonal to the axis X are 1 to y _max . . In addition, the luminance value of a pixel at an arbitrary coordinate (x, y) in the outer region A1 is represented as I (x, y).

これらの平均輝度値Ｘ，Ｙを求めた後、図８（ｂ），（ｃ）においてそれぞれ破線で示される閾値と平均輝度値Ｘ，Ｙとを比較して、閾値と等しい平均輝度値Ｘ，Ｙを示すｘ座標及びｙ座標をそれぞれ求める。そして、このように求められた各座標を通り、各座標が存在する軸Ｘ又は軸Ｙに垂直な直線によって囲まれる領域を、処理領域Ａ２に決定する。なお、外形領域Ａ１を複数に分割し、それぞれの領域について平均輝度値を求めることで、処理領域Ａ２を決定してもよい。 When determining the processing area A2, first, the average (average luminance value) Y of the luminance values I (1, y) to I (x _max , y) for a series of pixels parallel to the axis X direction is calculated in the axis Y direction. Calculate every time. Further, an average (average luminance value) X of the luminance values I (x, 1) to I (x, y _max ) for a series of pixels parallel to the axis Y direction is calculated for each axis X direction. The average luminance distributions L1 and L2 shown in FIGS. 8B and 8C represent changes in the average luminance values X and Y calculated in this way, respectively. The average luminance value Yn when the Y coordinate is n is expressed by the following equation (1), and the average luminance value Xm when the X coordinate is m is expressed by the following equation (2).

After obtaining these average luminance values X and Y, the threshold values indicated by the broken lines in FIGS. 8B and 8C are compared with the average luminance values X and Y, respectively, and the average luminance values X and X equal to the threshold values are compared. An x coordinate and a y coordinate indicating Y are respectively obtained. Then, a region that passes through the coordinates thus obtained and is surrounded by a straight line perpendicular to the axis X or the axis Y where the coordinates exist is determined as the processing region A2. The processing area A2 may be determined by dividing the outer area A1 into a plurality of areas and obtaining an average luminance value for each area.

ここで、式（３）におけるα，βは、それぞれ軸Ｘ，Ｙ方向における平均輝度値の寄与率となっている。マスタ画像Ｐ２を生成する際、寄与率α，βのうち平均輝度値Ｘ´，Ｙ´について優先させたい側の値を大きく設定することで、生成されるマスタ画像Ｐ２を調節することができる。なお、本実施形態では、軸Ｘ方向に沿う研磨筋の影響を考慮するため、βよりもαの値を大きくしている。 Subsequently, when proceeding to step 14, a master image (reference image) P2 is generated. When generating the master image P2, first, average luminance values X ′ and Y ′ are obtained again in the processing area A2. Based on these average luminance values X ′ and Y ′, the luminance value I ′ (m, n) of the pixel at the coordinates (m, n) is calculated for all the pixels in the processing area A2 from the following equation (3). A master image P2 is generated (see FIG. 7B).

Here, α and β in Equation (3) are the contribution ratios of the average luminance values in the directions of the axes X and Y, respectively. When generating the master image P2, the master image P2 to be generated can be adjusted by setting a larger value on the side of the contribution ratios α and β to be prioritized with respect to the average luminance values X ′ and Y ′. In the present embodiment, the value of α is made larger than β in order to take into account the influence of the polishing bars along the axis X direction.

  Subsequently, when proceeding to Step 15, for each pixel located at the corresponding coordinates in the inspection object image P1a and the master image P2 in the processing area A2, the master image P2 is calculated from the luminance value in each pixel of the inspection object image P1a. The luminance value at each pixel is subtracted. Thereby, the 1st correction image P3 is produced | generated (refer FIG.7 (c)), and a 1st correction image production | generation process is complete | finished.

また、ソーベル処理画像Ｐ４を生成した後、ソーベル処理画像Ｐ４における所定の画素とその画素の周囲における隣接画素とで、下記の式（５）による演算をソーベル処理画像Ｐ４の全ての画素について行う。これにより、ソーベル処理画像Ｐ４における各画素の輝度値Ｉ´の平均Ｍ１が求められ、ソーベル処理画像Ｐ４に生じているランダムなノイズを除去する平滑化が行われる。

なお、本実施形態では、階調を２５６階調（８ｂｉｔ）としているため、変化量Ｅ１を求める際に、変化量Ｅ１が下限値である０を下回る場合には変化量Ｅ１を下限値である０に固定し、変化量Ｅ１が上限値である２５５を上回る場合には変化量Ｅ１を上限値である２５５に固定する処理を行っているが、階調を２５６階調以外として同様の処理を行ってもよい。 (Defect detection first process)
Returning to FIG. 2 and proceeding to step 4, the first defect detection process is performed. In this defect detection first process, for example, a defect due to a crack in the inspection object 22 or a foreign substance attached to the inspection object 22 is detected. When the first defect detection process is started, the process proceeds to step 21 shown in FIG. 4, where the first corrected image P3 is processed by the Sobel filter in the processing area A2, and the change in density (edge) in the first corrected image P3 is processed. (Part) is detected. Specifically, the calculation according to the following equation (4) is performed for all the pixels of the first corrected image P3 with a predetermined pixel in the first corrected image P3 and adjacent pixels around the pixel. As a result, a change amount E1 of the luminance value I ′ of each pixel in the first corrected image P3 is obtained, and a Sobel processed image P4 in which edge detection is performed at a portion where a change in shading occurs is generated (FIG. 7). (See (d)).

In addition, after generating the Sobel processed image P4, the calculation according to the following equation (5) is performed for all the pixels of the Sobel processed image P4 with a predetermined pixel in the Sobel processed image P4 and adjacent pixels around the pixel. Thereby, the average M1 of the luminance values I ′ of each pixel in the Sobel processed image P4 is obtained, and smoothing is performed to remove random noise generated in the Sobel processed image P4.

In the present embodiment, since the gradation is set to 256 gradations (8 bits), when the change amount E1 is less than 0, which is the lower limit value, when the change amount E1 is obtained, the change amount E1 is the lower limit value. When the change amount E1 exceeds the upper limit value of 255, the process is performed to fix the change amount E1 to the upper limit value of 255. However, the same processing is performed with gradations other than 256 gradations. You may go.

  Subsequently, when proceeding to step 22, for each pixel located at the corresponding coordinates in the first corrected image P3 and the Sobel processed image P4 in the processing area A2, the Sobel processed image is calculated from the luminance value in each pixel of the first corrected image P3. The luminance value in each pixel of P4 is subtracted. Thereby, the 2nd correction image P5 is produced | generated (refer FIG.7 (e)). Further, after generating the second corrected image P5, the second corrected image P5 is smoothed in the same manner as in step 21 in order to remove random noise generated in the second corrected image P5.

  Subsequently, when proceeding to Step 23, the obtained second corrected image P5 is binarized by a predetermined threshold value, pixels having a luminance value equal to or higher than the threshold value are black, and pixels having a luminance value smaller than the threshold value are selected. A binary image P6 displayed as white is generated (see FIG. 7F). Here, the threshold value used in the binarization process in step 23 is a value obtained by conducting an experiment in advance (the threshold values used in steps 33 and 41 described later are also the same). In this embodiment, the threshold values in steps 23 and 33 are the same, and the threshold value in step 41 is larger than the threshold values in steps 23 and 33. This is because in the second and third corrected images before the binarization processing in steps 23 and 33, the luminance values in the regions other than the defective portion are substantially uniform by the Sobel filter and the differential filter. While the range in which the threshold can be set is wide, in the inspection object image P13a before the binarization processing in step 41, the luminance value in the region other than the defective portion is not substantially uniform, so the threshold is set. This is because the range that can be set is narrow. Then, when proceeding to step 24, the binarized image P6 is labeled, and a connected area where pixels displayed in white in the binarized image P6 are connected is specified. In step 24, processing is performed to detect the number of identified connected regions, the area (number of pixels) of each connected region, and the position of each connected region in the processing region A2.

  Subsequently, when proceeding to step 25, it is determined whether or not the area (number of pixels) of each connected region is smaller than a predetermined threshold for each connected region detected in step 24. As a result, if it is determined that the area of the connected region is smaller than the threshold value for all the connected regions, the first defect detection process ends. If not, the process proceeds to step 8 in FIG. It is determined that there is a defect in the inspection object 22 on which the inspection is performed, and the appearance inspection process is terminated. Here, the threshold value used in the determination of step 25 is also a value obtained by conducting an experiment in advance (the threshold values used in steps 35 and 43 described later are also the same). In this embodiment, the threshold values in steps 25 and 35 are the same, and the threshold value in step 43 is larger than the threshold values in steps 25 and 35. This is because the area (number of pixels) of the defect due to the crack that is the detection target in the defect detection first process and the defect that is the defect that is the detection target in the defect detection second process is the detection target in the defect detection third process. This is because the area (number of pixels) of the defect due to the peeling tends to be smaller. FIG. 7G shows each pixel of the inspection object image P1a and the binarized image P6 for each pixel located at coordinates corresponding to the inspection object image P1a and the binarized image P6 in the processing area A2. A defect detection result image P7 obtained by adding the respective pixels is obtained.

また、微分処理画像Ｐ９を生成した後、微分処理画像Ｐ９における所定の画素とその画素の周囲における隣接画素とで、下記の式（７）による演算を微分処理画像Ｐ９の全ての画素について行う。これにより、微分処理画像Ｐ９における各画素の輝度値Ｉ´の平均Ｍ２が求められ、微分処理画像Ｐ９に生じているランダムなノイズを除去する平滑化が行われる。

なお、欠陥検出第２処理においても、変化量Ｅ２が下限値を下回る場合には変化量Ｅ２を下限値に固定し、変化量Ｅ２が上限値を上回る場合には変化量Ｅ２を上限値に固定する処理を行っている。 (Defect detection second process)
Returning to FIG. 2 and proceeding to step 5, the second defect detection process is performed. In this defect detection first process, for example, a defect due to a chip of the inspection object 22 is detected. When the second defect detection process is started, the process proceeds to step 31 shown in FIG. 5 where the first corrected image P8 (see FIG. 9A) is processed by the differential filter in the processing area A2, and the first corrected image is processed. Processing for detecting a change in shading (edge portion) in P8 is performed. Specifically, the calculation according to the following equation (6) is performed for all the pixels of the first corrected image P8 with a predetermined pixel in the first corrected image P8 and adjacent pixels around the pixel. As a result, a change amount E2 of the luminance value I ′ of each pixel in the first corrected image P8 is obtained, and a differentiated image P9 is generated in which the edge is detected at the portion where the shade change occurs (FIG. 9). (See (b)).

Further, after generating the differential processed image P9, the calculation according to the following expression (7) is performed for all the pixels of the differential processed image P9 with a predetermined pixel in the differential processed image P9 and adjacent pixels around the pixel. Thereby, the average M2 of the luminance values I ′ of each pixel in the differential processing image P9 is obtained, and smoothing is performed to remove random noise generated in the differential processing image P9.

In the second defect detection process, the change amount E2 is fixed to the lower limit value when the change amount E2 is lower than the lower limit value, and the change amount E2 is fixed to the upper limit value when the change amount E2 exceeds the upper limit value. Processing to do.

  Subsequently, when the process proceeds to step 32, the luminance value and the differential processing image in each pixel of the first correction image P8 are determined for each pixel located at the corresponding coordinates in the processing area A2 in the first correction image P8 and the differential processing image P9. The luminance value in each pixel of P9 is added. Thereby, the 3rd correction image P10 is produced | generated (refer FIG.9 (c)). Further, after the generation of the third corrected image P10, the third corrected image P10 is smoothed in the same manner as in step 21 in order to remove random noise generated in the third corrected image P10.

  Subsequently, when proceeding to Step 33, the obtained third corrected image P10 is binarized by a predetermined threshold value, pixels having a luminance value equal to or higher than the threshold value are white, and pixels having a luminance value smaller than the threshold value are selected. A binary image P11 displayed as black is generated (see FIG. 9D). Then, when proceeding to step 34, the binarized image P11 is labeled, and a connected area where pixels displayed in white in the binarized image P11 are connected is specified. In step 34, processing is performed to detect the number of identified connected regions, the area (number of pixels) of each connected region, and the position of each connected region in the processing region A2.

  Subsequently, when proceeding to step 35, it is determined whether or not the area (number of pixels) of each connected region is smaller than a predetermined threshold for each connected region detected in step 34. As a result, if it is determined that the area of the connected area is smaller than the threshold value for all the connected areas, the second defect detection process ends. If not, the process proceeds to step 8 in FIG. It is determined that there is a defect in the inspection object 22 on which the inspection is performed, and the appearance inspection process is terminated. FIG. 9 (e) shows each pixel of the inspection object image and the binarized image P11 for each pixel located at coordinates corresponding to the inspection object image (not shown) and the binarized image P11 in the processing area A2. This is a defect detection result image P12 obtained by adding the respective pixels.

(Defect detection third process)
Returning to FIG. 2 and proceeding to step 6, a third defect detection process is performed. In the first defect detection process, for example, a defect due to peeling of the inspection object 22 is detected. When the third defect detection process is started, the process proceeds to step 41 shown in FIG. 6, where the inspection object image P13a in the processing area A2 is binarized by a predetermined threshold value, and the luminance value is equal to or higher than the threshold value. A binarized image P14 is generated in which the pixels are displayed as white and the pixels having a luminance value smaller than the threshold value are displayed as black (see FIG. 10B). Then, when proceeding to step 42, the binarized image P14 is labeled, and a connected region where pixels displayed in white in the binarized image P14 are connected is specified. In step 42, processing is performed to detect the number of identified connected regions, the area (number of pixels) of each connected region, and the position of each connected region in the processing region A2.

  Subsequently, when proceeding to step 43, it is determined whether or not the area (number of pixels) of each connected region is smaller than a predetermined threshold for each connected region detected in step 42. As a result, when it is determined that the area of the connected region is smaller than the threshold value for all the connected regions, the binarization process is completed, and the process proceeds to step 7 in FIG. It is determined that the object 22 is not defective, and the appearance inspection process is completed. On the other hand, if it is determined that the area of any of the connected regions is greater than or equal to the threshold, the process proceeds to step 8 in FIG. 2 to determine that the inspection object 22 currently undergoing visual inspection is defective. Then, the appearance inspection process ends.

  By the way, when an appearance inspection is performed on an object having a curved surface such as the inspection object 22, light incident on the inspection object 22 from the LED illuminations 18 a and 18 b is reflected according to the outer shape of the inspection object 22. . For this reason, when the inspection object 22 is imaged by the camera 12, light and darkness is locally generated, and the luminance distribution in the captured images P1 and P13 is not uniform. Further, since different polishing streaks are individually formed on the inspection object 22, the luminance distribution in the captured images P1 and P13 is different for each inspection object 22. Therefore, conventionally, there has been a problem that it is difficult to detect a defect of the inspection object 22.

  On the other hand, in the present embodiment, the average luminance distributions L1 and L2 are obtained for the X and Y directions for each inspection object 22, and the master image P2 is generated based on the average luminance distributions L1 and L2. Further, the first corrected image P3 is generated by subtracting the luminance value at each pixel of the master image P2 from the luminance value at each pixel of the inspection object image P1a in the processing area A2. For this reason, a master image P2 is generated according to each inspection object 22, and the inspection object image P1a in the processing area A2 is individually corrected by the master image P2. Therefore, the polishing streaks formed on the inspection object 22 Even if they are different from each other, the defect of each inspection object 22 can be detected. Further, the second corrected image P5 is generated by processing the first corrected image P3 using the Sobel processed image P4 generated by calculating the change amount of the luminance value of the first corrected image P3 by the Sobel filter. Therefore, it is possible to detect the defect of the inspection object 22 with high accuracy, and in particular, it becomes easy to detect the crack of the inspection object and the foreign matter attached to the inspection object.

  Further, in the present embodiment, for each pixel at a position corresponding to the inspection object image P1a and the master image P2 in the processing area A2, the luminance value in each pixel of the inspection object image P1a is used for each pixel of the master image P2. The first corrected image P3 is generated by subtracting the luminance value. For this reason, the amount of change in luminance value in the region other than the defective portion of the first corrected image P3 is extremely small (the luminance value is substantially uniform), while the amount of change in luminance value is extremely large in the region of the defective portion. As a result, the range in which a threshold value for binarizing the second corrected image P5 generated thereafter can be set, and the threshold value can be easily set.

  Further, in the present embodiment, for each pixel at a corresponding position in the first corrected image P3 and the Sobel processed image P4, the luminance value in each pixel of the Sobel processed image P4 from the luminance value in each pixel of the first corrected image P3. Is subtracted to generate a second corrected image. Therefore, by reflecting the amount of change in the luminance value of each pixel in the first corrected image P3 obtained by the Sobel filter (the luminance value of each pixel in the Sobel processed image P4) in the first corrected image P3, Since the difference between the magnitude of the brightness value and the magnitude of the brightness value other than the defective portion is widened, by binarizing the second corrected image P5, defects such as cracks and foreign matters attached to the inspection object are further detected. It becomes easy to do.

  In the present embodiment, since the third corrected image P10 is generated by processing the first corrected image P3 using the differential filter, the amount of change in the luminance value in the defective portion region takes a small value to a large value. It becomes easier to detect defects due to chipping.

  Further, in the present embodiment, for each pixel at a corresponding position in the differential processing image P9 and the first correction image P3, the luminance value in each pixel of the differential processing image P9 and the luminance value in each pixel of the first correction image P3. Is added to generate the third corrected image P10, and therefore, in the third corrected image P10, the difference between the magnitude of the luminance value in the defective area and the magnitude of the luminance value in the area other than the defective area is widened. This makes it easier to detect defects due to chipping.

  In the present embodiment, since the connection region is specified by binarizing the inspection object image P13a in the processing region A2, it is possible to detect even a defect due to peeling having a large area to the inspection object 22.

  Although the preferred embodiments of the present invention have been described in detail above, the present invention is not limited to the above-described embodiments. For example, in the present embodiment, the defect detection first to third processes are performed in the appearance inspection process, but only the defect detection first process may be performed among these processes. Further, the defect detection first and second processes may be performed, and the defect detection first and third processes may be performed. Furthermore, execution may be started from any of the processes of the defect detection first to third processes.

  In the present embodiment, the appearance inspection is performed on the inspection target 22 having a curved surface. However, the present invention is not limited to this, and the appearance inspection of a planar inspection target may be performed.

  In step 11, the outer region A1 may be determined by searching for the edge of the inspection object image P1a from four directions.

(A) is a perspective view which shows the structure of an external appearance inspection apparatus roughly, (b) is a perspective view of a test target object. It is a flowchart which shows the procedure from the start of an external appearance inspection process to completion | finish. It is a flowchart which shows the procedure from the start to the completion | finish of a correction image creation process. It is a flowchart which shows the procedure from the start of a 1st defect detection process to completion | finish. It is a flowchart which shows the procedure from the start of a 2nd defect detection process to completion | finish. It is a flowchart which shows the procedure from the start of a 3rd defect detection process to completion | finish. Each image in defect detection 1st processing is shown, (a) is a figure showing a picked-up image, (b) is a figure showing a master image, (c) is a figure showing the 1st amendment image, d) is a diagram showing a Sobel processed image, (e) is a diagram showing a second corrected image, (f) is a diagram showing a binarized image, and (g) is a defect detection result image. FIG. (A) is a figure which shows the test target object image for demonstrating the calculation method of an average luminance value, (b) is a figure which shows distribution of the average luminance value in the axis | shaft X direction, (c) is an axis | shaft Y It is a figure which shows distribution of the average luminance value in a direction. Each image in defect detection 2nd processing is shown, (a) is a figure showing the 1st amendment image, (b) is a figure showing a differential processing image, and (c) is a figure showing the 3rd amendment image. (D) is a figure which shows a binarized image, (e) is a figure which shows a defect detection result image. Each image in a defect detection 3rd process is shown, (a) is a figure which shows a captured image, (b) is a figure which shows a binarized image.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 ... Appearance inspection apparatus, 12 ... Camera, 14 ... Image processing part, 22 ... Inspection object, P1, P13 ... Captured image, P1a, P13a ... Inspection object image, P2 ... Master image, P3, P8 ... 1st correction | amendment Image, P4 ... Sobel processed image, P5 ... Second corrected image, P9 ... Differential processed image, P10 ... Third corrected image.

Claims (8)

An appearance inspection method for performing an appearance inspection of an inspection object,
Imaging the inspection object to obtain a captured image including the inspection object image;
For each first pixel column composed of a series of pixels parallel to a first direction along the outer edge of the inspection object image, an average of the luminance values of the series of pixels included in the first pixel column is calculated. Calculating an average of luminance values of the series of pixels included in the second pixel column for each second pixel column composed of a series of pixels parallel to the second direction orthogonal to the first direction. And calculating an average luminance distribution of the inspection object image in the first direction and the second direction;
Generating a reference image serving as a reference for the inspection object image based on the average luminance distribution;
By subtracting the luminance value at each pixel of the reference image from the luminance value at each pixel of the captured image for each pixel at a position corresponding to the captured image and the reference image, the captured image and the reference image Generating a first corrected image from:
Calculating a change amount of a luminance value of each pixel in the first corrected image by a Sobel filter to generate a Sobel processed image;
By subtracting the luminance value at each pixel of the Sobel processed image from the luminance value at each pixel of the first corrected image for each pixel at a corresponding position in the first corrected image and the Sobel processed image, Generating a second corrected image from the Sobel processed image and the first corrected image;
Binarizing the second corrected image to identify a connected area, and comparing the number of pixels in the connected area with a threshold ;
Calculating a change amount of a luminance value of each pixel in the first corrected image by a differential filter to generate a differential processed image;
For each pixel in a position corresponding to the differential processing image and the first correction image, by adding the luminance value in each pixel of the differential processing image and the luminance value in each pixel of the first correction image, Generating a third corrected image from the differentiated image and the first corrected image;
A visual inspection method comprising: binarizing the third corrected image to identify a connected area, and comparing the number of pixels in the connected area with a threshold value . The visual inspection method according to claim 1 , further comprising a step of binarizing the captured image to identify a connected area, and comparing the number of pixels in the connected area with a threshold value. An appearance inspection apparatus for inspecting the appearance of an inspection object,
Illuminating means for illuminating the inspection object;
Imaging means for imaging the inspection object illuminated by the illumination means;
The series included in the first pixel column for each first pixel column composed of a series of pixels parallel to the first direction along the outer edge of the inspection object image included in the captured image captured by the imaging unit. The series of pixels included in the second pixel column for each second pixel column consisting of a series of pixels parallel to a second direction orthogonal to the first direction. Means for calculating an average luminance distribution of the inspection object image in the first direction and the second direction by calculating an average of luminance values of the pixels;
Means for generating a reference image as a reference for the inspection object image based on the average luminance distribution;
By subtracting the luminance value at each pixel of the reference image from the luminance value at each pixel of the captured image for each pixel at a position corresponding to the captured image and the reference image, the captured image and the reference image Generating a first corrected image from:
Means for calculating a change amount of a luminance value of each pixel in the first corrected image by a Sobel filter and generating a Sobel processed image;
By subtracting the luminance value at each pixel of the Sobel processed image from the luminance value at each pixel of the first corrected image for each pixel at a corresponding position in the first corrected image and the Sobel processed image, Means for generating a second corrected image from the Sobel processed image and the first corrected image;
Means for binarizing the second corrected image to identify a connected area, and comparing the number of pixels in the connected area with a threshold ;
Means for calculating a change amount of a luminance value of each pixel in the first corrected image by a differential filter and generating a differential processed image;
For each pixel at a corresponding position in the differential processing image and the first correction image, by adding the luminance value in each pixel of the differential processing image and the luminance value in each pixel of the first correction image, Means for generating a third corrected image from the differentiated image and the first corrected image;
An appearance inspection apparatus comprising: means for binarizing the third correction image to identify a connected area, and comparing the number of pixels in the connected area with a threshold value . The visual inspection apparatus according to claim 3 , further comprising a step of binarizing the captured image to identify a connected area, and comparing the number of pixels in the connected area with a threshold value.   An appearance inspection method for performing an appearance inspection of an inspection object,
  Imaging the inspection object to obtain a captured image including the inspection object image;
  For each first pixel column composed of a series of pixels parallel to a first direction along the outer edge of the inspection object image, an average of the luminance values of the series of pixels included in the first pixel column is calculated. Calculating an average of luminance values of the series of pixels included in the second pixel column for each second pixel column composed of a series of pixels parallel to the second direction orthogonal to the first direction. And calculating an average luminance distribution of the inspection object image in the first direction and the second direction;
  A reference image serving as a reference for the inspection object image is obtained by multiplying the average luminance distribution of the inspection object in the first direction by a first contribution rate and the inspection object in the second direction. Generating based on the average luminance distribution of the object multiplied by a second contribution rate;
  By subtracting the luminance value at each pixel of the reference image from the luminance value at each pixel of the captured image for each pixel at a position corresponding to the captured image and the reference image, the captured image and the reference image Generating a first corrected image from:
  Calculating a change amount of a luminance value of each pixel in the first corrected image by a Sobel filter to generate a Sobel processed image;
  By subtracting the luminance value at each pixel of the Sobel processed image from the luminance value at each pixel of the first corrected image for each pixel at a corresponding position in the first corrected image and the Sobel processed image, Generating a second corrected image from the Sobel processed image and the first corrected image;
  A visual inspection method comprising: binarizing the second corrected image to identify a connected area, and comparing the number of pixels in the connected area with a threshold value.   On the outer surface of the inspection object, polishing streaks along the first direction are formed,
  6. The appearance inspection method according to claim 5, wherein in the step of generating the reference image, the first contribution rate is set to be larger than the second contribution rate.   An appearance inspection apparatus for inspecting the appearance of an inspection object,
  Illuminating means for illuminating the inspection object;
  Imaging means for imaging the inspection object illuminated by the illumination means;
  The series included in the first pixel column for each first pixel column composed of a series of pixels parallel to the first direction along the outer edge of the inspection object image included in the captured image captured by the imaging unit. The series of pixels included in the second pixel column for each second pixel column consisting of a series of pixels parallel to a second direction orthogonal to the first direction. Means for calculating an average luminance distribution of the inspection object image in the first direction and the second direction by calculating an average of luminance values of the pixels;
  A reference image serving as a reference for the inspection object image is obtained by multiplying the average luminance distribution of the inspection object in the first direction by a first contribution rate and the inspection object in the second direction. Means for generating based on the average luminance distribution of the object multiplied by a second contribution rate;
  By subtracting the luminance value at each pixel of the reference image from the luminance value at each pixel of the captured image for each pixel at a position corresponding to the captured image and the reference image, the captured image and the reference image Generating a first corrected image from:
  Means for calculating a change amount of a luminance value of each pixel in the first corrected image by a Sobel filter and generating a Sobel processed image;
  By subtracting the luminance value at each pixel of the Sobel processed image from the luminance value at each pixel of the first corrected image for each pixel at a corresponding position in the first corrected image and the Sobel processed image, Means for generating a second corrected image from the Sobel processed image and the first corrected image;
  An appearance inspection apparatus comprising: means for binarizing the second corrected image to identify a connected area, and comparing the number of pixels in the connected area with a threshold value.   On the outer surface of the inspection object, polishing streaks along the first direction are formed,
  8. The appearance inspection apparatus according to claim 7, wherein the means for generating the reference image is set so that the first contribution rate is larger than the second contribution rate.

Images