KR20150127997A - A micro-crack detection method using dynamic programming - Google Patents

Abstract

A micro-crack inspection method includes a scan step, a seam detection step, and an image banalization step. The scan step photographs an inspection object and obtains an image. The seam detection step detects a seam in the image. The image banalization step determines a micro-crack after the seam detection step.

Description

TECHNICAL FIELD [0001] The present invention relates to a micro-crack inspection method based on dynamic programming,

The present invention relates to a micro-crack inspection method based on a dynamic programming method, and more particularly, to a micro-crack inspection method for detecting a micro-crack using a dynamic programming method.

Recently, in industrial sites where mass production is required, automatic quality inspection is carried out using robots. Display, semiconductors, and wafer inspection of solar cells, and many studies have been conducted.

Generally, the manufacturing process of the solar cell is an automated production line composed of an in-line. However, when a micro crack is present in the solar cell, the solar cell may be broken inside the manufacturing facility. In this case, once the production line is stopped, it takes a long time to start again. This leads to a decrease in productivity and an increase in manufacturing cost, so a method of inspecting the solar cell to find micro cracks is needed.

In addition, since the boundary lines of the surface of the polycrystalline solar cell are irregular, it is difficult to apply the conventional inspection method. In the early days, microcrack inspection method using an infrared camera image was utilized. However, since an infrared camera is expensive and an infrared image acquisition speed is slow, an inspection using an industrial camera with a high image acquisition speed is being carried out.

Therefore, an image is obtained through the industrial camera as described above, and a crack is determined through a micro crack inspection method.

Tsai et al. Proposed an anisotropic diffusion model. The anisotropic diffusion model performs a diffusion process of flattening the micro cracks in the image and preserving the image without micro cracks. The pre-diffusion image and the diffused image are then subtracted from the microcrack to be inspected.

However, the anisotropic diffusion model of Tsai et al. Has disadvantages of discontinuously detecting micro cracks occurring in the diagonal direction because only the kernels in the north, south, east, and west directions are applied.

And Ko and Roheem applied the cross - directional kernel and the diagonal kernel in parallel, and improved the inspection method for polycrystalline solar cell by using the dual structure anisotropic diffusion filter. However, when the micro-crack images in the diagonal direction are discontinuously expressed in the image due to the characteristics of the camera lens and the sensor, micro cracks in the diagonal direction are discontinuously detected by the conventional method, There is a problem that productivity is reduced due to an increase in pixel and inspection execution time.

In this regard, Korean Patent Application No. 2013-0049321 discloses a method for detecting an effective microcrack by performing a step of removing a finger pattern, and Korean Patent Application No. 2013-0049320 discloses a method for detecting an effective microcrack based on an anisotropic diffusion model The present invention discloses an invention for detecting micro cracks even with a small number of repetitions by a method of inspecting micro cracks, but it is insufficient to remove all noise and finger patterns.

Therefore, it is necessary to study a new inspection method which can suppress the occurrence of noise and examine and judge the state of the polycrystalline solar cell in real time.

SUMMARY OF THE INVENTION Accordingly, the present invention has been made in view of the above problems, and it is an object of the present invention to provide a micro crack inspection method for inspecting the state of a polycrystalline solar cell by analyzing a surface image of the polycrystalline solar cell.

According to an embodiment of the present invention, a micro crack inspection method includes a scanning step, a detecting step, and an image binarizing step. The scanning step captures an image of the object to be inspected, and the detecting step detects an angle in the image, and the image binarizing step determines a micro-crack after the detecting step.

 In one embodiment, the detecting step includes defining an energy function for numerically representing an image feature of a micro-crack, applying the energy function to all pixels of the image to generate an energy map for the image And deriving the energy map by applying a dynamic programming method to the energy map.

In one embodiment, the angle may be in the form of lines having various shapes and orientations within the image.

In one embodiment, the dynamic programming method may detect a final jump corresponding to a micro-crack among the optimal droplets appearing horizontally and vertically in the image.

In one embodiment, the microcracks in the energy function of the dynamic programming scheme exhibit low energy values, and regions without noise or cracks may exhibit high energy values.

The micro-crack inspection method using dynamic programming and quadruple detection works on the low-level image features of micro cracks as well as the geometric characteristics and connectivity of micro cracks to enhance the detection power of micro cracks and suppress the isolated noise And the like.

In addition, the micro-crack inspection method has a faster detection time, lower image processing frequency and higher micro-crack detection rate than the conventional ADMCD (Anisotropic diffusion micro-crack detection) and IADMCD (Improved Anisotropic diffusion micro-crack detection) It has the effect of increasing the speed and accuracy of the mass production process and reducing the cost.

1A is an image of a surface showing a surface of a polycrystalline solar cell without cracks.
1B is an image of a surface showing a surface of a polycrystalline solar cell having microcracks formed diagonally.
2A is an image of a surface of a polycrystalline solar cell having a micro crack in a diagonal direction.
2B is an image of a surface of a polycrystalline solar cell having a microcrack formed by an anisotropic diffusion micro-crack detection (ADMCD).
2C is an image showing a surface of a microcracked polycrystalline solar cell detected by IADMCD (Improved Anisotropic diffusion micro-crack detection).
2D is an image of an image showing an energy map of the surface of a polycrystalline solar cell having micro cracks.
2E is an image showing a processed image obtained by image processing the energy map of FIG. 2D.
3A to 3D are perspective views showing objects edited by a seam carving.
4 is an image of a surface of a polycrystalline solar cell having micro cracks sensed by a dynamic programming method according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a micro crack inspection method according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings. While the present invention has been described in connection with what is presently considered to be the most practical and preferred embodiment, it is to be understood that the invention is not limited to the disclosed embodiments. It is to be understood, however, that the invention is not intended to be limited to the particular forms disclosed, but on the contrary, is intended to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. Like reference numerals are used for like elements in describing each drawing. The terms first, second, etc. may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. The terminology used in this application is used only to describe a specific embodiment and is not intended to limit the invention. The singular expressions include plural expressions unless the context clearly dictates otherwise. In the present application, the term "comprises" or "comprising ", etc. is intended to specify that there is a stated feature, figure, step, operation, component, But do not preclude the presence or addition of one or more other features, integers, steps, operations, components, parts, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries are to be interpreted as having a meaning consistent with the contextual meaning of the related art and are to be interpreted as either ideal or overly formal in the sense of the present application Do not.

 Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1A is a plan view of an image showing the surface of a crack-free polycrystalline solar cell.

Referring to FIG. 1A, the polycrystalline solar cell 100 includes a finger pattern 101, a first surface 102, and a second surface 103.

The finger pattern 101 is horizontally and horizontally formed on the surface of the polycrystalline solar cell 100. The finger pattern 101 is a passage through which electricity generated in the polycrystalline solar cell 100 moves. When the finger pattern 101 is photographed on a scan image, a bright and horizontal image is formed.

Further, the polycrystalline solar cell 100 represents a polycrystalline solar cell of zero defect, which is not cracked.

The surface of the polycrystalline solar cell 100 is formed of a plurality of grains. That is, the surface of the polycrystalline solar cell 100 is expressed by an irregular texture. Due to the heterogeneity of such a polycrystalline solar cell, time and effort are required to inspect and find the microcrack in a polycrystalline solar cell.

First, in the experiment for detecting micro cracks, the surface of a 6-inch polycrystalline solar cell is photographed with an area camera having a resolution of 2048 × 2048.

The K value of Equation 2 and the T value of Equation 6 to be described later are set to 200 and 65, respectively, and the detection results of the present invention are compared with the detection results of ADMCD and IADMCD.

1B is a plan view of an image showing the surface of a polycrystalline solar cell having microcracks formed diagonally.

Referring to FIG. 1B, the polycrystalline solar cell 100 includes micro-cracks 104 formed diagonally.

The micro crack 104 starts from the right center of the polycrystalline solar cell 100 and moves to the upper left corner of the polycrystalline solar cell 100 through the finger pattern 101.

In the case where the micro cracks 104 are diagonally formed as described above, unlike the micro cracks generated in the horizontal direction, the micro-cracks 104 are not connected in the diagonal direction due to the characteristics of the scanning lens and the CCD or CMOS sensor, It is expressed continuously. Therefore, when the micro cracks 104 are expressed discontinuously, there is a disadvantage that the micro cracks are discontinuously detected by the conventional micro crack detection algorithm.

2A is a plan view of an image showing a surface of a polycrystalline solar cell having a micro crack in a diagonal direction.

2A, a polycrystalline solar cell 200 includes a third surface 201, a fourth surface 202, a finger pattern 203, and a micro-crack 204.

The micro crack 204 is formed from the upper left corner of the polycrystalline solar cell 200 to the lower right via the third surface 201 and the fourth surface 202 and the finger pattern 203.

2B is a plan view of an image showing the surface of a microcracked polycrystalline solar cell detected by ADMCD (anisotropic diffusion micro-crack detection).

Referring to FIG. 2B, the ADMCD image 300 includes a finger pattern image 301, a micro crack image 302, and a surface image 303.

The ADMCD image 300 is inspected for defects in polycrystalline solar cells using an anisotropic diffusion model. The method uses the anisotropic diffusion model to flatten the microcracks in the defective image, to perform the diffusion process to preserve the image without microcracks, and to detect the microcracks by subtracting the pre-diffusion defect image and the diffused defect image .

However, the anisotropic diffusion model requires a lot of repetitive diffusion processes and generates false detection results and noise in a portion without micro cracks.

2A and 2B, the ADMCD image 300 shows the detection result of the micro-crack image 302, but the finger pattern image (the image of the finger pattern 203) 301 and the third and fourth surfaces 201, 202 also generate noise images.

2C is a plan view of an image showing the surface of a microcracked polycrystalline solar cell detected by IADMCD (Improved Anisotropic diffusion micro-crack detection).

Referring to FIG. 2C, an IADMCD image 400, a noise 401, a micro crack image 402, and a surface 403 are included.

The IADMCD image 400 shows less noise than the ADMCD image 300 of FIG. 2B in a manner of applying a cross-directional kernel and a diagonal direction kernel in parallel using a double-structured anisotropic diffusion filter. However, the computational load is rapidly increased and improved, but still noise is seen.

 That is, the discontinuity of the micro-crack image 402, the time for detecting the IADMCD image 400, the high initial cost for the computer hardware, and the plurality of noises 401 formed on the surface 403 There are still disadvantages that occur.

2D is a plan view of an image showing an energy map of a surface of a polycrystalline solar cell having micro cracks formed thereon.

Referring to FIG. 2D, the energy map 500 includes a first surface 501, a second surface 502, a micro crack image 503, and a finger pattern image 504.

The energy map 500 shows the distribution of materials that are divided according to different crystal directions, materials, and interfaces in the image segmentation as an energy distribution according to the frequency.

Referring to FIG. 2D, since the first and second surfaces 501 and 502 have different crystal orientations and interfaces, they exhibit different energy densities. The finger pattern image 504 is a metallic material such as Ag and exhibits a different energy density from silicon.

On the other hand, the micro-crack image 503 generated on the surface of the silicon solar cell exhibits a low energy density due to the absence of space formed therebetween, and the first and second surfaces 501 and 502, The energy density is different from that of the pattern image 504. Therefore, the micro crack image 503 can be detected through the energy map.

FIG. 2E is a plan view showing a processed image obtained by image processing the energy map of FIG. 2D. FIG.

Referring to FIG. 2E, the processed image 600 includes a micro-crack image 601 and noise 602.

Referring to FIGS. 2D and 2E, the micro-crack image 503 exhibits a remarkably low energy value with materials having different energy densities around the micro-crack image 503, and is formed of the micro-crack image 601 after image processing. However, there is a disadvantage in that the noise 602 is generated during the processing of the image through the energy map.

Thus, in order to detect the noise 602 and the micro-crack image 601 having high sharpness, it is possible to selectively remove low-priority images such as the noise 602 and the finger pattern image 301 of FIG. An image processing method called Seam carving can be used.

3A to 3D are perspective views showing objects edited by a seam carving.

In Fig. 3A, an image 700, a space 701, and a window 702 are displayed.

The images 700 are arranged with the space 701 at a certain distance. The slits 702 may be vertical or transverse. The slits 702 may be formed from the top to bottom or from left to right, and may be defined as connecting lines of pixels forming an image. And, the polynomial 702 can be calculated by dynamic programming.

On the other hand, the dynamic programming is a method of storing the results of calculations to simplify more complex results, and is used to derive the slope 702 from the seam carving. The slope 702 calculates the energy, density, or weight of each of the images 700 and the spaces 701, that is, the pixels, to derive the result. Various algorithms, slope size analysis, entropy, And the slope 702 is formed by a method such as visual salience.

When the energy of the images 700 and the spaces 701 is calculated, the spots 702 are formed in the images 700 or the spaces 701, The images 700 and the spaces 701 are connected in energy order, and the beams 702 having a shape of a line having a low energy have low importance. That is, the rods 702 having a low energy are preferentially formed in the spaces 701 and removed preferentially.

3A and 3B, when the sled 702 is formed from top to bottom, when the sled 702 having a low importance is removed, the sizes of the images 700 are maintained and the images 700 are formed so as to be adjacent to each other on the left and right sides.

Accordingly, when the spaces 701 of the low importance image are formed by forming the spots 702, the images 700 having a relatively high importance are maintained in a perfect shape, Only the space 701 which is a set of consecutively formed images is removed. Of course, since the importance can be switched according to the setting of the user, the images 700 can be arbitrarily set as an image with low importance.

Referring to FIGS. 3A and 3C, an image in which the images 700 are formed upside down is shown. That is, the images 700 are formed upside and down when the claws 702 are formed between the spaces 701 located between the upper and lower portions of the images 700.

Referring to FIGS. 3A and 3D, when the image 702 is formed and deleted in the spaces 701 between the images 700, as described above, Respectively.

Therefore, if the energy of all the pixels of the images is calculated through the seam carving and the dynamic programming, and the claws 702 represented by low-energy connection lines having low importance are first removed, Images 700 are maintained. Also, the user can preferentially remove the images 700 by switching the criterion of importance, and set the respective images or the set of pixels to arbitrarily delete a portion of the specific position.

The present invention utilizes the seam carving as described above and the dynamic programming method that finds the slope in the dynamic programming method as a method of finding a micro crack image. That is, it is possible to detect the micro crack images by discriminating whether the micro crack images are not obtained by detecting the slits on the image in the horizontal and vertical directions, respectively, through the dynamic programming method. Also, the energy function used in the above carving has different characteristics.

4 is a plan view of an image showing the surface of a microcracked polycrystalline solar cell sensed by a dynamic programming method according to an embodiment of the present invention.

Referring to FIG. 4, the detected image 800 includes a micro crack image 801 and a surface image 802.

The detection image 800 is derived from an energy optimization method used in carving as an improved method in MCD (Micro crack detection), and reconfigures a micro-crack detection problem as an energy minimization problem that can be solved through a dynamic programming method .

The detected image 800, which is detected through energy optimization using the dynamic programming method, includes the micro-crack image 801 and the surface image 802 of FIG. 2F.

Referring to FIG. 4, the micro crack image 801 appears clearly, and the space of the image in which the surrounding surface image 802 is formed shows an image significantly reduced compared to the noise 602 in FIG. 2E. Therefore, noise can be reduced by MCD (Micro crack detection) using the dynamic programming method, and the micro crack image 801 can be clearly detected.

The method of the micro-crack inspection method (10) using the dynamic programming method of the present invention operates not only with the low level image features of the micro cracks but also with the geometrical characteristics of the micro cracks, It is characterized as an element that enhances detection and suppresses isolated noises.

First, an energy optimization method is used to find an optimum value indicating an area where a micro crack is expected as a first step of the micro crack inspection method (10). In particular, an energy function representing the micro-crack having low-level characteristics is defined. The energy function is used to numerically represent an image characteristic of the micro-crack, and the energy function is applied to all pixels of the image to generate an energy map for the image. And the energy map is determined using the dynamic programming method with the minimum total energy.

However, since the cracks are formed even without the micro cracks, it is finally determined whether or not the cracks are micro cracks through image binarization. The surface of the polycrystalline solar cell is photographed using an area camera to perform the micro-crack inspection (10).

First, the process of the micro crack inspection method (10) for finding the optimum chip from the following equations can be calculated.

I represents a gray image of M × N size, and ∇I represents the slope size of the gray image.

[Equation 1]

The energy function at the position of (m x n) coordinates can be defined as follows.

&Quot; (2) "

K is a regularization parameter, the microcrack exhibits a low energy value, and regions without noise or cracks exhibit a high energy value. The energy function is different from the energy function used in conventional carving.

The distance to detect the micro crack in the vertical direction can be calculated by the following equation.

&Quot; (3) "

을 씸(seam) s*의 값으로 한다.expression

Is the value of 씸 (seam) s *.

&Quot; (4) "

The above can be found using the following dynamic programming method, wherein the first step traverses from the top to the bottom of the scanned image and determines the coordinates of the points (i, j) Calculate the minimum energy M accumulated to find.

&Quot; (5) "

(5)

(5)

The minimum value of M in the last line in the above equation (5) means the end of the arrow pointing in the vertical direction. Therefore, in the second stage, the trajectory is traced back from the M to find the following route.

On the other hand, the dots may be formed to cross the upper, the lower, the left, and the right of the image, and may be formed from the top of the image to the right or left end, or may be formed up to the center of the image. Therefore, in order to more accurately find various kinds of micro cracks, the following paths are varied. SV * and Sh * determined vertically and horizontally are found, and the energy map Is subjected to image processing in the following formula, the micro crack is detected.

&Quot; (7) "

(6)

(6)

[Table 1]

Table 1 shows the detection rate (%) of the micro crack inspection methods (10) of the present invention and the number of image processing times Iterations) and detection time (Time).

Referring to the detection rate, the microcrack inspection method 10 is 96.29%, which is 3.7% higher than 92.59% of the ADMCD and 11.11% higher than the IADMCD.

Referring to the image processing count, the micro-crack inspection method 10 to which the dynamic programming method is applied represents two processing processes in the vertical and horizontal directions, and requires a lower number of processes than the ADMCD and the IADMCD.

Referring to the detection time, the microcrack inspection method 10 represents 1.22 sec, and the detection speed is significantly faster than the rest of the ADMCD and the IADMCD.

Therefore, the micro-crack inspection method 10 has a higher detection rate than the ADMCD and the IADMCD in terms of the detection rate of the micro cracks at a lower number of processing steps.

According to the embodiments of the present invention as described above, the micro crack inspection method 10 using the dynamic programming method not only has a low level image feature of the micro crack but also has a connectivity with the geometric characteristic of the micro crack Thereby enhancing the detection capability of the micro cracks and suppressing the isolated noise.

In addition, the micro-crack inspection method 10 shows an anisotropic diffusion micro-crack detection (ADMCD), an improved anisotropic diffusion micro-crack detection (IADMCD), a fast detection time, a low image processing frequency, It is possible to increase the speed and accuracy of the mass production process of the polycrystalline solar cell, thereby reducing the cost.

Then, by using the above-mentioned dynamic programming method for the above-mentioned carving, which is not for the purpose of finding out a part in the image of the carving purpose and removing the unnecessary part, the carving is found on both sides of the image, As a method for finding the above-mentioned 씸.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the present invention as defined by the following claims. It can be understood that it is possible.

The microcrack inspection method for inspecting microcracks of a polycrystalline solar cell according to the present invention has industrial applicability that can be used in a factory for mass production and inspection of polycrystalline solar cells

10: Micro crack inspection method 100: Polycrystalline solar cell
101: finger pattern 102: first surface
103: second surface 104: micro crack
200: polycrystalline solar cell 201: third surface
202: fourth surface 203: finger pattern
204: Micro Crack 300: ADMCD video
301: finger pattern image 302: micro crack image
303: Surface 400: IADMCD image
401: noise 402: micro crack image
403: Surface 500: Energy map
501: first surface 502: second surface
503: micro crack image 504: finger pattern image
600: processed image 601: micro crack image
602: Noise 700: Image
701: Space 702: Seam
800: Detected image 801: Micro-crack image
802: Surface image

Claims (5)

A scanning step of photographing an object to be inspected to obtain an image;
Detecting a slope of the image; And
And an image binarization step of determining micro cracks after the detecting step. 2. The method according to claim 1,
Defining an energy function for numerically representing image features of a micro-crack;
Applying the energy function to all pixels of the image to generate an energy map for the image;
And applying a dynamic programming method to the energy map to derive an optimal slope. 3. The method according to claim 2,
Wherein a shape of a line having various shapes and orientations is formed inside the image. 3. The method of claim 2,
Wherein a final crack corresponding to a micro crack is detected among the optimum cracks appearing laterally and laterally in the image. 3. The method of claim 2,
Wherein the micro cracks in the energy function of the dynamic programming method exhibit a low energy value and the regions without noise and cracks exhibit a high energy value.

Images