JP2001296252A - Defect detection method of object surface, and device thereof - Google Patents

Abstract

PROBLEM TO BE SOLVED: To correctly detect a defect, even if a characteristic similar to the defect exists of if defect information has an interruption or a deletion. SOLUTION: This defect detection device is composed of a measuring device 2 for irradiating plural liner light beams at intervals on the side of a detection object 1, a TV camera 3 for imaging plural linear patterns obtained thereby, an image-processing device 6 for inspecting ruggedness defects on the side through images processing from the plural linear patterns. The image-processing device 6 has a processing function for extracting one or more defective character quantities, showing the shape of a defective characteristic part which is a deformation of each linear pattern has a deformation, evaluating the shape of a defect candidate, which is a group of plural defective characteristic parts ranging in one direction relative to each defective characteristic part, based on the defective characteristic quantities of the plural defective characteristic parts included in the defect candidate, for determining the degree and the kind of the defect on the side, and extracting an index as defect reliability from the defect candidate, to thereby determine whether the defect candidate is a defect.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for projecting a laser beam or the like onto a cylindrical inspection object and projecting the specular reflection light onto a screen. A method and apparatus for detecting a defect on a surface of an object, which generates a linear pattern that bends in accordance with a change in the surface direction of the object and detects an irregularity defect that may be present on the surface of the object by processing an image obtained by photographing the linear pattern. It is.

[0002]

2. Description of the Related Art In order to detect minute defects on the surface of a planar inspection object, a method of detecting the inclination of the surface of the inspection object is disclosed in Japanese Patent Application Laid-Open No. Hei 10-9838. A camera and a light source are arranged to capture the specular reflection component of the light projected on the object surface by the camera, and a plurality of images are taken while changing the angle of the light source and the camera,
A method is disclosed in which the light source angle at which the brightness at the same position on the object surface is maximized is determined.

As a method for inspecting the appearance of a cylindrical object, as shown in Japanese Patent Application Laid-Open No. 8-151163, a light source is arranged around the surface of the cylindrical object so as to project the band light, and the band light and its light are arranged. A method for detecting a defect from a change in brightness in the vicinity is disclosed.

As a method for detecting an extremely shallow concave / convex defect existing on the surface of a cylindrical object, Japanese Patent Laid-Open No. 11-2114 discloses a method.
As disclosed in Japanese Patent No. 42, a method is disclosed in which a laser beam is projected on the surface of an object, the specular reflection component is projected on a screen, and an image is taken by a camera.

A method of discriminating individual defects from an image in which a plurality of defects having different shapes are photographed in the same screen and determining the shape, defect type, degree and the like is as follows.
As disclosed in JP-A-5-280959, the image is binarized by threshold processing, area division is performed for each defective part, and information of the defective part is projected in the X and Y directions of the image. A method is disclosed in which dimensional information is used as peripheral distribution information of a defect to discriminate between a linear defect and a point-like defect, and a defect type or the like is determined based on the feature amount of each defect.

[0006]

DISCLOSURE OF THE INVENTION Problems to be Solved by the Invention
According to the method disclosed in Japanese Patent Publication No. 38-38, a fine uneven defect existing on the surface of a planar object can be detected very clearly, but the inspection method is a method suitable for inspecting a planar object. Therefore, it is difficult to apply the method to inspect unevenness defects existing on the surface of a cylindrical object such as a beverage can.

In the method disclosed in Japanese Patent Application Laid-Open No. H11-212442 as a means for detecting a minute defect on the surface of a cylindrical object, a laser beam is projected on the surface of the object, and the specular reflection component is projected on a screen. By using the imaging method, an extremely shallow defect existing on the surface of the cylindrical object can be detected without rotating the cylindrical object.

[0008] In the above method, since the inclination of the surface along the axial direction of the cylindrical object changes in a portion having irregularities on the surface of the cylindrical object, the regular reflection component of the emitted laser is projected on the screen. This is based on the fact that the position at which the image is projected changes and a linear projection pattern (hereinafter referred to as a linear pattern) bends. In addition, since the laser light hitting the bottom of the defect does not reach the screen at the irregularity defect having a relatively large depth, the bending of the linear pattern is observed only in the vicinity of the defect in the image taken of the screen. At the center of the part, the linear pattern is interrupted. Therefore, in this method, by detecting the portion where the interruption occurs in addition to the bending of the linear pattern, it is possible to distinguish the defective portion from the non-defect unevenness (very shallow unevenness).

However, in the above method, the linear pattern projected on the screen has an indistinct line shape due to the fine structure of the surface of the cylindrical object, and since the bent pattern is not observed in the linear pattern, it cannot be recognized as a defective portion. There is. That is, in the obtained linear pattern, for example, a long streak-like defect portion may be interrupted, and it may appear that there are a plurality of short streak-like defects.

Further, if foreign matter such as water droplets adheres to the surface of the inspection object, the linear pattern is interrupted, which may be erroneously recognized as a defective portion.

Further, in the method disclosed in Japanese Patent Application Laid-Open No. H8-15163 described above as another prior art relating to a method of inspecting the appearance of a cylindrical object, a belt-like light in which a light source is reflected on the surface of the cylindrical object and its vicinity. By observing the change in the brightness of the object, it is possible to inspect the cylindrical object.

However, in this method, the strip light is deformed due to the defect, and the defect is detected by using the crossing of the lightness measurement line arranged near the strip light. However, when a change appears in the data, there is a problem that erroneous detection occurs. For example, when water droplets do not adhere to a cylindrical object, the water droplets act as a lens and condense surrounding light, so that a light spot similar to the feature of a defect is formed on the surface of the cylindrical object. May be erroneously detected as a defect.

In the canning process of carbonated beverages,
Inject at a low temperature to suppress foaming, then return to normal temperature with warm water before shipping, but it is difficult to completely remove water droplets attached to the can in such a process,
It becomes necessary to distinguish between water droplets and defects. However,
The above method does not judge the size of the feature, etc., because it is a qualitative method to consider as a defect if there is a light spot near the band light,
The distinction between water droplets and defects is not possible and is not applicable in such situations.

As described above, Japanese Patent Application Laid-Open No. HEI 5-280959 discloses a method for processing an image in which there is a feature that is likely to be erroneously recognized as a defective portion and one defect appears to be divided into a plurality.
According to the method disclosed in Japanese Patent Application Laid-Open No. H10-209, a plurality of areas including one defect are generated by one-dimensionally projecting the inspection image in the X and Y directions and extracting information indicating linearity / pointiness. This method has the drawback that it can be applied only when the defective portion is horizontal or vertical. In addition, since a plurality of regions at the same position in the X and Y directions are all regarded as one defect, the length of the defective portion is reduced. There is a danger that the data will go out of order.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has a method of detecting a defect on an object surface capable of correctly detecting a defect even if a feature similar to the defect exists or the defect information is interrupted or missing. And an apparatus thereof.

[0016]

According to a first aspect of the present invention, a plurality of linear light beams are irradiated on a curved surface of an object to be inspected at intervals. By taking a plurality of linear patterns arranged at an angle along the surface by the shape light, from a plurality of linear patterns obtained by this shooting, through the image processing of the linear pattern processing step and the defect detection step, A defect detection method for detecting an irregular defect on a surface of the inspection object, wherein the linear pattern processing step includes, for each of a plurality of linear patterns obtained by the photographing, the linear pattern A defect feature detecting step of detecting the deformation as a defect feature, if any, and a defect feature extracting step of extracting at least one defect feature representing a shape of the defect feature. Excessive A defect candidate extracting step of extracting, as a defect candidate, a group of a plurality of defect feature parts connected in a direction intersecting with the direction of the linear pattern with respect to the plurality of defect feature parts obtained in the linear pattern processing step. And a defect for evaluating at least one of a degree and a type of a defect on the surface of the inspection object by evaluating a shape of the defect candidate based on defect feature amounts of a plurality of defect characteristic portions included in the defect candidate. A shape evaluation step, and a defect reliability evaluation step of extracting an index as defect reliability representing the likelihood of a defect from the defect candidate to determine whether the defect candidate is a defect. And

According to a second aspect of the present invention, in the method for detecting a defect on an object surface according to the first aspect, the linear pattern in the process of detecting the defect characteristic portion is a deformation extending in directions opposite to each other after the interruption. It is characterized by the following.

According to a third aspect of the present invention, in the method for detecting a defect on the surface of an object according to the first aspect, the defect feature amount in the defect feature amount extracting step includes the direction of a linear pattern including the defect feature portion and the The size of the defect feature portion is along at least one of the directions orthogonal to the direction.

According to a fourth aspect of the present invention, in the method for detecting a defect on an object surface according to the first aspect, the defect feature amount in the defect feature amount extracting step is the brightness of the defect feature portion.

According to a fifth aspect of the present invention, in the method for detecting a defect on the surface of an object according to the first aspect, the processing of the defect candidate extracting step includes the steps of:
If another linear pattern adjacent to or close to the linear pattern including the defect feature includes another defect feature connected to the defect feature in the intersecting direction, both of these defect features are replaced with the defect feature. It is characterized in that processing to be included in a group as a candidate is sequentially performed.

According to a sixth aspect of the present invention, in the method for detecting a defect on the surface of an object according to the fifth aspect, each of the plurality of defect features is close to a linear pattern including the defect feature. If another linear feature within a predetermined distance has another defect feature that is continuous with the defect feature in the intersecting direction, a process of including both of these defect features in the group as the defect candidates Are sequentially performed.

According to a seventh aspect of the present invention, in the method for detecting a defect on an object surface according to the fifth aspect, at least two of the plurality of defect candidates obtained by the processing are connected to form a new defect candidate. It is characterized by the following.

According to an eighth aspect of the present invention, in the method for detecting a defect on an object surface according to the first aspect, the defect feature amounts of the plurality of defect feature portions included in the defect candidate are statistical quantities of the plurality of defect feature amounts. And at least one of a distance between both ends of the defect candidate.

According to a ninth aspect of the present invention, in the method for detecting a defect on the surface of an object according to the first aspect, the defect reliability in the defect reliability evaluation step is based on the number of defect features included in the defect candidate. It is a value obtained by dividing by the number of a plurality of linear patterns corresponding to a plurality of defect features included in a defect candidate.

According to a tenth aspect of the present invention, there is provided a measuring means for irradiating a plurality of linear lights at intervals on a curved surface of an object to be inspected, and an angle along the surface by the plurality of linear lights. A photographing means for photographing a plurality of linear patterns to be arranged, and a plurality of linear patterns obtained by the photographing,
A defect detecting apparatus comprising image processing means for detecting an uneven defect on the surface of the inspection object through image processing of a linear pattern processing step and a defect detection step, wherein the linear pattern processing step is For each of the plurality of linear patterns obtained in the photographing, if there is a deformation in the linear pattern, a defect feature detection step of detecting the deformation as a defect feature, and a shape of the defect feature. Extracting at least one defect feature quantity to be represented, wherein the defect detection step comprises: detecting a plurality of defect feature parts obtained in the linear pattern processing step, and detecting a direction of the linear pattern. A defect candidate extraction process of extracting a group of a plurality of defect feature portions connected in a direction intersecting with the defect candidate as a defect candidate, and a defect feature amount of the plurality of defect feature portions included in the defect candidate. Evaluating the shape of the candidate, a defect shape evaluation step of determining at least one of the degree and type of a defect on the surface of the inspection object, and an index as defect reliability representing the likelihood of a defect from the defect candidate A defect reliability evaluation step of extracting and determining whether or not the defect candidate is a defect.

[0026]

FIG. 1 is a block diagram of an apparatus for detecting a defect on the surface of an object, FIG. 2 is a view showing an example of an optical pattern photographed by the TV camera of FIG. 1, and FIG. FIG. 4 is a diagram showing a state in which a linear pattern according to the surface shape of the inspection object is formed by the emitted linear light, and an embodiment of the present invention will be described below with reference to these drawings.

The defect detecting device shown in FIG. 1 detects the presence or absence of an irregular defect on the curved surface (the side surface in FIG. 1) of the inspection object 1 such as a beverage can. , TV camera 3, quantization device 4, storage device 5
And an image processing device 6.

The measuring device 2 comprises a conveyor 21 for placing the inspection object 1 with its axis oriented in a vertical direction and transporting the object 1 in one direction (to the left in the figure) at a constant speed. The inspection apparatus 1 includes a pass sensor 22 for detecting whether or not the inspection object 1 conveyed by the sensor 21 has passed a predetermined position, and a linear light source 23. This linear light source 23
A laser diode 231 as a point light source, and a cylindrical lens 232 that diffuses the point light from the laser diode 231 into a plane and converts it into a linear light. The side surface is irradiated with a plurality of linear lights at an oblique interval from above, and a plurality of linear patterns arranged in a fan shape as shown in FIG. 2 at an angle along the side surface by the plurality of linear lights are conveyed. 4 for forming (projecting) on the upper surface. Then, the passage sensor 22 and the linear light source 23 cause the linear light to be substantially parallel to the axis on the side surface of the inspection target 1 and, as shown in FIG. Are arranged at an angle with respect to the normal direction of the side surface of.

The TV camera 3 is for photographing a plurality of linear patterns to be formed on the upper surface of the conveyor 21 from above, and the quantizing device 4 converts the signal from the TV camera 3 into a digital signal. The image is converted into an image, and the storage device 5 stores the digital image.

Here, the formation of a plurality of linear patterns by irradiating a plurality of spaced linear lights on the side surface of the inspection object 1 may be realized using a plurality of linear light sources. In the present embodiment, in order to realize using one linear light source 23, a detection result indicating that the inspection object 1 has passed a predetermined position is detected by the passage sensor 22 by using the conveyance of the inspection object 1 by the conveyor 21. Is obtained, using the detection result as a trigger, the linear light source 23 intermittently emits linear light at predetermined time intervals, and the TV camera 3 releases the shutter until the required number of irradiations of the linear light is completed. The following configuration is adopted. As a result, a plurality of linear light beams emitted from the linear light source 23 at predetermined time intervals are reflected on the side surface of the inspection object 1, and a plurality of linear patterns arranged in a fan shape on the upper surface of the conveyor 4. Are substantially formed, and the plurality of linear patterns are used as one optical pattern in the TV camera 3.
Will be taken.

The image processing device 6 determines the presence or absence of an uneven defect on the side surface of the inspection object 1 from a plurality of linear patterns obtained by the above photographing through image processing in a linear pattern processing step and a defect detection step. It is to detect. The linear pattern processing step includes, for each of the plurality of linear patterns obtained by photographing with the TV camera 3, a defect characteristic part detecting step of detecting the deformation as a defect characteristic part if the linear pattern is deformed. Extracting at least one defect feature representing the shape of the defect feature. The defect detection process includes the steps of: extracting a plurality of defect features obtained in the linear pattern processing process; A defect candidate extraction process of extracting a group of a plurality of defect feature portions connected in a direction intersecting with the direction of the pattern as a defect candidate, and a defect candidate extraction process based on the defect feature amounts of the plurality of defect feature portions included in the defect candidate. A defect shape evaluation process for evaluating the shape and determining at least one of the degree and type of a defect on the side surface of the inspection object 1 and extracting an index as a defect reliability indicating the probability of the defect from the defect candidate. Defect candidate and contains a defect reliability evaluation process determines whether the defect. The linear pattern processing process and the defect detection process will be described later in detail.

Here, the principle of various kinds of detection or discrimination used in the image processing will be described. A more detailed principle will be described later.

First, the principle of detection of an uneven defect will be described. As shown in FIG. 3A, when linear light is applied to the side surface of the inspection object 1, the linear light is substantially parallel to the axis of the side surface of the inspection object 1, so that the inspection object A portion Pn1 which is a substantially straight line of the linear pattern Pn is formed by the portion 11 which is a complete cylindrical side surface on one side surface.

On the other hand, if the linear light impinges on the uneven defect portion on the side surface of the inspection object 1, the linear light makes an angle with respect to the normal direction of the side surface of the inspection object 1. , The linear pattern Pn is deformed. That is, due to the downwardly inclined surface portion 12 on the side surface of the inspection object 1, a deformation Pn2 that turns left toward the conveyor 21 is generated in the linear pattern Pn and becomes a part of the linear pattern Pn, and the upward deformation on the side surface of the inspection object 1 occurs. By the part 14 of the surface inclined to
A deformation Pn4 that turns right toward the conveyor 21 occurs in the linear pattern Pn and becomes a part thereof.

If the portion 13 between the portions 12 and 14 is deep, an interruption as shown in FIG. 3A occurs at the position of the linear pattern Pn corresponding to this portion. A deformation Pn3 that extends further on one side is generated. This deformation Pn3 may exceed the shooting range of the TV camera 3, or the linear light may not reach the defective portion.

Therefore, the deformation Pn is added to the linear pattern Pn.
2, if Pn4 occurs and there is an interruption between them,
In other words, if deformations that extend in opposite directions occur after the interruption, it can be determined that there is a defect F1 deeper than a predetermined level on the side surface of the inspection object 1. Furthermore, if the portion 13 between the portions 12 and 14 is shallow, there is no interruption, and a dent (see FIGS. 5 and 6 described later) in which sinusoidal deformation occurs in which the leading ends extending left and right are inverted and connected. It can be considered as a defect.

Next, the principle of discriminating between the defect F1 and foreign matter will be described. As shown in FIG. 3A, if a foreign matter 15 such as dirt or water drops adheres to the side surface of the inspection object 1, a portion Pn5 that is simply interrupted in the linear pattern occurs. Therefore, the defect F1 and the foreign matter can be discriminated by simply seeing whether or not there is a break or deformations extending in opposite directions after the break. Thus, it is possible to solve the problem that foreign matter such as dirt and water droplets is erroneously detected as a defect.

If a streak-like defect F2 having a steep angle is present in the portion 16 of the streak defect shown in FIG. 3A, the interruption can be detected as in Pn6.
In some cases, it is not possible to detect pattern shapes extending in opposite directions after the interruption. In such a case, the foreign matter and the defect F2 cannot be distinguished. Next, the principle of these determinations will be described with reference to the drawings.

FIG. 4 is a perspective view of an inspection object as an example, FIG. 5 is an example of an optical pattern obtained for the inspection object of FIG. 4, and FIG. 6 is a linear pattern of the optical pattern of FIG. It shows the result obtained by dividing and extracting and normalizing the length. However, A is a streak defect including the defect F1, B is a point defect including the defect F1, C is a very shallow unevenness outside the detection target,
D indicates a foreign substance, and E in the streak defect A indicates a portion including the defect F2. In FIGS. 5 and 6, A in FIG.
The code is used for each part corresponding to .about.E.

First, from the optical pattern shown in FIG.
As shown in, each linear pattern is cut out to generate an image whose length is normalized, and thereafter, into individual linear patterns,
If there is a stagger pattern corresponding to the defect F1 (see Pf shown in FIG. 3A), the stagger pattern is detected as a defect feature. As a result, each portion of the very shallow unevenness C and foreign matter D is excluded from the defects to be detected, and at this stage, the portion E including the defect F2 to be detected is also excluded from the defects to be detected.

Subsequently, a group of a plurality of defect feature parts connected in a direction orthogonal to the direction of the linear pattern is extracted as a defect candidate. At this time, if another linear pattern adjacent to or close to the linear pattern including each defect characteristic portion has another defect characteristic portion connected to the defect characteristic portion in the orthogonal direction,
The process of including both of these defect feature parts in a group as defect candidates is sequentially performed. As a result, the portion E can be included in the streak defect A, the portion of the foreign matter D and the portion E of the defect F2 can be determined, and finally, the streak defect A and the point defect B can be determined. Can be detected.

FIG. 7 is a diagram summarizing the criteria for discriminating streak-like and dot-like defects, very shallow irregularities, foreign matter, and vertical surfaces (non-defective parts). Streak-like and dot-like defects and foreign matter can be distinguished by the size of the staggered pattern, that is, by the presence or absence of deformations extending in opposite directions after the break. Further, streak-like and dot-like defects and shallow irregularities can be distinguished by the presence or absence of line breaks.

If the interval between a plurality of linear lights on the side surface of the inspection object 1 is known, the length of the defect (La, L in FIG. 6) is determined from the number of linear patterns included in the defect candidate.
Since b) is known, the streak defect A and the point defect B can be quantitatively distinguished.

The processing of the linear pattern processing step and the defect detection step using the above-described principle of detection or discrimination is executed by the image processing device 6. Hereinafter, the processing procedure of the linear pattern processing process and the defect detection process will be described with reference to the drawings.

FIG. 8 is a flowchart showing a processing procedure of a linear pattern processing step and a defect detection step executed by the image processing apparatus, FIG. 9 is a flowchart of a defect characteristic part detection step and the like, and FIG. FIG. 11 is an explanatory diagram of a defect feature amount extraction process, FIG. 11 is an explanatory diagram of a defect candidate tracking process, FIG. 12 is a flowchart of a defect candidate extraction process,
FIG. 13 is an explanatory diagram of the defect shape evaluation process and the defect reliability evaluation process, and FIG. 14 is a flowchart of the defect shape evaluation process and the defect reliability evaluation process.

As shown in FIG. 8, the defect detection process includes a linear pattern processing step S1 for extracting a defect feature included in each linear pattern, and a positional relationship between the linear patterns of the defect feature. A defect detection process S2 for connecting defect feature parts and performing defect determination is basically classified. The linear pattern processing step S1 is divided into a defect feature detection step S11 and a defect feature extraction step S12.

In the defect feature detection step S11, the defect feature is detected by capturing the bending of the linear pattern generated by the vertical shape change of the side surface of the inspection object 1. If there is a defect F1 to be detected, a staggered pattern with an interruption of the linear pattern is seen.
Using the pattern, as shown in FIG. 10, a portion L1 where the linear pattern straddles the center line is detected, and positions Pr and Pl where the bending of the linear pattern is maximum above and below are detected (see FIG. 10). 9 (S110, S111). Here, if the width and depth described later are sufficiently large, it is regarded as a defect feature portion.

In the defect feature extraction step S12, a defect feature amount is determined for the defect feature detected in the defect feature detection step S11 so that the size of the defect can be quantitatively evaluated. As a defect feature amount, a distance (hereinafter, “depth”) L in a direction perpendicular to the direction of the linear pattern between points Pr and Pl
Two distances L3 along the direction of the linear pattern between the points 2 and Pr-Pl (hereinafter, "width") L3 are extracted (S11 in FIG. 9).
2). If the depth L2 and the width L3 are sufficiently large, it is determined to be a defect feature (Y in S113 in FIG. 9). Thereby, the bent shape of the linear pattern can be obtained.

Further, as shown in FIG. 10, the portion inside the staggered pattern, that is, the value of the average lightness I between the line segments Pr-Pl is also obtained as a feature value (S11 in FIG. 9).
4). By determining the brightness at the defect feature portion in this way, the brightness of the linear pattern is reduced at the defective portion as shown by I3 in FIG. 10, so that the interruption of the linear pattern can be determined. .

After step S114, the position, width, depth, and brightness of the defective feature are recorded (S115 in FIG. 9). The above process is repeated until the bottom of the linear pattern is reached (S116 in FIG. 9).

Next, the process proceeds to the defect detection step S2, which is a defect candidate extraction step S21.
And a defect shape evaluation process S22. However, the defect shape evaluation step S22 includes a defect reliability evaluation step described later.

In the defect candidate extraction step S21, the defect feature parts extracted in the linear pattern processing step S1 are connected to generate a defect candidate (see FIG. 11).

The process of the defect candidate extraction step S21 is performed according to the procedure shown in FIG. (1) As shown in FIG. 11, in order from the leftmost linear pattern, a defect feature d10 serving as a tracking start point is searched for (FIG. 1).
2 (S120, S121). (2) With the defect feature d10 on the linear pattern lm as a starting point, it is checked whether or not the defect feature d11 exists within a range of ± t above and below the position d10 on the linear pattern lm + 1 on the right side (S122, S123). (3) When the defect feature d11 exists (Y in S123), it is checked whether the defect feature d12 exists within a range of ± t above and below the position on the right side linear pattern lm ± 2 (S123). . (4) If it does not exist, it is checked whether or not the defect feature part d12 exists within a range of ± t above and below the position of d11 on the further right-hand linear pattern lm + 3 (S126). (5) The above-mentioned steps (3) and (4) are repeated, and if no defect features are found successively in n linear patterns, a group of defect features tracked (d10 to d17 in FIG. 11).
Are combined into one defect candidate A1 (S124). (6) Detect defect feature d20 serving as the next tracking start point,
Steps (2) to (5) are repeated. (7) Steps (1) to (6) are repeated until there is no defect feature portion serving as a tracking start point in the image (the tracking start point search reaches the bottom of the rightmost linear pattern) (S12).
8).

In the above processing, the numerical value of the condition n for determining the end of tracking in the procedure (5) is an allowable value of discontinuity included in one defect candidate. Unlike the method of extracting a region, it is possible to extract a defect candidate with a break. By increasing n, a sparser set of defect features can be extracted as defect candidates.

Next, a defect candidate combining process to cope with a case that cannot be dealt with by the method using the tracking method allowing a short break as described above will be described with reference to FIG. A defect candidate relatively densely including a defect feature portion, such as A2, has a discontinuity having a width of n or more (n = 2 in FIG. 13) between the two defect candidates A21 and A22.
May be detected separately. In such a case,
Distance d1 between end points of A21 and A22 and difference d in height d
2 are found, and if these are within a predetermined allowable value, a defect candidate A2 combining A21 and A22 is generated in addition to A21 and A22, and the connection relationship is recorded.

Next, the defect shape evaluation process and the defect reliability evaluation process will be described with reference to FIG. (11) For each defect candidate, an average value of the distance between both ends of the defect candidate (hereinafter, “length”) and the defect feature amounts (width, depth, brightness) of the defect feature part included in the defect candidate is obtained ( S220 and S221 in FIG. 14). It should be noted that the defect determination process in the defect shape evaluation process and the defect reliability evaluation process for the defect candidate is preferentially executed for the defect candidate (A2 in FIG. 13) generated by the above-described defect candidate combining process. (12) For each defect candidate, a defect reliability (density) represented by (length of defect feature / length of entire defect candidate) is obtained (S222). (13) Defect candidates with low defect reliability are excluded as false alarm parts (S223). (14) A defect candidate having a high brightness is excluded as a false alarm part (S224). (15) The defect type is determined based on the length (S22).
5 to S227). (16) Determining the degree of defect based on the depth and width (S
228). (17) Unnecessary determination information is excluded based on the connection relationship between defect candidates. The defect candidate currently focused on (A2 in FIG. 13)
1 or A22) is determined to be a defective portion by the above procedure, it is generated by the above-described defective portion combining process and is one of longer defect candidates (A2 in FIG. 13) previously determined to be a defective portion. If it is a defective portion, the defective portion currently focused on is not regarded as a defect (S229).

In the above procedure, procedure (13) is a defect reliability evaluation process, and procedures (14) to (17) are defect shape evaluation processes. Here, in FIG. 13, defect candidate A2 is a streak defect portion, A21 and A22 are portions where defect candidate A2 appears to be separated as described above, B2 is a point defect, C1 is a shallow unevenness, and D1 is a very small point. The false alarm part, D2, which is generated due to the continuous irregularities in the shape of a circle, is a foreign substance. These defect candidates are classified as follows. (21) The defect candidate D1 with low defect reliability is false information. Further, the foreign matter D2 is not recognized as a defective portion. (22) Based on the brightness information, the defect candidate C1 having a bright brightness at the defective portion is regarded as a false report. (23) According to the classification based on the length, A2 and C1 are classified as streak defects, and A21, A22 and B2 are classified as point defects. (24) Based on the depth and width information, A2 is determined to be a true streak defect, and A21, A22, and B2 are determined to be true point-like defects. (25) A21 and A22 are excluded from the defect determination because A2 obtained by combining them is determined to be a true stripe defect.

As described above, the defect A
2 and the point defect B2 alone are determined to be true defects,
Others are excluded as false information or unnecessary information.
As described above, by making a determination in consideration of not only the defect shape of the defect candidate but also the defect reliability, a correct defect can be determined.

[0059]

As is apparent from the above description, according to the first aspect of the present invention, a plurality of linear light beams are radiated at intervals to the curved surface of the inspection object. By taking a plurality of linear patterns arranged at an angle along the surface by the shape light, from a plurality of linear patterns obtained by this shooting, through the image processing of the linear pattern processing step and the defect detection step, A defect detection method for detecting an irregular defect on a surface of the inspection object, wherein the linear pattern processing step includes, for each of a plurality of linear patterns obtained by the photographing, the linear pattern If there is a deformation, a defect feature detection step of detecting the deformation as a defect feature, and a defect feature extraction step of extracting at least one defect feature representing the shape of the defect feature,
In the defect detecting step, a group of a plurality of defect feature parts connected in a direction intersecting with the direction of the linear pattern is extracted as a defect candidate for the plurality of defect feature parts obtained in the linear pattern processing step. A defect candidate extraction step, and evaluating a shape of the defect candidate based on defect feature amounts of a plurality of defect feature parts included in the defect candidate, and at least one of a degree and a type of a defect on the surface of the inspection object. And a defect reliability evaluation step of extracting an index as defect reliability representing the likelihood of a defect from the defect candidate to determine whether the defect candidate is a defect. Therefore, for example, a portion where the plane direction is changed is detected as a defect feature from each linear pattern, and a defect feature amount for the portion is extracted, and the obtained defect feature is obtained. In the tracking, a group of defect features is extracted as a defect candidate, and the degree and type of the defect are determined by using both the defect shape and the defect reliability for the defect candidate. Even when the image is unclear and includes noise, correct defect detection can be performed.

According to a second aspect of the present invention, in the method for detecting a defect on the surface of an object according to the first aspect, the deformation extending in opposite directions to the deformation of the linear pattern in the step of detecting the defect feature portion is performed. Therefore, it is possible to prevent the interruption of a simple linear pattern from being erroneously recognized as a defective feature.

According to a third aspect of the present invention, in the method for detecting a defect on the surface of an object according to the first aspect, the defect feature amount in the defect feature amount extracting step is defined as a direction of a linear pattern including the defect feature portion. And the size of the defect feature portion along at least one of the directions orthogonal to this direction. By extracting such a defect feature amount, it is possible to obtain useful information for defect determination in subsequent processing.

According to a fourth aspect of the present invention, in the defect detecting method for an object surface according to the first aspect, the defect feature amount in the defect feature amount extracting step is the brightness of the defect feature portion. By determining the discontinuity of the linear pattern generated at the center of the defect feature based on the original, it is possible to distinguish between the deep unevenness and the shallow unevenness.

According to a fifth aspect of the present invention, in the method for detecting a defect on the surface of an object according to the first aspect of the present invention, the defect candidate extraction process is performed on each of the plurality of defect characteristic portions. If another linear pattern adjacent to or close to the linear pattern including the portion has another defect characteristic portion connected to the defect characteristic portion in the intersecting direction, both of these defect characteristic portions are regarded as the defect candidates. The process of including in a group is sequentially performed. For example, if defect features of adjacent linear patterns are sequentially tracked and connected, and a defect candidate is generated, not only a horizontal defect candidate but also a bent defect candidate is extracted. be able to.

According to a sixth aspect of the present invention, in the method for detecting a defect on an object surface according to the fifth aspect, for each of the plurality of defect features, a linear pattern including the defect feature is provided. If another linear pattern that is within a predetermined distance as being close to another has another defect feature connected to the defect feature in the intersecting direction, both of these defect features are grouped as the defect candidate. Since the inclusion process is performed sequentially, even if the defect candidate is interrupted and it appears that there are a plurality of short defect candidates, tracking can be performed without being affected by the short interruption.

According to a seventh aspect of the present invention, in the method for detecting a defect on an object surface according to the fifth aspect, at least two of the plurality of defect candidates obtained by the processing are connected to form a new defect candidate. Since it is assumed that the defect candidate is interrupted and there are a plurality of short defect candidates, it is possible to prevent erroneous recognition that there are two short defects even when the interruption is large.

According to an eighth aspect of the present invention, in the method for detecting a defect on an object surface according to the first aspect, the defect feature amounts of the plurality of defect feature portions included in the defect candidate are the same as those of the plurality of defect feature amounts. Since it is at least one of the statistic and the distance between both ends of the defect candidate, using the length of the defect candidate or the shape feature of the defect feature portion included in the defect candidate, for example, to distinguish between a streak defect and a point defect, It is possible to determine the degree of defect (the width, depth, etc. of the defective portion).

According to a ninth aspect of the present invention, in the method for detecting a defect on an object surface according to the first aspect, the defect reliability in the defect reliability evaluation step is determined by the number of defect features included in the defect candidate. Since the value is obtained by dividing by the number of the plurality of linear patterns corresponding to the plurality of defect features included in the defect candidate, the presence / absence of a defect is determined using this value (density of the defect features). This makes it possible to prevent erroneous detection of a defect candidate having many discontinuities that occurs when fine irregularities are continuous as a defect, and prevent a defect candidate from being detected longer than originally due to noise.

According to the tenth aspect of the present invention, the measuring means for irradiating a plurality of linear lights at intervals on the curved surface of the object to be inspected, and the measuring means along the surface by the plurality of linear lights Photographing means for photographing a plurality of linear patterns arranged at an angle, and a plurality of linear patterns obtained by the photographing, through image processing in a linear pattern processing step and a defect detection step, to obtain an image of the inspection object. A defect detection apparatus configured by an image processing unit that detects an uneven defect on a surface, wherein the linear pattern processing step is performed on each of the plurality of linear patterns obtained by the imaging. If the linear pattern is deformed, a defect feature detection step of detecting the deformation as a defect feature, and a defect feature extraction step of extracting at least one defect feature representing the shape of the defect feature are included. , The defect detection process for a plurality of defect features obtained in the linear pattern process,
A defect candidate extraction process of extracting a group of a plurality of defect feature portions connected in a direction intersecting with the direction of the linear pattern as a defect candidate, and a defect feature amount of the plurality of defect feature portions included in the defect candidate. Evaluating the shape of the defect candidate, a defect shape evaluation step of determining at least one of the degree and type of a defect on the surface of the inspection object, and a defect reliability representing the likelihood as a defect from the defect candidate Since the method includes a defect reliability evaluation step of extracting an index and determining whether or not the defect candidate is a defect, the irregularity defect on the object surface can be correctly detected.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a defect detection device for an object surface.

FIG. 2 is a diagram illustrating an example of an optical pattern captured by the TV camera in FIG. 1;

FIG. 3 is a diagram illustrating a state in which a linear pattern according to the surface shape of an inspection target is formed by linear light emitted from the linear light source in FIG. 1;

FIG. 4 is a perspective view of an inspection object as an example.

FIG. 5 is a diagram illustrating an example of an optical pattern obtained for the inspection target in FIG. 4;

FIG. 6 is a diagram showing the linear pattern of the optical pattern of FIG. 5 cut out and extracted, and the length thereof is normalized.

FIG. 7 is a diagram summarizing discrimination criteria for streak-like and point-like defects, very shallow irregularities, foreign matters, and vertical surfaces (non-defective parts).

FIG. 8 is a flowchart showing a processing procedure of a linear pattern processing step and a defect detection step executed by the image processing apparatus.

FIG. 9 is a flowchart of a defect feature detection process and the like.

FIG. 10 is an explanatory diagram of a defect feature portion detection process and a defect feature value extraction process.

FIG. 11 is an explanatory diagram of a tracking process of a defect candidate.

FIG. 12 is a flowchart of a defect candidate extraction process.

FIG. 13 is an explanatory diagram of a defect shape evaluation process and a defect reliability evaluation process.

FIG. 14 is an explanatory diagram of a defect shape evaluation process and a defect reliability evaluation process.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Inspection object 2 Measuring device 3 TV camera 4 Quantization device 5 Storage device 6 Image processing device 21 Conveyor 22 Passage sensor 23 Linear light source 231 Laser diode 232 Cylindrical lens

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Ryosuke Mitaka 1048 Kadoma Kadoma, Osaka Prefecture Matsushita Electric Works, Ltd. (72) Inventor Osamu Hirakawa 1048 Kazuma Kadoma, Osaka Pref. Matsushita Electric Works, Ltd. (72) Inventor Hiroshi Hemi 1-23-1, Azumabashi, Sumida-ku, Tokyo Asahi Beer Co., Ltd. AA07 AA21 AA23 AA25 AA49 AA67 BB08 BB15 CC00 FF01 FF02 FF09 FF42 GG06 GG08 GG12 HH05 HH12 HH14 JJ03 JJ19 JJ26 LL08 LL30 MM03 NN02 PP15 QQ03 QQ21 QQ23 QQ14 AQ04 AQ04 AQ04 AQ07 AQ04 CA06 CA12 CA16 CB02 CB06 CB12 CB16 CC04 CH08 DA03 DB02 DB05 DB08 DC03 DC08 DC09

Claims (10)

[Claims] 1. A method of irradiating a plurality of linear lights at intervals on a surface of a test object, which is a curved surface, and forming a plurality of linear patterns arranged at an angle along the surface by the plurality of linear lights. A defect detection method for performing imaging and performing image processing in a linear pattern processing step and a defect detection step from a plurality of linear patterns obtained by the imaging to detect uneven defects on the surface of the inspection object. The linear pattern processing step includes, for each of the plurality of linear patterns obtained by the photographing, a defect characteristic part detecting step of detecting the deformation as a defect characteristic part if the linear pattern is deformed. And a defect feature quantity extraction step of extracting at least one defect feature quantity representing a shape of the defect feature part. The defect detection step includes the step of extracting a plurality of defect feature parts obtained in the linear pattern processing step. versus A defect candidate extraction step of extracting a group of a plurality of defect feature portions connected in a direction intersecting with the direction of the linear pattern as a defect candidate; and a defect feature amount of the plurality of defect feature portions included in the defect candidate. A defect shape evaluation step of evaluating the shape of the defect candidate to determine at least one of a degree and a type of a defect on the surface of the inspection object; and a defect reliability representing a probability as a defect from the defect candidate. A defect reliability evaluation step of extracting whether or not the defect candidate is a defect by extracting the index as a defect candidate. 2. The method for detecting a defect on an object surface according to claim 1, wherein the deformation of the linear pattern in the step of detecting the defect feature portion is a deformation extending in mutually opposite directions with a break. 3. The defect feature value in the defect feature value extraction process is a size of the defect feature portion along at least one of a direction of a linear pattern including the defect feature portion and a direction orthogonal to the direction. 2. The method for detecting defects on an object surface according to claim 1, wherein: 4. The method according to claim 1, wherein the defect feature amount in the defect feature amount extraction step is the brightness of the defect feature portion. 5. A process of the defect candidate extraction step, wherein each of the plurality of defect features intersects with another linear pattern adjacent to or close to a linear pattern including the defect feature. 2. The defect on the object surface according to claim 1, wherein if there is another defect feature in the direction, the defect feature is connected to the defect candidate group. Detection method. 6. For each of the plurality of defect features,
If another linear pattern within a predetermined distance as being close to the linear pattern including the defect characteristic portion has another defect characteristic portion connected to the defect characteristic portion in the intersecting direction, both of these are included. 6. A process for sequentially including the defect feature part in the group as the defect candidate is performed.
The method for detecting a defect on an object surface according to the above description. 7. The method according to claim 5, wherein at least two defect candidates among the plurality of defect candidates obtained in the processing are connected to form a new defect candidate. 8. The defect feature amount of a plurality of defect feature portions included in the defect candidate is at least one of a statistic of the plurality of defect feature amounts and a distance between both ends of the defect candidate. The method for detecting defects on an object surface according to claim 1. 9. The defect reliability in the defect reliability evaluation step is determined by changing the number of defect features included in the defect candidate by a plurality of linear patterns corresponding to the plurality of defect features included in the defect candidate. 2. The method according to claim 1, wherein the value is a value obtained by dividing the number of defects. 10. A measuring means for irradiating a plurality of linear lights on a curved surface of an object to be inspected at intervals, and a plurality of lines arranged at an angle along the surface by the plurality of linear lights. Imaging means for imaging a linear pattern, and detecting, from a plurality of linear patterns obtained by the imaging, an uneven defect on the surface of the inspection object through image processing in a linear pattern processing step and a defect detection step The linear pattern processing step includes, for each of the plurality of linear patterns obtained by the photographing, if the linear pattern is deformed. A defect feature detection step of detecting the deformation as a defect feature, and a defect feature extraction step of extracting at least one defect feature representing a shape of the defect feature. A defect candidate extraction step of extracting, as defect candidates, a group of a plurality of defect feature parts connected in a direction intersecting with the direction of the linear pattern, for a plurality of defect feature parts obtained in the linear pattern processing step; A defect shape evaluation step of evaluating a shape of the defect candidate based on defect feature amounts of a plurality of defect feature portions included in the defect candidate and determining at least one of a degree and a type of a defect on the surface of the inspection object; And a defect reliability evaluation step of extracting an index as defect reliability representing the probability of the defect from the defect candidate and determining whether or not the defect candidate is a defect. Surface defect detection device.