JP2008241650A - Device for inspecting defect - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and a device for detecting defect stably by reducing misinformation, without generating information missing, even when a stain adheres onto a lens. <P>SOLUTION: This device has a defect candidate position 3D coordinate comparison means between frames, in addition to a defect candidate position frame coordinate comparison means between frames. The device also has a defect mask information generation means and a lens stain alarm generation means for determining that a defect candidate is a lens fouling, when a defect candidate having the same 3D coordinates in another frame does not exist relative to the defect candidate, when the same spot is imaged at least twice or more number of times. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a defect inspection apparatus that performs image processing and detects defects.

  Conventionally, it has been generally performed to detect defects on the surface of an object by performing image processing on an image captured by a camera. However, in machining sites, there is a problem that facets, oil, water droplets, etc. adhere to the lens and are mistakenly recognized as defects.

  In Patent Document 1, such dirt adhering to the lens is detected by image processing, and this dirt causes a false report (detected as a defect even though it is not actually a defect) in defect detection. A technique for avoiding this is disclosed. Here, two frames of images continuously captured by a camera that shoots the inspection object are compared, and when a defect is detected at the same position (frame coordinates) in both frames, it is determined that this is lens contamination. Thus, an alarm is issued or a defect detected at this position is interpreted as a false alarm due to lens contamination and masked (not treated as a defect).

Japanese Patent Laid-Open No. 6-148083

  In the method disclosed in Patent Document 1, when a true defect exists in a portion hidden by lens dirt, a missing report (missing the true defect) occurs.

  Also, if there is a true defect in the same frame coordinates by chance in consecutive frames, these are recognized as lens contamination and a missing report occurs. Furthermore, since the defect recognized as the lens contamination is masked after that, the defect reflected in the frame coordinates is not treated as a defect and will be reported as missing.

  In many defect inspection systems, missing information is more problematic than false information, and missing a defect is fatal to the defect inspection system. In detecting lens dirt, it is a problem to reduce false alarms while suppressing an increase in missing reports.

  An object of the present invention is to provide a defect inspection apparatus that can stably detect defects by reducing false alarms without generating missing alarms even when dirt is attached to a lens.

  In order to solve the above-described problems, the present invention provides a defect detection unit that detects a defect candidate from image data for each frame imaged by an imaging unit, a defect candidate storage unit that stores a detected defect candidate, and a defect candidate The defect candidate detected from the image data of the previous frame stored in the storage means and the defect candidate detected from the image data of the current frame are compared in two-dimensional coordinates, the identity of the defect candidate is determined, and the determination result is The defect candidate comparison unit that generates the two-dimensional coordinate comparison result information, the defect candidate stored in the defect candidate storage unit and the defect candidate detected from the image data of the current frame are compared in three-dimensional coordinates, and the defect candidate is identical. 3D defect candidate comparison means for determining the property and generating 3D coordinate comparison result information as a determination result, and based on the 2D coordinate comparison result information and the 3D coordinate comparison result information Is determined to be the contamination of the imaging means, based on the imaging means defect alarm signal generating means for generating and outputting the imaging means defect alarm signal, and the two-dimensional coordinate comparison result information and the three-dimensional coordinate comparison result information When it is determined that the defect candidate is a defect of the inspection object, a defect alarm signal generating unit that generates and outputs a defect alarm signal is provided.

  It is possible to provide a defect inspection apparatus that can detect defects stably by reducing false alarms without generating missing alarms even when dirt is attached to the lens.

  FIG. 2 shows the appearance of an embodiment of the defect inspection apparatus according to the present invention.

  The inspection object is the inner surface of the cylinder 11. The inner surface of the cylinder 11 is photographed while inserting the scope head 12 into the fixed cylinder 11. Control signals and image information relating to movement of the scope head 12 and photographing are exchanged with the PC 13 through the control signal line 14. The movement of the scope head 12 and the timing of photographing are controlled by the PC 13, the captured image is sent to the PC 13, and defect detection and the like are performed by image processing in the PC 13.

  FIG. 3 shows in detail the part other than the PC of the defect inspection apparatus.

  The scope head 12 is fixed to a linear motor movable element 17 and can be accurately moved by a length specified in the front-rear direction with respect to a linear motor rail 18 fixed to a pedestal or the like. Control relating to the movement of the linear motor is performed by the PC 13 via a position control signal. When the linear motor completes the designated operation, the PC can know the completion of the operation by the position control response signal.

  At the tip of the scope head 12, an LED illumination 15 and a digital camera 16 as an image pickup device are attached. The inside of the cylinder can be illuminated with the LED and the inner surface of the cylinder can be photographed with the camera. The shooting timing can be controlled from the PC via a shooting control signal. The captured image information is sent to the PC 13 for image processing.

  The PC 13 can execute a sequence of moving the scope head 12 using these series of control signals, capturing images when the movement is completed, and moving the scope head 12.

  FIG. 4 shows a structure when the tip of the scope head 12 is viewed from the front.

  A digital camera 16 is arranged at the center of the cylinder of the scope head 12, and several LED lights 15 are arranged around the digital camera 16. A diffusion filter, a light shielding plate, or the like may be disposed in front of or around the LED illumination 15 as necessary so that the inside of the cylinder can be illuminated with uniform brightness.

  FIG. 5 shows an example of the tip of the scope head 12. A lens 19 is exposed on the front surface of the digital camera 16.

  FIG. 6 shows another example of the tip of the scope head 12. The lens 19 of the digital camera 16 is not exposed, and a transparent cover 20 is disposed so as to surround the scope head 12.

FIG. 7 shows another example of the tip of the scope head 12. A conical mirror 21 is disposed in front of the lens 19, and an image of the inner surface of the cylinder reflected by the mirror is captured by the digital camera 16.

  The lens dirt targeted by the present invention includes not only dirt adhered to the lens but also all optical systems such as the cover 20 and the mirror 21 when the digital camera 16 takes an image. Also, the dirt represents not only deposits such as oil and chips, but also the entire factors that hinder normal photographing such as scratches on the lens 19. Therefore, oil droplets adhering to the cover 20 and scratches on the mirror 21 are also categories of lens contamination targeted by the present invention.

  FIG. 8 shows a photographing method inside the cylinder in the defect inspection apparatus.

The three-dimensional view shows the cylinder 11 and the scope head 12 as viewed from the side. (A), (b),
(C) is shooting while moving the camera by the length d. Frame images (a ′), (b ′), and (c ′) are images taken at respective positions (a), (b), and (c) in the three-dimensional view. The camera acquires the object captured in the field of view 23 as an image, but the object far from the camera is small and the resolution is low. Therefore, in order to obtain high-resolution information convenient for image processing, it is necessary to move the camera and take a picture in a state close to the camera.

  In this defect inspection apparatus, a cylindrical inner surface having a width d at a certain distance from the camera is used as image information. Since the scope head 12 is sent (moved) by the length d, an image of the entire cylinder inner surface can be acquired by connecting the image information of the cylinder inner surface having the width d.

  In the frame image, as the camera advances (a ′) to (c ′), the contour shape of the hole bottom 24 gradually becomes a circle with a large radius, but the image capture portion of the width d has a constant radius. And is taken as a ring with a width.

  In the plurality of frame images thus captured, only the information of the image capturing portion 22 is cut out and subjected to geometric transformation to generate a developed view. The developed view is an image virtually cut in the feed direction on the inner surface of the cylinder and opened in the circumferential direction. The ring-shaped image capturing portions 22 of each frame image are all converted into strip-shaped images having a width d and a length equal to the circumference of the cylindrical inner surface.

  A method for creating this development is described in detail in Japanese Patent Laid-Open No. 11-66316.

  FIG. 9 shows another imaging method inside the cylinder in the defect inspection apparatus.

  The symbols and the like are the same as in FIG. 8, but FIG. 9 differs from FIG. 8 in that the feed width is not changed from d, whereas the width of the image capturing portion 22 is doubled to 2d. is there. Since the width of the image capturing portion 22 is twice the feed width, duplication occurs in the image capturing portion 22 of (a) and (b) or (b) and (c), and all the points on the inner surface of the cylinder 11 are It is included in both two consecutive frames of the image capturing portion 22.

  When the image capturing portion 22 of the frame image obtained in this way is developed into a strip shape, it becomes a strip having a width of 2d, and when pasting together while shifting the feed width by d, the two strips overlap at all points. Become. The same part of two overlapping strips has the same image information geometrically and overlaps exactly (no image shift). However, due to lighting conditions, the side closer to the camera appears brighter and the far side appears darker, resulting in a brightness difference.

  If the feed width and the capture width are equal as shown in FIG. 8, there is no overlap margin between the strips, and each strip has a brightness difference in the feed direction. Occurs. This horizontal line may adversely affect later image processing. However, when there is an overlap margin between the strips as shown in FIG. 9, a good image free from horizontal lines due to a drop in luminance can be obtained by blending the luminances of the two strips at a blend ratio that continuously changes.

  FIG. 10 illustrates an image acquired when the lens, which is an image pickup unit, is contaminated in this defect inspection apparatus (which means the dirt 26 of the cover here).

The shooting conditions are the same as those in FIG. 9 except that the cover 20 is attached and the cover is dirty. Here, it is assumed that the cover 20 has no optical action. That is, it is not considered that the light is refracted or reflected by the cover 20 and the image is distorted or illumination is reflected.

  The cover stain 26 attached to the cover 20 is difficult to distinguish from a defect on the inner surface of the cylinder in the frame image, and it is impossible to determine whether this is a defect or a lens stain with only one frame image. Therefore, an object that appears in the frame image like a defect is called a defect candidate. Further, if the defect candidate is not a defect but a lens stain, it is considered that a false defect 25 exists on the inner surface of the cylinder overlapping the lens stain. The false defect 25 is shown in white on the three-dimensional view. On the other hand, true defects are shown in black.

FIG. 11 shows the behavior of the false defect 25 when photographing is performed while moving the scope head 12 in the feeding direction.

  Since the relative position of the lens dirt with the camera does not change, the false defect 25 does not move on the frame image due to the movement of the camera. Considering a three-dimensional view, the false defect 25 seems to move forward by the amount sent by the camera.

  It is based on this principle that a defect candidate whose frame coordinate (position in a frame image) does not change in consecutive frames in Patent Document 1 taken up as a conventional technique is determined as lens contamination.

FIG. 12 shows the behavior of the true defect 27 when photographing is performed while moving the scope head 12 in the feeding direction.

  Since the true defect 27 does not move with the camera, the position of the true defect 27 in the three-dimensional view is unchanged. The true defect position in the frame image moves in the direction opposite to the feed direction (that is, outside the screen) when the camera is moved in the feed direction.

  FIG. 13 shows each behavior when photographing is performed while moving the scope head 12 in the feeding direction when the true defect 27 and the false defect 25 overlap.

  In the figure, the portion where the true defect 27 and the false defect 25 overlap is indicated by shading. The fact that the true defect 27 and the false defect 25 overlap each other means that the lens stain and the true defect 27 appear in the same position on the frame image.

  However, when the true defect 27 and the false defect 25 overlap, they are separated by sending a camera. That is, even if they overlap each other in a certain frame, the true defect 27 and the false defect 25 are captured at different positions when the image is taken from another position. The true defect 27 moves in the direction opposite to the feeding direction, and the false defect 25 does not move.

In such a case, by masking the false defect 25 which does not move in the method of Patent Document 1 taken up as the prior art, the overlapped true defect 27 is also masked, and a missing report is generated.

  However, even in such a case, the true defect 27 is captured in a state separated from the false defect 25 in another frame (in this example, the next frame). It is possible to stop.

  FIG. 14 shows a case where two true defects 27 are accidentally moved to the same location in a continuous frame when photographing is performed while moving the scope head 12 in the feeding direction.

  Even in such a case, since there are defect candidates whose frame coordinates (positions in the frame image) do not change in successive frames, the technique of Patent Document 1 masks these defects. Although the frame coordinates are not actually contaminated with a lens, the defect appearing at this position will be masked thereafter, resulting in a missed report.

  Even in such a case, the true defect 27 can be detected without missing it by looking at another frame (in this example, the next frame and the next frame) in which each true defect 27 is reflected.

  FIG. 1 shows a processing configuration and a data flow in the defect inspection apparatus of the present invention.

  Images are taken while shifting the camera 1 as the imaging means on the inner surface of the cylinder as the inspection object, and the image data for each photographed frame is sent to the defect detection means 2.

  The defect detection means 2 performs various image processing on the image data for each frame taken by the camera 1 to detect defect candidates. Specifically, a low pass filter such as a Gaussian filter or a medium filter is applied to remove noise superimposed on the image, and then the edge is emphasized with a Sobel filter or Laplacian filter, and an appropriate threshold value is set. Processing such as binarization, labeling for each connected component, and making a lump (connected component) having a certain size as a defect candidate is common. Various defect detection methods have been proposed, and an appropriate method may be used according to the target image. In particular, the defect detection method is not limited in the present invention. As a result of defect detection, position coordinates of defect candidates, that is, frame coordinates and 3D coordinates (three-dimensional coordinates in a three-dimensional view), and in some cases, defect shape information, that is, size, moment, or local image information of defects, etc. Is included. Some of these pieces of information do not necessarily need to be obtained by the defect detection means 2. For example, 3D coordinates can be calculated later if the camera position information and frame coordinates are known. Further, the size and moment of the defect can be calculated from the local image information of the defect. The designer should decide where to generate such information in consideration of the amount of calculation.

  The defect information storage means 3 for each frame, which is a defect candidate storage means, stores information on defect candidates detected by the defect detection means 2. An example of the storage method is shown in FIG. The defect candidate No. in the table is for identifying defect candidates, and different defect candidates are assigned different numbers. However, since the identity determination between frames is not performed at this stage, the same defect candidate Have different numbers in multiple frames. The frame No. is a number assigned to each frame and identifies a frame in which a defect candidate exists. As described above, the frame coordinate is frame coordinate information in which a defect candidate exists on two-dimensional coordinates, and the 3D coordinate is three-dimensional coordinate information in which a defect candidate exists on three-dimensional coordinates. The defect information is information describing whether the defect candidate is a true defect or a false defect, or a hold if it is not yet determined. In other properties, information on the shape of the defect, information for grouping the same defect candidate across frames, and the like are stored as necessary. The identity of defect candidates is determined by coincidence of 3D coordinates in the case of a true defect, and coincidence of frame coordinates in the case of lens contamination (that is, recognition as the same lens contamination).

  A table in which information related to defects is associated and collected in this way is called a defect table. The defect information storage unit 3 for each frame writes information that can be obtained within a frame among the information stored in the defect table.

  For each defect candidate, the defect candidate position frame coordinate comparison means 4 between the frames, which is a defect candidate comparison means, and each defect candidate of the previous frame stored in the defect information storage means 3 for each frame and the defect candidate detected in the current frame Are compared with the two-dimensional frame coordinates to determine the identity of the defect candidates (determining whether there is one with the same coordinates). As the determination result, two-dimensional coordinate comparison result information is generated and output. When the defect shape information is stored in the defect table, not only the coordinate comparison but also the shape comparison can be performed. The comparison result of this shape is useful for making a more accurate determination when determining a true defect and a false defect later. Specifically, it is possible to determine the identity with higher accuracy by using not only the coordinates but also the shape information when determining the identity in the frame image of the defect candidate (whether it is the same lens stain or not). Become. In addition, information for grouping lens stains on the defect table between frames is written by a defect candidate position frame coordinate comparison unit between frames.

  The defect candidate position 3D coordinate comparison means 7 between the frames, which is a three-dimensional defect candidate comparison means, for each defect candidate, the defect candidates stored in the defect information storage means 3 for each frame and the defect candidates detected in the current frame. Compare three-dimensional 3D coordinates, determine the identity of defect candidates (determine whether there is one with the same three-dimensional coordinates), and generate and output three-dimensional coordinate comparison result information as the determination result To do. Specifically, it is searched whether the defect candidate has already been photographed in another frame and whether it was a defect candidate in the photographed another frame. When the defect shape information is stored in the defect table, not only the coordinate comparison but also the shape comparison can be performed. The comparison result of this shape is useful for making a more accurate determination when determining a true defect and a false defect later. Specifically, when determining the identity in a three-dimensional view of defect candidates (whether or not they are the same defect), it is possible to determine identity with higher accuracy by using not only coordinates but also shape information. . When shape information is used to determine the identity of a defect candidate in a three-dimensional view, it is better to use the shape of a developed view, not the shape of a frame image, that is, the shape that can be seen when facing a defect three-dimensionally. desirable. This is because the shape of the same defect changes depending on where it appears in the frame coordinates. Information for grouping true defects in the defect table across frames is written by the defect candidate position 3D coordinate comparison means 7 between frames.

  The defect mask information generation unit 5 suppresses the generation of defect candidates at positions where the lens contamination is present in the defect detection unit 2 based on the two-dimensional coordinate comparison result information and the three-dimensional coordinate comparison result information output from the two coordinate comparison units. To do. Since the determination of whether or not the lens is dirty is performed with reference to the defect table for each defect candidate, the defect mask information generation means 5 is not necessarily required, but it is possible to determine the defect by providing (fake defect or true defect). The processing of defects) can be reduced. However, this may increase the number of missing reports. As a contrivance to avoid this, the defect mask information generation means 5 does not suppress the generation of defect candidates at the position where the lens contamination is present in the defect detection means 2, but the defect candidates having the same 3D coordinates exist in different frames. In such a case, a method such as determining a defect may be used.

  The lens dirt alarm generation means 6 which is an imaging means defect alarm signal generation means is based on the two-dimensional coordinate comparison result information and the three-dimensional coordinate comparison result information output from the two coordinate comparison means, and the defect candidate is the imaging means. If it is determined that the lens of the camera is dirty, a lens contamination alarm that is an imaging means defect alarm signal is generated. Specifically, when it is determined that the defect candidate is lens dirt, an alarm instruction is issued to the lens dirt alarm means. When it is determined that there is a lens dirt, the lens dirt alarm means 8 that is an imaging means defect alarm means outputs the lens dirt alarm generated by the lens dirt alarm generation means 6 to inform the user of the presence of the lens dirt. . Specifically, it is notified by a buzzer or lighting. Alternatively, a defect candidate that is determined to be lens dirt by displaying a developed image is displayed with a different color or the like so that the user can recognize that the possibility of lens dirt is high. Alternatively, instead of notifying the user, a frame having a defect candidate determined to be lens dirt does not use an image near the lens dirt but creates an unfolded image using only an image of another frame so that the user does not feel the lens dirt. There is also a method. It is also possible to automatically start cleaning of the lens and the like.

  The defect alarm generation means 10 which is a defect alarm signal generation means is based on the two-dimensional coordinate comparison result information and the three-dimensional coordinate comparison result information output from the two coordinate comparison means, and the cylindrical inner surface whose defect candidate is the inspection object. When it is determined that the defect is a defect, a defect alarm signal is generated, and the defect alarm signal generated by the defect alarm generator 10 is output to the defect alarm unit 9 to issue an alarm instruction. The defect alarm means 9 issues various defect alarm signals in order to notify the user of the existence of a defect. Specifically, it is notified by a buzzer or lighting. Alternatively, the defect candidates that are determined as defects by displaying the developed image are displayed in different colors so that the user can recognize that the possibility of the defect is high. It is also possible to automatically eliminate defective products.

  FIG. 16 shows the flow of processing in the defect inspection apparatus of the present invention.

First, the defect detection means 2 detects defect candidates in the frame from the image data for each frame captured by the camera 1. If no defect candidate is detected here, it is determined that the cylinder that is the inspection object has no defect. The following description is based on the assumption that a defect candidate has been detected. The following processing is performed for each defect candidate. For example, loop processing is performed for the number of defect candidates. The “calculation of frame coordinates and 3D coordinates” may be performed by the defect detection means 2, and as described above, the calculation of 3D coordinates may be performed until the 3D coordinates are compared in the subsequent processing.

  “Compare the frame coordinates with each defect candidate in the previous frame” is performed by the defect candidate position frame coordinate comparison means 4 between frames. If the two-dimensional coordinate comparison result information, which is the result of the comparison, is information that there is no defect candidate with the matching frame coordinates in the previous frame, it is determined that the defect candidate is a defect (true) and the cylinder of the inspection object It is determined that the inner surface is defective.

  In the case of information that there is a defect candidate whose two-dimensional coordinate comparison result information matches, the result of performing “search for defect candidates of the same 3D coordinates in different frames” in the defect candidate position 3D coordinate comparison means 7 between frames Thus, if the image has not been shot in another frame, the determination in this frame is suspended and the determination is left to a later frame. When the image has been photographed, if there is a 3D coordinate defect candidate having the same 3D coordinate comparison result information, it is determined that there is a defect inside the cylinder that is the object to be inspected. Judge that there is dirt.

  If there is no defect candidate with the same frame coordinate in the previous frame, it may be better not to determine that it is a defect. This is because, in the first frame on which dirt is attached, there is no defect candidate having the same frame coordinate in the previous frame, so that it is always determined as a defect. At this time, if it is not determined as a defect, the determination may be put on hold. However, in this case, there is a possibility that the number of missed reports increases because the ratio that the lens is not a defect but a lens stain is increased. Which is better depends on how much misinformation can be tolerated. On the other hand, if the lens contamination is strictly determined, it can be determined that the lens contamination is present only when a defect candidate exists in the same frame coordinates in a predetermined N frame instead of two consecutive frames.

  When the image data of the previous frame is not stored in the defect candidate storage means, the comparison cannot be performed, but the defect determination is suspended and the defect inspection is performed after waiting for the next frame.

  In the present invention, since it is possible to determine whether or not there is a defect inside the cylinder in consideration of the defect due to the lens contamination in this way, even if the lens adheres to the lens, it is possible to reduce false alarms without generating a missing notification. Thus, it is possible to provide a defect inspection apparatus that stably detects defects.

  It should be noted that the defect mask information generating unit 5, the lens dirt alarm generating unit 6, and the defect alarm generating unit 10 may select the defect candidate determination method independently. You should choose appropriately according to each role.

  FIGS. 17, 18 and 19 show examples in which even if the take-in width is twice the feed width, it will not work.

  In FIG. 17, true defects are arranged at intervals of d. In FIG. 18, there is a lens stain at a position where the false defects appear to be at the interval d. In FIG. 19, true defects are arranged at intervals of d, and lens dirt exists so as to overlap one of them.

  Any of the above processing will determine that the defect is a true defect and will give a false alarm. Such a case is considered to be rare, but if all the defect candidates are photographed in 3 frames by changing 2d to 3d in FIGS. Is drastically reduced. How many times the take-in width should be set should be determined appropriately according to the required defect candidate determination accuracy.

  In this embodiment, an inspection device for a cylindrical inner surface is taken up. In this case, it is easy to search for defect candidates in another frame in which all the observation sites are photographed N times or more and the 3D coordinates match. However, the same processing is possible even in the case of a defect inspection of a general inspection object, for example, an automobile body as in Patent Document 1.

It is a figure which shows the processing structure and flow of data of one Example of the defect inspection apparatus which concerns on this invention. It is a figure which shows the external appearance of one Example of the defect inspection apparatus which concerns on this invention. It is a figure explaining the detail of one Example of the defect inspection apparatus which concerns on this invention. It is a figure which shows the front of the scope head of the defect inspection apparatus which concerns on this invention. It is a figure which shows one detailed example of the scope head of the defect inspection apparatus which concerns on this invention. It is a figure which shows the other detailed example of the scope head of the defect inspection apparatus which concerns on this invention. It is a figure which shows the other detailed example of the scope head of the defect inspection apparatus which concerns on this invention. It is a figure which shows an example of the imaging | photography method of the defect inspection apparatus which concerns on this invention. It is a figure which shows the other example of the imaging | photography method of the defect inspection apparatus which concerns on this invention. It is a figure which shows an example of an image when there exists lens dirt by the imaging | photography method of the defect inspection apparatus which concerns on this invention. It is a figure explaining the behavior of the false defect in the imaging | photography method of the defect inspection apparatus which concerns on this invention. It is a figure explaining the behavior of the true defect in the imaging | photography method of the defect inspection apparatus which concerns on this invention. It is a figure explaining the behavior when the true defect and the false defect overlap in the imaging method of the defect inspection apparatus according to the present invention. It is a figure explaining the behavior when there are two true defects in the imaging method of the defect inspection apparatus according to the present invention. It is a figure which shows an example of the defect table of the defect inspection apparatus which concerns on this invention. It is a figure which shows an example of the flow of the defect determination process of the defect inspection apparatus which concerns on this invention. It is a figure explaining one Example in which the defect determination of the defect inspection apparatus which concerns on this invention is impossible. It is a figure explaining the other Example which cannot determine the defect of the defect inspection apparatus which concerns on this invention. It is a figure explaining the other Example which cannot determine the defect of the defect inspection apparatus which concerns on this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Camera 2 Defect detection means 3 Defect information storage means for every frame 4 Defect candidate position frame coordinate comparison means between frames 5 Defect mask information generation means 6 Lens dirt warning generation means 7 Defect candidate position 3D coordinate comparison means between frames 8 Lens dirt Alarm means 9 Defect alarm means 10 Defect alarm generation means

Claims (9)

In the defect inspection apparatus for detecting defects by imaging the cylindrical inner surface that is the inspection object while shifting the imaging means,
Defect detection means for detecting defect candidates from image data for each frame imaged by the imaging means;
Defect candidate storage means for storing the detected defect candidates;
The defect candidate detected from the image data of the previous frame stored in the defect candidate storage means is compared with the defect candidate detected from the image data of the current frame in two-dimensional coordinates, and the identity of the defect candidate is determined. , Defect candidate comparison means for generating two-dimensional coordinate comparison result information as a determination result;
The defect candidate stored in the defect candidate storage means and the defect candidate detected from the image data of the current frame are compared in three-dimensional coordinates, the identity of the defect candidate is determined, and a three-dimensional coordinate comparison result is determined as a determination result. 3D defect candidate comparison means for generating information;
An imaging unit defect alarm that generates and outputs an imaging unit defect alarm signal when the defect candidate is determined to be dirt on the imaging unit based on the two-dimensional coordinate comparison result information and the three-dimensional coordinate comparison result information Signal generating means;
Defect alarm signal generation means for generating and outputting a defect alarm signal when it is determined that the defect candidate is a defect of an inspection object based on the two-dimensional coordinate comparison result information and the three-dimensional coordinate comparison result information And a defect inspection apparatus. The defect inspection apparatus according to claim 1,
The defect inspection signal generation unit determines that the inspection target is defective when the information indicating that there is no defect candidate having the same two-dimensional coordinate comparison result information, and generates a defect alarm signal. The defect inspection apparatus according to claim 1,
In the case of information that the two-dimensional coordinate comparison result information has the same defect candidate and the information that the three-dimensional coordinate comparison result information has the same defect candidate, the defect warning signal generation means, A defect inspection apparatus that determines that an inspection object has a defect and generates a defect alarm signal. The defect inspection apparatus according to claim 1,
In the case where the imaging means defect warning signal generating means is information that the two-dimensional coordinate comparison result information has the same defect candidate, and the information that the three-dimensional coordinate comparison result information does not have the same defect candidate. Is a defect inspection apparatus that determines that the imaging means is dirty and generates an imaging means defect alarm signal. The defect inspection apparatus according to claim 1,
The defect candidate storage means includes a defect candidate number for identifying a defect candidate, a frame number, frame coordinate information in which a defect candidate exists on two-dimensional coordinates, and three-dimensional coordinate information in which a defect candidate exists on three-dimensional coordinates. And a defect table in which a defect table in which defect information describing whether a defect candidate is true, false, or pending is described is stored. The defect inspection apparatus according to claim 1,
The defect candidate comparison means compares the defect candidate detected from the image data of the previous frame stored in the defect candidate storage means with the defect candidate detected from the image data of the current frame in two-dimensional coordinates and shape, A defect inspection apparatus that determines the identity of the defect candidate and generates two-dimensional coordinate comparison result information as a determination result. The defect inspection apparatus according to claim 1,
The three-dimensional defect candidate comparison unit compares the defect candidate stored in the defect candidate storage unit with the defect candidate detected from the image data of the current frame in three-dimensional coordinates and shape, and determines the identity of the defect candidate. A defect inspection apparatus that determines and generates three-dimensional coordinate comparison result information as a determination result. The defect inspection apparatus according to claim 1,
In the defect candidate comparison unit and the three-dimensional defect candidate comparison unit, if the image data of the previous frame to be compared is not stored in the defect candidate storage unit, the defect inspection apparatus holds the defect determination and waits until the next frame. The defect inspection apparatus according to claim 1,
A defect inspection apparatus that determines that an inspection object has no defect when the defect detection means cannot detect a defect candidate.

Images