JP3130462B2 - Passage obstruction inspection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a passage obstruction inspection method for inspecting whether or not an inside of a curved passage whose inner surface reflects light is inspected.

[0002]

2. Description of the Related Art For example, when inspecting the quality of an article, a binarized image is extracted from a two-dimensional planar image of the article, and the binarized image is identified to judge the quality. For example, although not known, an inspection means shown in FIG. 3 has been proposed. In the above-mentioned inspection means, when inspecting whether or not the inside of the passage (2b) is blocked in the engine head cylinder (1) having the bent inverted T-shaped passage (2) whose inner surface reflects light, first, the light connected to the light source The fiber (3) is inserted into the passage entrance (2a), and the diffused light (La) is projected inside the passage (2b). Then, if the inside of the passage (2b) is not closed, the diffused light (La) is diffusely reflected on the inner surface and exits from the passage exit (2c) as it is. Then, the emitted light (Lb) is received and imaged on a two-dimensional plane by the camera (4), and is captured at a predetermined threshold value.
After the binarization, a binarized image (Wa) shown in FIG. 4A is extracted. On the other hand, if the core (m) or the like during the molding of the head cylinder is left inside the passage (2b) and closed,
The amount of light (Lb) emitted from the passage outlet (2c) decreases.
Then, as shown in FIG. 4B, when the plane image taken at the passage exit (2c) is binarized with the same threshold value, the binarized image (Wb) is reduced.

[0003] Therefore, image features such as the area and perimeter of the binarized image are measured, and the measured data is compared with normal data (for example, a reference area) to identify the image features, and at the same time, the inside of the passage is identified. The presence or absence of the blockage in (2b) is determined.

[0004]

The problem to be solved is to extract the binarized images (Wa) and (Wb) from the planar image of the passage (2) by binarization using a predetermined threshold value. Since the threshold value is fixed, false information is generated without accurately extracting data of the passage (2), and a good product may be erroneously determined as a defective product. For example, when the inner surface (2d) inside the passage (2b) is smooth, the surface is shining, so that the binarized image (W) of the passage exit (2c) shown in FIG.
Get a). However, when the inner surface (2d) is rough, the reflectance decreases and the surface does not shine, so that the amount of light emitted from the passage outlet (2c) decreases even if the inside of the passage (2b) is not closed. Therefore, when the captured image is binarized with the same threshold value, a reduced binarized image (Wb) similar to that at the time of occlusion shown in FIG. The problem that a false information is generated occurs.

[0005]

According to the present invention, in inspecting the inside of a curved passage whose inner surface reflects light, the diffused light is emitted from the passage entrance and the light emitted from the passage exit is inspected. A step of receiving and capturing an image on a two-dimensional plane, measuring an image feature amount of the captured image, determining the conformity of the measured data to normal data by fuzzy inference, and inspecting whether there is a blockage inside the passage. It is characterized by including.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a passage obstruction inspection method according to the present invention will be described below with reference to FIGS.
First, in the above inspection, as shown in FIG.
The light projecting end of the optical fiber (3) is inserted into the passage entrance (2a), and diffused light (La) is projected into the passage inside (2b). Then, the light diffusely reflects inside the passage (2b) and exits again from the exit (2c), and the emitted light (Lb) is reflected by the camera (4).
The image is received and imaged on a two-dimensional plane. Therefore, the binarization threshold (H) is automatically set in accordance with the path (2), and the planar captured image is binarized by the threshold (H) to have a luminous intensity equal to or higher than a predetermined level. Take out the digitized image.
Then, feature values such as the brightness area value and the perimeter of the binarized image are measured, and the degree of conformity of the measured data to the normal data is determined by fuzzy inference to check whether the inside of the passage (2b) is blocked. .

In the setting of the threshold value, first, a membership function of fuzzy inference is created by measuring distributions of feature amounts such as the area and perimeter of an image serving as a plurality of references of a passage as an object to be inspected in advance. deep. Further, as shown in FIGS. 1A and 1B, a plurality of binarization thresholds (Ha) (Ha) (Ha) ( Set Hb) and (Hc). Then, in the case of a non-defective product, as the level of the threshold (H) decreases, the binarized image area of each level gradually increases. Conversely, in the case of a defective product, the binarized image area is constant or FIG. It becomes 0 as shown in d).

Therefore, the plane image picked up by the camera (4) is first binarized with the maximum threshold (Ha), and a binarized image (Da) shown in FIG. The binarized image area (Su) at that time is determined by a fuzzy inference or the like with respect to the corresponding reference area, and if it substantially matches the reference area,
The threshold value (Ha) is set as the optimum threshold value (Ho), and the process proceeds to the next blockage inspection. Also, binarized image area (Su)
Is larger than the reference area, the level of the maximum threshold (Ha) is set to a still higher level and readjusted.

Next, when the binarized image area (Su) is smaller than the reference area, the inspected object is defective or the luminance distribution as shown by the dotted line in FIG. This is the case where even though the passage is good, the inner surface of the passage (2d) is rough or satin-like, so that the amount of emitted light decreases. in this case,
The level of the binarized image is re-measured at the next higher threshold value (Hb) by lowering the level by one step. Binary image area at that time (Sv)
Is larger than the previous threshold value (Ha), or approximately equal to the corresponding reference area, it is judged as good and the threshold value (Hb) at that time is optimized. Set as the threshold (Ho). Also, if it gets too big,
The threshold slightly raised from the threshold (Hb) is set as the optimum value. And even if the threshold is lowered, the binarized image area (Sv)
Is constant or 0, it is determined to be defective.
The above operation is repeated a plurality of times, for example, about five times sequentially,
If the binarized image area is constant or 0, it is determined to be defective. In this manner, the binarized image area is determined, and the threshold level is automatically adjusted as appropriate to set an optimal binarized threshold (Ho).

There is also a means for directly setting the optimum threshold value (Ho) using the center of gravity calculation of fuzzy inference. For example, as shown in FIG. 2A, the membership functions (Ma) and (Mb) of the binarized image area and the threshold are set as the input / output unit. Next, the binarized image area (Sa) is extracted from the luminance distribution shown in FIG. 1A by a predetermined threshold value, and the fitness (Ta) (Tb) is calculated from the membership function (Ma) of the input unit.
Is determined. Then, the fitness (Ta) (Tb) is substituted into the membership function (Mb) of the output section, and as shown in FIG. 2 (b), the optimum is calculated by the center of gravity calculation of the corresponding combined trapezoidal area (shaded area). Calculate the threshold (Ho).

Once the optimum binarization threshold (Ho) is set, the threshold (Ho) is then set as shown in FIG.
Then, the binarized image (Do) is extracted from the image captured by the camera (4) based on, and the feature amount of the image (Do), for example, the area and the perimeter are measured. Then, as shown in FIG. 2D, a membership function (Mc) of fuzzy inference is set in advance for each passage exit, and for example, measured area data is substituted into the membership function (Mc). Therefore, the degree of conformity (α) of the measured data to the normal data is derived from the membership function (Mc) of fuzzy inference. Therefore, the degree of conformity (α) is compared with the reference value (A), and when α> A, it is determined to be a non-defective product, and the presence or absence of the blockage inside the passage (2b) is inspected. At this time, a plurality of passage exits may be imaged at a time, and the suitability of the plurality of images for each exit may be combined, and for example, the presence or absence of occlusion may be determined from the multiplied value. Further, in addition to the area and the perimeter, the position of the center of gravity of the image and the like are added as discriminating factors, and multiplication by each degree of conformity can improve the inspection accuracy for those at the boundary between good and bad.

The membership function (Mc) measures the area and perimeter of a plurality of non-defective products or defective products, and draws a probability distribution with (Na) as a normal region and (Nb) and (Nc) as closed regions. It is created for each passage exit and for each feature quantity such as area and perimeter. For example, if the measured area is (Pa) in the area distribution, the degree of conformity to the normal data is (Qa), and (Qa) is compared with (A). Based on the magnitude, the blockage of the inside of the passage (2b) is determined. Determine the presence or absence. Note that the inside of the area (Nb) indicates a closed area generated when the amount of light decreases due to the exit side of the passage or the core, and the area (Nb)
The inside of c) shows a closed area generated by an increase in the amount of light due to a closed flash at the middle of the passage or at the entrance.

[0013]

According to the present invention, when inspecting the inside of a passage such as a head cylinder having a curved passage whose inner surface reflects light, diffused light is emitted from the passage entrance to irregularly reflect the inside. The light passing through the passage is received at the exit of the passage on a two-dimensional plane and imaged, and the degree of conformity of the measured data of the image feature amount of the captured image to the normal data is determined by fuzzy inference to check for the presence of obstruction inside the passage. As a result, the inspection error can be reduced to prevent the occurrence of false alarm, and the erroneous determination is eliminated, thereby improving the inspection accuracy. Further, re-inspection becomes unnecessary, and the number of steps is reduced.

[Brief description of the drawings]

FIG. 1A is a graph showing an example of a brightness distribution of a captured image and respective binarization thresholds showing an embodiment of a passage obstruction inspection method according to the present invention. (B) is another example of the brightness distribution of the captured image and a graph of each binarization threshold, showing the embodiment of the passage obstruction inspection method according to the present invention. (C) is a diagram showing an example of an image binarized by the threshold values of FIGS. 1 (a) and (b). (D) is a diagram showing a case where an image cannot be taken out with the threshold values of FIGS. 1 (a) and (b).

FIG. 2A shows each membership function of an input / output unit for setting a threshold value by fuzzy inference. (B) is a membership function showing an example of calculating the center of gravity of fuzzy inference.
(C) is a diagram showing an example of a binarized image. (D) is a waveform diagram showing an example of a fuzzy judgment membership function.

FIG. 3 is a side view showing an example of a passage.

FIG. 4A is a diagram showing an image binarized by a conventional threshold at the passage exit of FIG. 3; FIG. 4B is a diagram showing another image binarized by a conventional threshold value at the passage exit in FIG. 3.

[Explanation of symbols]

 Ba, Bb luminance distribution

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06T 1/00 G06T ⁷ /00-7/60 F02F 1/24-1/42 G01B 11/24-11 / 255 G01N 21/49-21/53 G01N 21/88-21/958 G01J 1/44

Claims (1)

(57) [Claims] When inspecting the inside of a curved passage whose inner surface reflects light, the diffused light is projected from the passage entrance and the light emitted from the passage exit is received on a two-dimensional plane. A passage that includes an imaging step and a step of measuring an image feature amount of the captured image, determining a degree of conformity of the measured data to normal data by fuzzy inference, and examining whether or not the passage is closed; Obstruction inspection method.