JP2018128261A - Inspection apparatus and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an inspection apparatus and method capable of reducing over-detection other than cracks and capable of efficiently detecting cracks.SOLUTION: An inspection apparatus includes: illuminating means 2 for illuminating an object to be inspected O; image pickup means 3 for capturing the object to be inspected O; corner detecting means 35 for detecting a corner from crack candidates constituted by consecutive pixels by using a captured image obtained by the image pickup means 3; and determination means 36 which determines whether the crack candidate is a crack or not on the basis of the number of corners being detected.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a technique for inspecting a defect appearing in an appearance.

  Various types of inspection devices have been developed that detect defects such as cracks (fissures, tears) that occur in fine patterns formed on the surface of semiconductor substrates, metal substrates, tape carriers, and the like. As one of crack inspection methods, a method of performing image processing using an image taken from a camera is used. In an inspection method using image processing, generally, only the defective portion is extracted using image processing by utilizing the fact that the degree of reflection of illumination light is different between the defective portion and other portions (for example, patents). Reference 1).

JP 2005-114671 A

  In the inspection of cracks (fissures and fissures) using image processing, a technique is generally used in which a dark part is detected as a crack using the fact that a crack is dark and a normal part appears bright. However, there is a problem in that defects other than cracks that appear dark are overdetected.

  FIG. 6 is a diagram showing a crack detection method in Patent Document 1. In FIG. FIG. 6A shows a method for calculating the linearity. In FIG. 6A, a solid broken line indicates the inspection object. When detecting whether or not the inspection object is a crack, first, a total length S1 that is an actual length is obtained. In the case of a polygonal line as shown in FIG. 6A, the total length S1 is the total length of the straight line segments constituting the line. Further, the total length T1 is obtained. The full length is a distance between one end point of the inspection target and the other end point, as indicated by a broken line. Then, the linearity is obtained by dividing the total length T1 by the total length S1. When this linearity is equal to or greater than the threshold value, it is determined as a crack.

  FIG. 6B is a diagram schematically showing cracks, and FIG. 6C is a diagram schematically showing other than cracks such as scratches. 6B and 6C, the solid line indicates the inspection object (total length S1), and the broken line indicates the total length T1. In the method of Patent Document 1, it is determined that the scratches and the like as shown in FIG. 6C are not cracks because the linearity St is highly likely to be less than the threshold value. However, a crack that is relatively straight as shown in FIG. 6B is also highly likely to have a linearity St that is less than the threshold value, and is determined not to be a crack. On the other hand, if the threshold value is changed to a small value in order to detect a crack as shown in FIG. 6B, there is a problem that a scratch or the like is detected as a crack.

  Therefore, an object of the present invention is to provide an inspection apparatus and method capable of reducing cracks other than cracks and detecting cracks efficiently.

In order to solve the above problems, in the present invention,
Illumination means for illuminating the inspection object;
Imaging means for imaging the inspection object;
Corner detection means for detecting a corner from a crack candidate composed of continuous pixels, using a captured image obtained by the imaging means;
Determining means for determining whether the crack candidate is a crack based on the number of detected corners;
There is provided an inspection apparatus characterized by comprising:

  According to the present invention, there is provided an illuminating unit that illuminates the inspection object and an imaging unit that images the inspection object, and the crack is configured by continuous pixels using the captured image obtained by the imaging unit. Since corners are detected from candidates, and whether or not the crack candidates are cracks is determined based on the number of detected corners, the number of corners, which is a feature with few scratches and many cracks Therefore, it is possible to reduce the overdetection other than cracks and detect cracks efficiently.

  In addition, the present invention is characterized in that the determination means determines that a crack is present when the number of corners per unit length of the crack candidate is equal to or greater than a threshold value. According to the present invention, when the number of corners per unit length of the crack candidate is equal to or greater than the threshold value, it is determined that it is a crack, so it is possible to detect the crack more efficiently. .

The present invention also provides:
Noise removing means for removing noise from the captured image obtained by the imaging means;
Binarization means for obtaining a binary image obtained by binarizing the captured image after noise removal;
Based on the number of pixels in the predetermined luminance range in the captured image after noise removal, narrowing means for specifying the target area,
The binary image further includes thinning means for performing thinning processing on the identified target area;
The corner detection means detects a corner from a crack candidate obtained by the thinning process.

    According to the present invention, noise is removed from a captured image, a binary image obtained by binarizing the captured image after noise removal is obtained, and based on the number of pixels in a predetermined luminance range in the captured image after noise removal, Since the region is specified and the thinning process is performed on the specified target region in the binary image and the corner is detected from the crack candidate obtained by the thinning process, the crack can be detected more efficiently. Is possible.

In the present invention,
Noise removing means for removing noise from the captured image obtained by the imaging means;
Binarization means for obtaining a binary image obtained by binarizing the captured image after noise removal;
Narrowing means for specifying the target area based on the number of pixels in the predetermined luminance range in the captured image after noise removal,
In the binary image, thinning means for performing thinning processing on the identified target area,
Corner detection means for detecting a corner from a crack candidate obtained by the thinning process;
Determining means for determining whether or not the crack candidate is a crack based on the number of detected corners;
A program for operating a computer is provided.

  According to the present invention, by incorporating in a computer, the computer detects corners from crack candidates composed of continuous pixels using the captured image obtained by the imaging unit, and based on the number of detected corners. In addition, since it is determined whether or not the crack candidate is a crack, it is possible to reduce over-detection other than the crack and efficiently detect the crack.

In the present invention,
A computer removing noise from a captured image obtained by the imaging means;
A computer obtains a binary image obtained by binarizing the captured image after noise removal;
A computer specifying a target region based on the number of pixels in a predetermined luminance range in a captured image after noise removal;
The computer performing a thinning process on the identified target area in the binary image;
A computer detecting a corner from a crack candidate obtained by the thinning process;
A computer determining, based on the number of detected corners, whether the crack candidate is a crack; and
There is provided an inspection method characterized by comprising:

  According to the present invention, a computer detects a corner from a crack candidate composed of continuous pixels using a captured image obtained by an imaging unit, and the crack candidate is cracked based on the number of detected corners. Therefore, it is possible to detect cracks efficiently by reducing over-detection other than cracks.

  According to the present invention, excessive detection other than cracks can be reduced, and cracks can be detected efficiently.

It is a figure showing the composition of the inspection device concerning one embodiment of the present invention. It is a figure which shows the hardware constitutions of the image processing apparatus which comprises an inspection apparatus. It is a functional block diagram of the image processing apparatus which comprises an inspection apparatus. It is a flowchart which shows the processing operation of the test | inspection apparatus which concerns on one Embodiment of this invention. It is a figure which shows an example other than the crack used as a detection target, and a crack. It is a figure which shows the detection method of the crack in the past.

<1. Device configuration>
DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments of the invention will be described in detail with reference to the drawings.
FIG. 1 is a diagram showing a configuration of an inspection apparatus according to an embodiment of the present invention. In FIG. 1, 1 is a conveying means, 2 is an illuminating means, 3 is an imaging means, 4 is an imaging timing detector, 5 is an image processing device, 6 is a control device, 7 is an operation / display unit, 8 is an alarm device, 9 Is an automatic discharge device. In FIG. 1, the inspection object O made of resin is arranged on the transport unit 1 and is illuminated by the illumination unit 2. In the present embodiment, the inspection object O is assumed to be made of a resin that easily causes cracks. The transfer means 1 is a transfer stage that can move in a uniaxial direction (the depth direction of the drawing), and is realized by various known techniques. In the present embodiment, a flat table on which the inspection object O can be placed is moved by a driving mechanism (not shown) to convey the inspection object. And the illumination means 2 and the imaging means 3 are installed above the predetermined position of the conveyance means 1, and the inspection object O on the conveyance means 1 can be imaged.

  As the illuminating means 2, various known light sources can be used as long as the light that illuminates the inspection object O reaches the imaging means 3. It is preferable to use the illumination. The dome of the illuminating means 2 is provided with a hole on the upper side of the imaging means 3, and the entire inspection object O on the conveying means 1 is included in the imaging range of the imaging means 3 through this hole. ing. That is, the coaxial means that the center of the imaging unit 3, the center of the hole above the illumination unit 2, and the center of the object O are located on an axis extending in the vertical direction (vertical direction in the drawing). Moreover, LED illumination is used as an illumination system. By using the dome-type and coaxial illumination, there is no reflection by the LED, and the object can be imaged by the imaging means 3 located on the same axis. As the image pickup means 3, various known devices using an image pickup device such as a CCD or a CMOS can be used as long as it is possible to pick up an image of the inspection object O and obtain a picked-up image. The image sensor array may be one-dimensional or two-dimensional, but in this embodiment, a line sensor having a one-dimensional image sensor array is used.

  The imaging timing detector 4 is for detecting the imaging timing, and for example, a rotary encoder that detects the rotation of the rotating shaft that rotates in synchronization with the traveling of the conveying means 1 can be used. The image processing device 5 performs image processing on a captured image obtained by capturing the upper surface of the inspection object O imaged by the imaging unit 3 and determines whether or not there is a crack. The control device 6 controls the entire inspection device by controlling the illumination unit 2, the imaging unit 3, the imaging timing detector 4, and the image processing device 5. The control device 6 also controls the operation / display unit 7, the alarm device 8, the automatic discharge device 9, and the like. The operation / display unit 7 operates the control device 6 and outputs a display from the control device 6, and is realized by, for example, a touch panel monitor. The alarm device 8 is for outputting an alarm when it is determined that there is a crack as a result of the inspection, and is realized by, for example, a patrol lamp or a buzzer. The automatic discharge device 9 is for discharging from the line when it is determined that there is a crack as a result of the inspection.

  FIG. 2 is a hardware configuration diagram of the image processing apparatus 5 which is a main part of the inspection apparatus. The image processing device 5 may be any device as long as it can perform image processing on a captured image. In this embodiment, the image processing device 5 is realized by a general-purpose computer such as a personal computer. Yes. As shown in FIG. 2, a CPU (Central Processing Unit) 21, a RAM (Random Access Memory) 22 that is a main memory of the computer, a hard disk for storing programs and data executed by the CPU 21, a flash memory, etc. A device 23, a communication unit 24 for communicating with the outside, and a data input / output I / F (interface) 25 for data communication with an external device such as a data storage medium. Connected via bus. The RAM 22 and the storage device 23 store the captured image received from the imaging unit 3 via the communication unit 24 in the process.

  The control device 6 may be any device as long as it can control a part of the inspection device. In the present embodiment, the control device 6 is realized by a PLC (programmable logic controller). The control device 6 has a hardware configuration similar to that of the image processing device 5 shown in FIG. 2, and can also be realized by incorporating a dedicated program into a general-purpose computer such as a personal computer. Further, the image processing device 5 and the control device 6 can be realized by a single computer.

  FIG. 3 is a functional block diagram showing the configuration of the image processing apparatus 5. In FIG. 3, 31 is noise removing means, 32 is binarizing means, 33 is narrowing means, 34 is thinning means, 35 is corner detecting means, and 36 is judging means. Further, the noise removing means 31, the binarizing means 32, the narrowing means 33, the thinning means 34, the corner detecting means 35, and the judging means 36 read the program into the RAM 22 shown in FIG. It is realized.

  The noise removing unit 31 performs noise removal on the captured image obtained by the imaging unit 3 that is an imaging unit in order to easily narrow down the target area to be inspected for cracks. The binarizing means 32 binarizes the captured image from which noise has been removed. The narrowing-down means 33 narrows down the target area from the binarized image with the number of pixels in a predetermined luminance range. The thinning means 34 performs a thinning process on the target area. The corner detection means 35 detects the corner of the target area after thinning. The determination means 36 determines whether or not it is a crack based on the detected corner.

  Each component shown in FIG. 3 is actually realized by mounting a dedicated program on hardware such as a computer and its peripheral devices as shown in FIG. That is, the computer executes the contents of each means according to a dedicated program. In the present specification, the computer means a device having an arithmetic processing unit such as a CPU and capable of data processing, and is a concept including a PLC for realizing the control device 6.

  In the storage device 23 shown in FIG. 2, a dedicated program for operating the CPU 21 and causing the computer to function as the image processing device 5 is installed. By executing this dedicated program, the CPU 21 realizes functions as a noise removing unit 31, a binarizing unit 32, a narrowing unit 33, a thinning unit 34, a corner detecting unit 35, and a determining unit 36. .

<2. Transport processing and imaging>
Next, the inspection method according to the present embodiment will be described together with the processing operation of the inspection apparatus. First, when the inspection apparatus is activated, the inspection object O is sequentially transferred by the transfer means 1 using a single axis stage. Then, according to the output signal from the imaging timing detector 4, the control device 6 gives an imaging instruction to the imaging means 3. The imaging means 3 images the inspection object O and sequentially transmits it to the image processing device 5. In the image processing device 5, the received one-dimensional line images are integrated for a predetermined number of lines, and acquired as a two-dimensional captured image.

<3. Inspection process>
Next, the inspection method according to the present embodiment will be described together with the processing operation of the inspection apparatus. FIG. 4 is a flowchart showing the processing operation of the inspection apparatus according to the present embodiment. As described above, the line images acquired by the imaging unit 3 are sequentially transmitted to the image processing device 5. The image processing apparatus 5 integrates a predetermined number of line images received from the imaging unit 3 and acquires a captured image (step S1).

In this embodiment, the line sensor which is an imaging means images the target object illuminated using dome illumination. Therefore, in the obtained image, the crack portion is imaged darkly and the luminance value is small, and the portion other than the crack is imaged brightly and the luminance value is large. Although this differs depending on the characteristics of the line sensor used as the image pickup means 3, in this embodiment, since the image is taken with 8 bits per pixel, the value of each pixel of the picked-up image is obtained as a luminance value of 256 gradations. In the present embodiment, a line sensor having 2000 pixels per column and a resolution of 6 μm is used as the imaging unit 3. Then, the image processing apparatus 5 acquires 2000 columns as one captured image. Therefore, one captured image is 2000 × 2000 pixels and images a range of 1.2 cm × 1.2 cm.

Subsequently, the noise removing unit 31 removes noise from the captured image (step S2). This noise removal is performed to facilitate extraction of the target area. Various known methods can be used for noise removal. In the present embodiment, noise is detected and removed by performing expansion / contraction processing. Specifically, a filter having a predetermined size (number of pixels) is prepared, and the expansion process is performed by performing a process of giving each pixel a central pixel and the maximum value of the neighboring pixels of the central pixel. Then, with each pixel as the central pixel, the reduction process is performed by performing a process of giving the central pixel the minimum value of the neighboring pixels of the central pixel. This expansion process and contraction process are repeated. As a result, an image from which noise such as dirt on the resin portion that appears dark is removed is generated. The size of noise that can be removed depends on the size of the expansion filter (reduction filter) and the number of processing times.

If noise removal is performed, the binarization means 32 next performs a process of binarizing the image from which noise has been removed (step S3). Specifically, the value of each pixel having a luminance value of 256 gradations is binarized using a predetermined threshold value. Then, a new value of a pixel whose 256 gradation luminance value is equal to or greater than the threshold is “0” (black), and a new value of a pixel whose 256 gradation luminance value is less than the threshold is “1” (white). ) Is generated. As a result, the bright portion of the captured image becomes dark when the pixel value is “0”, and the dark portion of the captured image becomes bright when the pixel value is “1”. An image is obtained. The size of the obtained binary image is 2000 × 2000 pixels similarly to the captured image. In the following description, for convenience of explanation, a portion where the pixel value is “0” is assumed to be a black pixel and a pixel value “1” is assumed to be a white pixel.

Next, the narrowing-down means 33 narrows down the target area in the captured image after noise removal (step S4). Specifically, the target area is narrowed down based on the number of pixels in a predetermined luminance range in the captured image after noise removal. In this embodiment, the number of pixels in which a pixel having a luminance value less than a predetermined value is continuous is greater than or equal to a predetermined number in a 256-gradation captured image after noise removal in order to set a darker portion as a target region. Narrow down only the points. That is, an area where the area of the darkly reflected part is large to some extent is set as the target area. The predetermined luminance range and the number of pixels in the predetermined luminance range can be appropriately set according to the situation, but in the present embodiment, the predetermined luminance range has a luminance value of less than 50, and the number of pixels is 50 pixels or more. And Therefore, in a 256-gradation captured image after noise removal, a region in which 50 or more pixels having a luminance value of less than 50 are continuous (area is 50 or more) is narrowed down as a target region.

Next, the thinning means 34 performs thinning on the binary image using the target area narrowed down using the picked-up image after noise removal (step S5). As a specific method for thinning, various known methods can be used. In the present embodiment, a certain white pixel is a target pixel, and among 8 neighboring pixels other than the target pixel in the 3 × 3 pixels, the predetermined pixel is a white pixel and the predetermined pixel on the opposite side across the target pixel is a black pixel. In this case, the target pixel is replaced with a black pixel. That is, the pixel value of the target pixel is replaced from “1” to “0”. In this way, the line composed of white pixels is thinned by narrowing the width of the portion where the white pixels are continuous. In this embodiment, thinning is performed so that the number of pixels in the width direction of the white pixel line becomes one pixel. A set of continuous white pixels obtained by thinning becomes a crack candidate.

Next, the corner detection means 35 performs corner detection in the binary image in which the portion where the white pixels are continuous is thinned (step S6). A corner is a location where the local direction changes in a shape in which white pixels after thinning are continuously formed. Corner detection is performed by detecting a portion in which the angle in the direction of successive pixels is greatly different in a crack candidate that is a set of thinned white pixels. As a specific method of corner detection, various known methods can be used. In the present embodiment, each white pixel after thinning is realized by applying a corner detection filter to a local region centered on the white pixel. For example, a SUSAN filter can be applied as the corner detection filter. The SUSAN filter obtains an absolute difference between the white pixel that is the target pixel and its peripheral pixels, and the difference absolute value is less than or equal to a threshold value among pixels existing in a substantially circular mask centered on the target pixel. A certain pixel is counted. In the present embodiment, since the images are binary images, the same white pixels are counted. Then, when the number of white pixels that are the SUSAN filter output values is equal to or less than a certain ratio with respect to the number of all pixels in the mask, it is determined that the white pixel as the target pixel is a corner.

If it is determined whether or not all white pixels are corners, then the determination unit 36 determines whether or not the number of corners per unit length is a certain number or more for each crack candidate (step) S7). Specifically, first, the number of crack candidates that are a set of thinned white pixels is counted, and the detected number of corners is divided by the counted number of pixels. As a result, the number of corners per unit length is calculated for the crack candidate. And it compares with the threshold value set beforehand. The threshold value is influenced by the characteristics of the inspection object and the resolution of the image. Therefore, the corresponding threshold value is set in consideration of the resolution of the line sensor. For example, when the inspection object is a resin, it is preferable to determine that it is a crack when there are three or more corners per 10 mm. For example, when the resolution of the imaging means is 6 μm, the length corresponding to 10 mm is about 1667 pixels. In this case, 3/1667 is set as a threshold value and used for determination in step S7.

  FIG. 5 is a diagram illustrating an example of a crack and a flaw other than a crack. FIG. 5A shows a crack, and FIG. 5B shows something other than a crack such as a scratch. The cracks and scratches shown in FIGS. 5 (a) and 5 (b) are the same as those shown in FIGS. 6 (b) and 6 (c), respectively. In the above description, the scratched part is processed with white pixels. However, in FIG. 5, the scratched part is shown in black in conjunction with FIG. 6, and white and black are opposite to the above description. Yes. In the case of a scratch as shown in FIG. 5A, there are three corners as surrounded by a broken-line circle. As described above, when there are a large number of corners, the corners are detected in step S6, and in step S7, the number of corners is considered to be a certain number or more in relation to the length of the scratch, and it is determined as a crack. In the case of FIG. 5B, no corner exists. Thus, when there is no corner, in step S7, the number of corners may be less than a certain number in relation to the length of the scratch, and it is determined that there is no crack. That is, according to the present invention, it is possible to distinguish between a crack and a scratch that were difficult to distinguish with the conventional method.

  In step S7, when the number of corners per unit length is a certain number or more, the determination means 36 determines that there are cracks (step S8), and the number of corners per unit length is less than the certain number. The determination means 36 determines that it is other than a crack (step S9).

  The preferred embodiments of the present invention have been described above. However, the present invention is not limited to the above embodiments, and various modifications can be made. For example, in the above-described embodiment, a line sensor in which image pickup elements are arranged one-dimensionally is used as the image pickup unit, and several lines are integrated to obtain one picked-up image. However, the image pickup elements are arranged in two dimensions. Alternatively, one captured image may be obtained by one imaging.

DESCRIPTION OF SYMBOLS 1 ... Conveyance means 2 ... Illumination means 3 ... Imaging means 4 ... Imaging timing detector 5 ... Image processing apparatus 6 ... Control apparatus 7 ... Operation / display part 8 ...・ Alarm device 9 ... Automatic discharge device 21 ... CPU (Central Processing Unit)
22 ... RAM (Random Access Memory)
23 ... Storage device 24 ... Communication unit 25 ... Data input / output I / F
31 ... Noise removing means 32 ... Binarization means 33 ... Narrowing means 34 ... Thinning means 35 ... Corner detection means 36 ... Determination means

Claims (5)

Illumination means for illuminating the inspection object;
Imaging means for imaging the inspection object;
Corner detection means for detecting a corner from a crack candidate composed of continuous pixels, using a captured image obtained by the imaging means;
Determining means for determining whether the crack candidate is a crack based on the number of detected corners;
An inspection apparatus comprising:   The inspection apparatus according to claim 1, wherein the determination unit determines that a crack is present when the number of corners per unit length of the crack candidate is equal to or greater than a threshold value. Noise removing means for removing noise from the captured image obtained by the imaging means;
Binarization means for obtaining a binary image obtained by binarizing the captured image after noise removal;
Based on the number of pixels in the predetermined luminance range in the captured image after noise removal, narrowing means for specifying the target area,
The binary image further includes thinning means for performing thinning processing on the identified target area;
The inspection apparatus according to claim 1, wherein the corner detection unit detects a corner from a crack candidate obtained by the thinning process. Noise removing means for removing noise from the captured image obtained by the imaging means;
Binarization means for obtaining a binary image obtained by binarizing the captured image after noise removal;
Narrowing means for specifying the target area based on the number of pixels in the predetermined luminance range in the captured image after noise removal,
In the binary image, thinning means for performing thinning processing on the identified target area,
Corner detection means for detecting a corner from a crack candidate obtained by the thinning process;
Determining means for determining whether or not the crack candidate is a crack based on the number of detected corners;
As a program to make the computer function as. A computer removing noise from a captured image obtained by the imaging means;
A computer obtains a binary image obtained by binarizing the captured image after noise removal;
A computer specifying a target region based on the number of pixels in a predetermined luminance range in a captured image after noise removal;
The computer performing a thinning process on the identified target area in the binary image;
A computer detecting a corner from a crack candidate obtained by the thinning process;
A computer determining, based on the number of detected corners, whether the crack candidate is a crack; and
An inspection method characterized by comprising:

Images