1. A method for detecting defects of a transparent substrate film, comprising:

obtaining a substrate image of a substrate having the transparent substrate film by a processor;

performing image preprocessing on the substrate image through the processor to determine an interested area;

performing, by the processor, image segmentation processing on the region of interest in the substrate image to distinguish a peripheral region and a material region in the region of interest and generate a mask image; and

comparing the mask image with the substrate image through the processor to analyze whether the material pattern in the material area of the substrate image has pattern defects or not and further judge whether the substrate has defects or not.

2. The defect detection method of claim 1, wherein the step of determining the region of interest comprises:

and performing edge detection on the substrate image through the processor to detect the substrate edge in the substrate image, and dividing the background area and the region of interest in the substrate image according to the substrate edge.

3. The defect detection method of claim 1, wherein the image segmentation process performed on the region of interest in the substrate image comprises:

inputting the substrate image marked with the region of interest into a segmentation network module through the processor so that the segmentation network module outputs the mask image.

4. The defect detection method of claim 3, wherein the mask image is a binary image,

wherein the peripheral region in the mask image corresponding to the region of interest has a first value and the material region in the mask image corresponding to the region of interest has a second value.

5. The defect detection method of claim 1, wherein comparing, by the processor, the mask image and the substrate image to analyze whether the material pattern in the material region of the substrate image has the pattern defect comprises:

comparing the mask image with the substrate image by the processor to capture the material pattern from the material region of the substrate image; and

inputting, by the processor, the material pattern to a dense convolution network module to cause the dense convolution network module to output a determination of whether the material pattern has the pattern imperfection.

6. The defect detection method of claim 5, wherein the pattern defect comprises having a non-uniform area or a pattern of impurities in the material pattern.

7. The defect detection method of claim 1, wherein comparing, by the processor, the mask image and the substrate image to analyze whether the material pattern in the material region of the substrate image has the pattern defect comprises:

comparing the mask image with the substrate image by the processor to capture the material pattern from the material region of the substrate image;

performing, by the processor, line segment detection on the material pattern to obtain edge line segments of the material pattern; and

the processor compares a reference edge line segment in a template image corresponding to the substrate image with the edge line segment to determine whether the material pattern has the pattern defect of dislocation defects.

8. The defect detection method of claim 7, wherein comparing the reference edge line segment with the edge line segment to determine whether the material pattern has the pattern defect of the misalignment defect comprises:

calculating, by the processor, a reference distance of the reference edge line segment relative to a reference fiducial icon in the template image;

calculating, by the processor, a distance of the edge segment relative to a reference icon in the substrate image; and

and judging whether the distance difference between the distance and the reference distance is larger than a threshold value through the processor so as to judge whether the material pattern has the dislocation defect.

9. The defect detection method of claim 1, wherein the material pattern corresponds to at least one of at least one layer of titanium dioxide material and at least one layer of dye formed on the substrate.

10. The defect detection method of claim 1, wherein the substrate image is obtained by photographing the substrate placed in the closed housing with a camera through illumination light.

11. A system for detecting defects in a transparent substrate film, comprising:

a storage device for storing a plurality of modules; and

a processor coupled to the storage device and executing the modules to:

obtaining a substrate image of a substrate having the transparent substrate film;

performing image preprocessing on the substrate image to determine an interested area;

performing image segmentation processing on the region of interest in the substrate image to distinguish a peripheral region and a material region in the region of interest and generate a mask image; and

comparing the mask image with the substrate image to analyze whether the material pattern in the material region of the substrate image has a pattern defect, and further determining whether the substrate has a defect.

12. The defect detection system of claim 11, wherein the processor determining the region of interest comprises:

and carrying out edge detection on the substrate image so as to detect the substrate edge in the substrate image, and according to the background area and the region of interest in the substrate image of the substrate edge part.

13. The defect detection system of claim 11, wherein said processor performing said image segmentation process on said region of interest in said substrate image comprises:

inputting the substrate image marked with the region of interest into a segmentation network module so that the segmentation network module outputs the mask image.

14. The defect detection system of claim 11, wherein the mask image is a binary image,

wherein the peripheral region in the mask image corresponding to the region of interest has a first value and the material region in the mask image corresponding to the region of interest has a second value.

15. The defect detection system of claim 11, wherein the processor comparing the mask image to the substrate image to analyze whether the material pattern in the material region of the substrate image has the pattern defect comprises:

comparing the mask image with the substrate image to capture the material pattern from the material region of the substrate image; and

the material pattern is input to a dense convolution network module such that the dense convolution network module outputs a determination of whether the material pattern has the pattern imperfection.

16. The defect detection system of claim 15, wherein the pattern defect comprises having a non-uniform area or impurity pattern in the material pattern.

17. The defect detection system of claim 11, wherein the processor comparing the mask image to the substrate image to analyze whether the material pattern in the material region of the substrate image has the pattern defect comprises:

comparing the mask image with the substrate image to capture the material pattern from the material region of the substrate image;

performing line segment detection on the material pattern to obtain an edge line segment of the material pattern; and

comparing the reference edge line segment in the template image corresponding to the substrate image with the edge line segment to determine whether the material pattern has the pattern defect of dislocation defect.

18. The system of claim 17, wherein the processor compares the reference edge line segment with the edge line segment to determine whether the material pattern has the pattern defect of the misalignment defects comprises:

calculating a reference distance of the reference edge line segment relative to the reference icon according to the reference icon in the template image;

calculating the distance of the edge line segment relative to the reference icon according to the reference icon in the substrate image; and

and judging whether the distance difference between the distance and the reference distance is larger than a threshold value so as to judge whether the material pattern has the dislocation defect.

19. The defect detection system of claim 11, wherein the material pattern corresponds to at least one of at least one layer of titanium dioxide material and at least one layer of dye formed on the substrate.

20. The defect detection system of claim 11, wherein the substrate image is obtained by a camera shooting the substrate placed in the enclosed housing through illumination light.

Background

Dye-Sensitized Solar cells (DSSC) are one of the major development directions in the Solar Cell field at present. However, in the process of manufacturing the dye-sensitized solar cell, at least one titanium dioxide material film and at least one dye film are coated on the transparent conductive substrate of the dye-sensitized solar cell to serve as a conductive layer and a light absorbing layer. In this regard, the production yield of the conductive substrate relates to whether the conductive layer and the light absorbing layer are uniformly coated, have impurities, or are coated in the correct position. However, at present, it is determined manually (with a high error rate) whether the conductive layer and the light absorbing layer on the conductive substrate are coated correctly, which results in high production cost and poor production yield of the dye-sensitized solar cell. In addition, transparent substrates having thin film material patterns on their surfaces, such as display panels and touch panels, have similar problems of thin film yield. In view of this, several embodiments of solutions will be presented below.

Disclosure of Invention

The invention provides a flaw detection method and a system thereof for a transparent substrate film, which can automatically and effectively detect whether the substrate of the transparent substrate film has flaws.

The flaw detection method of the transparent substrate film comprises the following steps: obtaining, by a processor, a substrate image of a substrate having a transparent substrate film; performing image preprocessing on the substrate image through a processor to determine an interested area; performing, by a processor, an image segmentation process on a region of interest in the substrate image to distinguish a peripheral region and a material region in the region of interest and generate a mask image; and comparing the mask image with the substrate image by the processor to analyze whether the material pattern in the material region of the substrate image has a pattern defect, and further determining whether the substrate has a defect.

The flaw detection system of the transparent substrate film comprises a storage device and a processor. The storage device is used for storing a plurality of modules. The processor is coupled with the storage device and executes the plurality of modules to perform the following operations: obtaining a substrate image of a substrate having a transparent substrate film; performing image preprocessing on the substrate image to determine an interested area; performing image segmentation processing on a region of interest in the substrate image to distinguish a peripheral region and a material region in the region of interest and generate a mask image; and comparing the mask image with the substrate image to analyze whether the material pattern in the material region of the substrate image has a pattern defect, and further determining whether the substrate has a defect.

In view of the above, the method and system for detecting defects in a transparent substrate film according to the present invention can perform image processing and image analysis on a substrate image of a substrate having a transparent substrate film, and can effectively determine whether the substrate having a transparent substrate film has defects based on the analysis result of the substrate image.

In order to make the aforementioned and other features and advantages of the invention more comprehensible, embodiments accompanied with figures are described in detail below.

Drawings

FIG. 1 is a schematic diagram of a fault detection system according to an embodiment of the invention.

Fig. 2 is a schematic view of an image of a substrate according to an embodiment of the invention.

FIG. 3 is a schematic view of a substrate image according to an embodiment of the invention.

FIG. 4 is a flowchart illustrating a method for detecting defects according to an embodiment of the present invention.

Fig. 5 is a schematic diagram of a mask image according to an embodiment of the invention.

FIG. 6 is a schematic illustration of a material pattern according to an embodiment of the invention.

FIG. 7 is a schematic view of a substrate image according to another embodiment of the present invention.

Fig. 8 is a schematic diagram of a template image according to an embodiment of the invention.

Fig. 9 is a schematic illustration of a material pattern of another embodiment of the present invention.

Wherein the symbols in the drawings are briefly described as follows:

100: a flaw detection system; 110: a processor; 120: a storage device; 121: an edge detection module; 122: dividing the network module; 123: a dense convolutional network module; 124: a line segment detection module; 200: sealing the box body; 210: a conductive substrate; 220: a camera; 230. 240: a light source; 300: imaging the substrate; 301: a substrate portion; 302: a background portion; 303: a peripheral region; 304: a region of material; 311 to 314, 611 to 614, 811 to 814: a pattern of material; 3111. 3112, 3121, 3122, 3131, 3132, 3141, 3142: an edge line segment; 500: masking the image; 510. 520, the method comprises the following steps: an image area; 601: a non-uniform region; 602-605: an impurity pattern; 800: a template image; 8111. 8112, 8121, 8122, 8131, 8132, 8141, 8142: a reference edge line segment; D1-D8: a distance; K1-K8: a reference distance; P1-P4: a reference icon; PA-PD: a reference fiducial icon; s410, S420, S430, S440: a step of; t1: a first direction; t2: a second direction.

Detailed Description

In order that the present disclosure may be more readily understood, the following specific examples are given as illustrative of the invention which may be practiced in various ways. Further, wherever possible, the same reference numbers are used in the drawings and the description to refer to the same or like parts.

FIG. 1 is a schematic diagram of a fault detection system according to an embodiment of the invention. Referring to fig. 1, the defect detection system 100 includes a processor 110 and a storage device 120. The processor 100 is coupled to a storage device 120. In the embodiment, the storage device 120 may store the edge detection module 121, the segmentation network module 122, the dense convolution network module 123 and the line segment detection module 124, but the invention is not limited thereto. In other embodiments of the present invention, the storage device 120 may further store other modules or algorithms related to image processing and image analysis, and may further be used to store the image data, the image processing result, and the image analysis result according to the embodiments of the present invention. In still other embodiments of the present invention, the storage device 120 may store only the dense convolutional network module 123 or the line segment detection module 124.

In the embodiment, the processor 110 may be, for example, a digital and/or analog Processing circuit, an integrated circuit, or a chip with data Processing capability, such as a Graphics Processing Unit (GPU), a Central Processing Unit (CPU), a Microprocessor (MCU), a Field Programmable Gate Array (FPGA), or the like. The storage device 120 may be a Memory (Memory) and may be used for the processor 110 to access data, so that the processor 110 may perform the methods and operations according to the embodiments of the present invention.

It should be noted that although the following description is made by taking a solar cell substrate (conductive substrate) as an example, those skilled in the art will understand that the present invention is also applicable to other transparent substrates (such as a display panel or a touch panel) having a thin film material pattern (transparent substrate film) on the surface. In the present embodiment, the defect detecting system 100 can be used for detecting defects of a Dye-Sensitized Solar Cell (DSSC), and particularly, analyzing a material pattern on a transparent conductive substrate of the DSSC to automatically determine whether the conductive substrate has defects. In this embodiment, the conductive substrate may include, for example, a transparent substrate and at least one of at least one layer of Titanium dioxide (TiO 2) material and at least one layer of Dye (Dye) coated on the transparent substrate. In this embodiment, the transparent substrate may be glass or plastic material, but the invention is not limited thereto. The titanium dioxide material may be coated on the conductive substrate as a conductive layer, and the dye may then be coated on the conductive layer of the conductive substrate as a light absorbing layer. The extent and shape of the conductive layer and the light absorbing layer may be uniform. Also, the conductive substrate may be sequentially coated with a plurality of conductive layers and a plurality of light absorbing layers, wherein the plurality of conductive layers and the plurality of light absorbing layers are alternately stacked.

It is noted that the defect inspection system 100 of the present invention can perform image inspection on the substrate image of the conductive substrate coated with only a single layer of the electrode layer of the titanium dioxide material, and can also perform image inspection on the electrode layer coated with a single layer of the titanium dioxide material and the substrate image coated with a single layer of the dye. Even more, the defect inspection system 100 of the present invention can perform image inspection on the electrode layer coated with the multi-layered titanium dioxide material and the substrate image coated with the multi-layered dye. The defect detection system 100 of the present invention can detect whether the electrode layer and/or the dye of the titanium dioxide material is completely coated or is coated on the correct substrate position.

Fig. 2 is a schematic view of an image of a substrate according to an embodiment of the invention. FIG. 3 is a schematic view of a substrate image according to an embodiment of the invention. Referring to fig. 1-3, in some embodiments of the invention, the flaw detection system 100 may further include the camera 220 of fig. 2 and the light sources 230, 240, and the processor 110 is coupled to and controls the light sources 230, 240. Alternatively, in other embodiments of the present invention, the fault detection system 100 does not include the camera 220 and light sources 230, 240 of FIG. 2. The defect inspection system 100 receives the image from the camera 220 via the input interface for defect inspection.

In the image capturing process, the conductive substrate 210 of the dye-sensitized solar cell may be placed on a plane in the closed box 200, wherein the plane may be paved with black flannelette or a plane object with a specific color, for example. The closed housing 200 is a dark box and is provided with light sources 230, 240 (the number and location of the light sources are not limited in the present invention). During image capture, the light sources 230, 240 may illuminate toward the conductive substrate 210. In this regard, because the titanium dioxide material is a nearly transparent color and the background is, for example, black lint, the color shift effect caused by the light emitted from the light sources 230 and 240 causes the portion of the titanium dioxide material to appear dark green. In this way, the camera 220 can capture the image 300 of the substrate placed in the closed box 200 by illuminating the conductive substrate 210 with the illumination light, as shown in fig. 3. The substrate image 300 may include a substrate portion 301 and a background portion 302. The substrate portion 301 of the substrate image 300 may include material patterns 311-314 and reference icons P1-P4. It is noted that the areas of the substrate image 300 corresponding to the material patterns 311-314 of the titanium dioxide material may be dark green, and the areas outside the material patterns 311-314 are different colors. The border of the substrate portion 301 of the substrate image 300 may also show the border of the conductive substrate 210.

FIG. 4 is a flowchart illustrating a method for detecting defects according to an embodiment of the present invention. Referring to fig. 1, 3 and 9, the defect detecting system 100 may perform the following defect detecting operations of steps S410 to S440. In step S410, the processor 110 of the defect inspection system 100 may obtain a substrate image of a substrate having a transparent substrate film. In this regard, the following description will be given with reference to an exemplary embodiment of the defect inspection system 100 for obtaining a substrate image 300 of a conductive substrate of a dye-sensitized solar cell as shown in fig. 3. In step S420, the processor 110 may perform image preprocessing on the substrate image to determine a Region of Interest (ROI). In the present embodiment, the processor 110 may execute the edge detection module 121 to perform edge detection on the substrate image 300 to detect a substrate edge (e.g., a glass substrate edge) in the substrate image 300 and distinguish a background region (i.e., the background portion 302 of the substrate image 300) and a region of interest (i.e., the substrate portion 301 of the substrate image 300) in the substrate image 300 according to the substrate edge. The processor 110 may effectively define the area and location of the substrate portion 301 and the background portion 302 of the substrate image 300.

In step S430, the processor 110 may perform an image segmentation process on the region of interest in the substrate image 300 to distinguish the peripheral region 303 and the material region 304 in the region of interest, and generate a mask image 500 as shown in fig. 5. Fig. 5 is a schematic diagram of a mask image according to an embodiment of the invention. In the present embodiment, the processor 110 may execute the segmentation network module 122 to perform image processing on the substrate image 300 and generate a mask image 500 corresponding to the substrate portion 301 of the substrate image 300. The segmentation network module 122 may include an algorithm of a deep learning segmentation network architecture, such as a U-Net model. The segmentation network module 122 can find a material region 304 in the region of interest of the substrate image 300 (i.e., the substrate portion 301 of the substrate image 300), and output a corresponding mask image 500 according to the material region 304. In the present embodiment, the mask image 500 is a binary image. Image region 520 of mask image 500 corresponds to peripheral region 303 of substrate image 300 and has a first value (e.g., a binary value of "0"). Image area 510 of mask image 500 corresponds to material area 304 of substrate image 300 and has a second value (e.g., a binary value of "1").

In step S440, the processor 110 may compare the mask image 500 and the substrate image 300 to analyze whether the material pattern in the material region 304 of the substrate image 300 has a pattern defect, and further determine whether the substrate has a defect. In the present embodiment, the processor 110 may compare the mask image 500 with the substrate image 300 to extract the material patterns 611-614 shown in FIG. 6 from the material region 304 of the substrate image 300. FIG. 6 is a schematic illustration of a material pattern according to an embodiment of the invention. The processor 110 may input the material patterns 611-614 to the dense convolutional network module 123 and/or the line detection module 124 to determine whether the substrate image 300 has an impurity pattern defect. Specifically, the dense convolutional network module 123 may output a determination result whether the material pattern has a pattern defect of a non-uniform pattern and/or an impurity pattern. The line segment detection module 124 may output a determination result of whether the material pattern has a misalignment defect. The dense Convolutional network module 123 may include an algorithm of a Convolutional Neural Network (CNN) architecture, such as a dense wired network (densneet) model. The Line Segment detection module 124 may include an algorithm of a Line Segment Detector (LSD). The defect inspection system 100 of the present invention can determine whether the substrate image 300 has at least one of the following exemplary pattern defects.

For pattern defects having non-uniform patterns and/or impurity patterns, the material patterns 611, 613 as in fig. 6 are uniformly coated patterns. The material pattern 612 has pattern imperfections of non-uniform areas 601, wherein the non-uniform areas 601 may be caused, for example, by non-uniform coating of titanium dioxide material and/or dye. The material pattern 614 has pattern defects of the impurity patterns 602-605, wherein the impurity patterns 602-605 may be caused by, for example, an ash layer or impurities. Thus, the dense convolutional network module 123 may be trained in advance to discriminate or classify uniform patterns, non-uniform patterns, and impurity patterns. Through the operation and classification of the dense convolutional network module 123, the dense convolutional network module 123 may output the material patterns 612, 614 as operation results with pattern defects, and may output the material patterns 611, 613 as operation results without pattern defects. In the present embodiment, the processor 110 may generate or provide defect detection information of the corresponding conductive substrate based on the calculation result of the pattern defect.

Referring to fig. 7, fig. 7 is a schematic diagram of a substrate image according to another embodiment of the invention. The processor 110 may perform line segment detection in the first direction T1 on the material patterns 311-314 (i.e., the material patterns 611-614 of fig. 6 have been identified for comparison with the substrate image 300) in the region of interest 301 of the substrate image 300 to obtain edge line segments 3111, 3112, 3121, 3122, 3131, 3132, 3141, 3142 of the material patterns 311-314. However, the present invention is not limited to the line segment detection in the first direction T1, and the processor 110 may also perform the line segment detection in the second direction T2, or in any direction, on the material patterns 311-314 in the region of interest 301 of the substrate image 300. Next, referring to fig. 8, fig. 8 is a schematic diagram of a template image according to an embodiment of the invention. The template image 800 may be a predetermined reference pattern for the titanium dioxide material and/or dye applied by the manufacturer. In the present embodiment, the processor 110 compares the reference edge line segments 8111, 8112, 8121, 8122, 8131, 8132, 8141, 8142 and the edge line segments 3111, 3112, 3121, 3122, 3131, 3132, 3141, 3142 in the template image 800 corresponding to the substrate image 300 to determine whether the material patterns 311-314 have pattern defects with misalignment defects.

For example, the template image 800 also includes material patterns 811-814 and reference fiducial icons PA-PD, and the material patterns 811-814 have reference edge line segments 8111, 8112, 8121, 8122, 8131, 8132, 8141, 8142. The processor 110 may calculate the reference distances K1K 8 of the reference edge line segments 8111, 8112, 8121, 8122, 8131, 8132, 8141, 8142 relative to the reference icon PA according to the reference icon PA in the template image 800. Processor 110 may calculate distances D1-D8 of edge line segments 3111, 3112, 3121, 3122, 3131, 3132, 3141, 3142 from reference icon P1 based on reference icon P1 (corresponding to reference icon PA) in substrate image 300. Then, the processor 110 can determine whether the distance difference between the distances D1-D8 and the reference distances K1-K8 is greater than the threshold value one by one. And when the distance difference between a certain distance and a corresponding certain reference distance is larger than a threshold value, judging that the corresponding material pattern has dislocation flaws. In addition, the distance and the threshold value of the embodiment may be in units of pixels, meters or other length units, and the invention is not limited thereto.

However, the way of calculating the distance and the reference distance by the processor 110 of the present invention is not limited to the above. The processor 110 may also calculate the distance between the edge line segments of the first direction T1 and/or the second direction T2 of the material patterns 311 to 314 and the corresponding reference icons according to at least one of the reference icons P1 to P4 in the substrate image 300, and calculate the reference distance between the edge line segments of the first direction T1 and/or the second direction T2 of the material patterns 811 to 814 and the corresponding reference icons according to at least one of the reference icons PA to PD in the template image 800.

Taking the material pattern 311 having the dislocation defect as an example, with reference to fig. 9, fig. 9 is a schematic diagram of a material pattern according to another embodiment of the invention. As shown in fig. 9, the material pattern 311 has an offset of the second direction T2, and thus edge line segments of both sides of the material pattern 311 along the first direction T1 do not overlap with the reference edge line segment. In this regard, the distance D1 calculated by the processor 110 has a distance difference from the reference distance K1, and the distance D2 calculated by the processor 110 has a distance difference from the reference distance K2. Therefore, the processor 110 may determine whether the distance difference between the distance D1 and the reference distance K1 is greater than a threshold value, and determine whether the distance difference between the distance D2 and the reference distance K2 is greater than the threshold value. If the distance difference is not greater than the threshold, the offset is within an acceptable range. On the contrary, if the distance difference is larger than the threshold value, the processor 110 will issue an alarm indicating that the material pattern 311 has a misalignment defect.

It is noted that the misalignment defect of the present invention is not limited to that shown in fig. 9, and the processor 110 is configured to align each side of the material pattern 311 independently. Also, the processor 110 may compare each side of the material pattern 311 without being limited to the comparison of the edge line segments in the first direction T1. As shown in fig. 9, the material pattern 311 further has an offset in the first direction T1, so the processor 110 can also compare the edge line segment of the side of the material pattern 311 along the second direction T2 with the reference edge line segment to determine whether the material pattern 311 has a misalignment defect.

In summary, the defect detection method and system for a transparent substrate film of the present invention can perform a fast, accurate and automatic defect detection on a substrate image of a transparent substrate having a thin film material pattern on a surface thereof, so as to effectively determine whether the thin film material pattern in the substrate image is correctly coated. The method and the system for detecting the defects of the transparent substrate film can analyze whether the film material pattern in the substrate image has at least one of the pattern defects of non-uniform patterns and/or impurity patterns and the pattern defects with dislocation defects, and further judge whether the transparent substrate corresponding to the substrate image has the defects.

The above description is only for the preferred embodiment of the present invention, and it is not intended to limit the scope of the present invention, and any person skilled in the art can make further modifications and variations without departing from the spirit and scope of the present invention, therefore, the scope of the present invention should be determined by the claims of the present application.