A comprehensive method and device for image detection of solar cell after coating

Abstract

The present invention discloses a comprehensive method and device for image detection of solar cell sheets after coating. The solar cell sheet to be detected is horizontally transported into the image detection device through the motion track 12 according to the horizontal motion direction 13 for detection. Positioning judgment is performed according to the camera comprehensive detection method. When the sensor a14 senses the product, the light source 3 lights up, and the linear array camera a5 starts to collect images at the same time. When the sensor b15 senses the product, the near-infrared laser 9 light source lights up, and the linear array camera b6 starts to collect images at the same time. The interval between the front and rear of the solar cell sheet product should be greater than the distance between the scanning lines of the linear array camera a5 and the linear array camera b6. The present invention combines appearance defect detection and internal defect detection. In terms of hardware, the appearance image and the internal image of the product are collected simultaneously in the same set of equipment. The collection is performed by a line scanning camera, which can greatly reduce the size of the equipment.

Description

Technical Field

[0001] The present invention relates to the field of industrial visual inspection, and in particular to an inspection device and method for inspecting appearance defects, color classification and internal defects of a solar cell after coating. Background Art

[0002] In order to inspect the quality of solar cells in the intermediate process, especially the cells after coating before screen printing, defect detection is usually carried out after the process.

[0003] Previous detection methods mainly focus on color sorting and surface dirt detection, and cannot detect internal product defects. Alternatively, photoluminescence detection devices are used to detect internal defects alone, which may result in many misjudgments. In addition, internal defects may be repaired through subsequent processes, which will affect the judgment of internal defects. Summary of the invention

[0004] Based on the above-mentioned deficiencies of the prior art, the present invention provides a detection device and method for detecting appearance defects, color classification and internal defects of solar cells after coating. The invention aims to solve the problem of single detection items of existing equipment and the problem of many misjudgments of a single device.

[0005] The technical solution of the present invention is: a method for image detection of solar cells after coating, wherein the solar cell products to be detected are horizontally transported into an image detection device for detection through a motion track 12 in a horizontal motion direction 13, and positioning is determined according to a comprehensive camera detection method. When the sensor a14 senses the product, the light source 3 lights up, and the linear array camera a5 starts to collect images. When the sensor b15 senses the product, the near-infrared laser 9 light source lights up, and the linear array camera b6 starts to collect images. The interval between the front and rear of the solar cell products should be greater than the distance between the scanning lines of the linear array cameras a5 and b6.

[0006] Further, the image detection device includes a solar cell, a light source 3, an infrared blocking lens 4, a linear array camera a5, a linear array camera b6, an infrared lens amplification lens 7, a high-pass filter 8, a near-infrared laser 9, a sensor a14, and a sensor b15;

[0007] The light source 3 and the near-infrared laser 9 are arranged relative to each other at a certain angle inwards, and the line array camera a5 and the line array camera b6 both collect images vertically downwards. An infrared blocking lens 4 is arranged below the line array camera a5, and an infrared amplification lens 7 is arranged below the line array camera b6. A high-pass filter 8 is arranged at the bottom of the infrared amplification lens 7 where the line array camera b6 is located. The light source 3 and the near-infrared laser 9, the line array camera a5 and the line array camera b6 are located above the motion track 12, and the sensor a14 and the sensor b15 are located below the motion track 12; the light source 3 and the line array camera a5 point to the sensor a14 below, and the near-infrared laser 9 and the line array camera b6 point to the sensor b15 below;

[0008] Furthermore, the line array camera b6 is an infrared camera, and the resolution is generally a phase scan camera of 1024*1; the line array camera a5 is a visible light camera, and the resolution is generally 2048*3 or 4096*3.

[0009] The specific process of the image detection method of a solar cell after coating according to the present invention for positioning judgment based on the camera comprehensive detection method is as follows:

[0010] Step 1: Linear camera a5 collects images and transmits them to computer software for processing. First, a unique ID is bound to the image to find the corner positioning points; Linear camera b6 collects photoluminescent images and transmits them to computer software for processing. First, the same ID is bound to the image to find the corner positioning points;

[0011] Step 2: Perform secondary positioning on the image with abnormal positioning;

[0012] Step 3: Line array camera a5 collects images for appearance defect detection and color classification, and line array camera b6 collects photoluminescence images for defect detection;

[0013] Step 4: Calculate the coordinate transformation matrix between the two images by locating the points;

[0014] Step 5: Finally, make a comprehensive judgment on the defective areas detected in the two images, give the detection results, and count the process sections where the defects occur.

[0015] Further, the step 2 comprises the following steps:

[0016] Step 2.1: Determine whether there is a positioning failure;

[0017] Step 2.2: Determine whether there is a lack of plating: that is, when a cell without a coating on the surface flows into the test, the visible light camera image shows that the entire surface of the cell is overexposed, and the infrared camera image shows that the entire surface is too dark. Since the background of the visible light image is white and the background of the infrared image is black, it will lead to failure in finding the corner point. In this case, the product can be directly judged as "reworkable";

[0018] Step 2.3: Determine whether there are black pieces and concentric circles: The visible light camera image is a normal image, and the infrared camera image is a black piece if the entire surface is too dark. The infrared camera image has concentric circles, which is a concentric circle defect. This will cause the infrared image to fail to find the corner point, and the product can be directly judged as "not reworkable";

[0019] Step 2.4: Determine whether there is edge overexposure and dark corners: In the visible light camera image, uneven coating thickness at the edges of the product causes the color to be too light, which will lead to edge overexposure. In the infrared camera image, there are dark corners, that is, the corners are too dark. Both situations will lead to failure in finding corner points or deviations. In this case, it is necessary to find the four sides and fit the straight lines, and use the intersection of the straight lines as the corner point for positioning.

[0020] Further, the step 4 comprises the following steps:

[0021] Step 4.1: Use affine transformation of multi-point positioning between the imaging of line array camera a5 and line array camera b6 to find the sub-pixel coordinates of multiple corner points in the four corners of the solar cell, and set corresponding IDs for the corner points in the two images;

[0022] Step 4.2: Considering the factors of point finding failure, it is necessary to discard the corner points with the same ID at the same time to ensure that the calculated points are all corresponding points;

[0023] Step 4.3: Since there is a conversion error from low pixels to high pixels, the transformation matrix of the camera coordinates of high pixels to low pixels is used for coordinate transformation, and finally the affine transformation matrix of the two images is calculated.

[0024] Further, the step 5 comprises the following steps:

[0025] Step 5.1: Black pieces, concentric circles, and cracks are irreparable defects and cannot be reworked, so no comprehensive judgment is required;

[0026] Step 5.2: For appearance defects, it is necessary to refer to whether there are defects at the same position inside. If there are no defects inside, it can be classified according to the normal appearance defect threshold. If there are internal defects at the same position, it is necessary to judge whether the contrast and area of ​​the internal defects meet the threshold setting, and make a comprehensive judgment to classify. For internal defects, it is also necessary to refer to whether there are defects at the same position. If there are no defects at the same position, it is judged whether the threshold setting is met and classified;

[0027] Step 5.3: The comprehensive algorithm needs to set three types of thresholds for judging three situations, namely, defects in both the appearance and the interior at the same position, defects in the appearance and no defects in the interior, and defects in the appearance and no defects in the interior. The three situations and defect types can also be used to count the process segments where defects occur and make statistics; if there are mechanical defects in the appearance and the interior, it can be counted as a problem with the mechanical part of the coating section, and it is more serious; if there are mechanical defects in the appearance and no defects in the interior, it can be counted as a relatively minor mechanical problem in the mechanical part of the coating section; if there are no defects in the appearance and mechanical defects in the interior, the process segment with problems can be judged and counted according to the defect type and position; if there are process defects in the appearance and the interior, it can be counted as a serious coating abnormality; if there are process defects in the appearance and no defects in the interior, it can be counted as a minor coating process defect; if there are no defects in the appearance and defects in the interior, the process segment with problems can be judged according to the defect type.

[0028] The present invention has the following technical effects:

[0029] The present invention combines appearance defect detection with internal defect detection. In terms of hardware, appearance images and internal images of products are simultaneously collected in the same set of equipment. The collection is performed using a line scan camera, which can greatly reduce the size of the equipment.

[0030] The software matches the coordinates of the two images to achieve comprehensive judgment of defects, and can count the characteristic performance of a defect in two tests, so as to classify defects more finely, reduce the misjudgment rate, and improve the detection accuracy. This can reduce the production cost of the production line, reduce the use of silver paste in the subsequent process, and further reduce the use of human resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a schematic diagram of the device layout according to the present invention;

[0032] Figure 2 This is a flow chart of the image positioning and calibration module;

[0033] Figure 3 This is a flow chart of the defect comprehensive judgment module;

[0034] Figure 4 It is a schematic diagram of the overall process of defect detection algorithm;

[0035] Figure 5 It is a detailed flowchart of the judgment operator in the comprehensive judgment module.

[0036] 1. Solar cell a: product to be tested

[0037] 2. Line light path

[0038] 3. Light source: white line light source

[0039] 4. Infrared blocking lens: In order to prevent the visible light image from being affected by infrared, a lens that blocks infrared light is used.

[0040] 5. Linear scan camera

[0041] 6. Linear scan camera b

[0042] 7. Infrared lens: a lens used to increase the transmittance of near-infrared light.

[0043] 8. High-pass filter: a near-infrared 950nm high-pass filter that blocks visible light

[0044] 9. Near-infrared laser: 808nm near-infrared laser, which uses the irradiation of the laser to make the cell produce photoluminescence effect to obtain photoluminescence images;

[0045] 10. Laser optical path

[0046] 11. Solar cell b: Tested finished product

[0047] 12. Motion track: the track on which the product is carried

[0048] 13. Movement direction: the direction of product movement

[0049] 14. Sensor a: trigger 5, 3 to start collecting images

[0050] 15. Sensor b: trigger 6 and 9 to start collecting images DETAILED DESCRIPTION

[0051] like Figure 1 As shown, the solar cell a1 is transported into the detection mechanism for detection through the motion track 12 in the horizontal motion direction 13, the sensor a14 senses the solar cell a1 product, the light source 3 lights up, and the line array camera a5 starts to collect images, the sensor b15 senses the solar cell b11 product, the near-infrared laser 9 light source lights up, and the line array camera b6 starts to collect images, and the distance between the solar cell a1 and the solar cell b11 products should be greater than the distance between the scanning lines of the line array camera a5 and the line array camera b6.

[0052] like Figure 2As shown, the positioning module: the commonly used infrared camera, i.e., the linear array camera b6, is a phase scan camera with a resolution of generally 1024*1. The visible light camera, i.e., the linear array camera a5, has a resolution of generally 2048*3 and 4096*3. The affine transformation of multi-point positioning is used between the two camera images to find the sub-pixel coordinates of multiple corner points in the four corners of the solar cell, and the corner points in the two images are set to corresponding IDs. Considering the factor of point finding failure, it is necessary to discard the corner points with the same ID at the same time to ensure that the calculated points are all corresponding points, and finally calculate the affine transformation matrix of the two images. Because there is a conversion error from low pixels to high pixels, the transformation matrix of the camera coordinates of high pixels to the camera coordinates of low pixels is used for coordinate transformation. The following describes the special cases of positioning: 1. Missed plating: that is, the cell with no coating on the surface flows into the inspection. The visible light camera image shows that the entire surface of the cell is overexposed, and the infrared camera image shows that the entire surface is too dark. Since the background of the visible light image is white and the background of the infrared image is black, it will cause the failure of finding the corner point. In this case, the product can be directly judged as "reworkable". 2. Black film and concentric circles: The visible light camera image is a normal image, and the infrared camera image is a black film if the entire surface is too dark. The infrared camera image has black concentric circles, which is a concentric circle defect. This situation will cause the infrared image to fail to find the corner point, and the product can be directly judged as "not reworkable". 3. Edge overexposure and dark corners: In the visible light camera image, the uneven coating thickness of the product edge causes the color to be too light, which will cause the edge to be overexposed. In the infrared camera image, there is a dark corner, that is, the corner is too dark. Both situations will cause the corner point to fail or there is a deviation. In this case, it is necessary to find the four sides and fit the straight line, and use the intersection of the straight lines as the corner point for positioning.

[0053] Summary of comprehensive judgment module: Appearance defect classification includes: too thin coating, too thick coating, scratches, suction cup marks, belt marks, finger marks, white spots, rainbow flakes, uneven coating, winding plating, etc.; internal defect classification includes: black flakes, concentric circles, white spots, block shadows, top tooth marks, over-engraving, black spots, scratches, cracks, pitting, friction scratches, belt marks, winding plating, finger marks, water marks, watermarks, medicine marks, dirt, boat scratches, boat marks, etc. Product grades are divided into: good, reworkable, and non-reworkable. Among them, the defects that can be comprehensively judged are: scratches, suction cup marks, belt marks, finger marks, white spots, pitting, black spots, rainbow flakes, etc. Among the internal defects, except for black flakes, concentric circles, cracks, and boat scratches, which do not require comprehensive judgment, the remaining defects all need to be jointly judged.

[0054] like Figure 3As shown, the comprehensive judgment logic is: black pieces, concentric circles, and cracks are irreparable defects and cannot be reworked, so no comprehensive judgment is required. Appearance defects need to refer to whether there are defects at the same position inside. If there are no defects inside, they can be classified according to the normal appearance defect threshold. If there are internal defects at the same position, it is necessary to judge whether the contrast and area of ​​the internal defects meet the threshold setting at the same time, and make a comprehensive judgment for classification. Internal defects also need to refer to whether there are defects at the same position. If there are no defects at the same position, it is judged whether the threshold setting is met and classified. The comprehensive algorithm needs to set three types of thresholds for judging three situations, namely, there are defects in both the appearance and the interior at the same position, there are defects in the appearance and no defects inside, and there are no defects in the appearance and defects inside.

[0055] Through these three situations and defect types, the process segments where defects occur can also be counted and counted. If there are mechanical defects in both the appearance and the interior, it can be counted as a problem with the mechanical part of the coating section, and it is more serious; if there are mechanical defects in the appearance and no defects in the interior, it can be counted as a relatively minor mechanical problem in the mechanical part of the coating section; if there are no defects in the appearance and mechanical defects in the interior, the process segment where the problem occurs can be judged and counted according to the defect type and location. If there are process defects in both the appearance and the interior, it can be counted as a serious coating abnormality; if there are process defects in the appearance and no defects in the interior, it can be counted as a minor coating process defect; if there are no defects in the appearance and defects in the interior, the process segment where the problem occurs can be judged according to the defect type.

[0056] like Figure 4 As shown, the line array camera a5 collects images and transmits them to the computer software for processing. First, a unique ID is bound to the image, then the appearance defect detection and color classification are performed, and the corner positioning points are found. The line array camera b6 collects photoluminescence images and transmits them to the computer for processing. First, the same ID is bound to the image, then the defect detection is performed, and the corner positioning points are found. Through the positioning points, the coordinate transformation matrix between the two images is calculated. Finally, a comprehensive judgment is made on the defect areas detected in the two images, and the detection results are given, and the process sections where the defects are generated are counted.

[0057] like Figure 5 As shown in the figure, the detailed process of judging operators in the comprehensive judgment module

[0058] The algorithm obtains the defect information of two images of the same product obtained by the previous calculation, retrieves the defect types in the two images respectively, and divides the results into two categories, one is the defects that are directly judged without comprehensive judgment, and the other is the defects that need to be judged comprehensively. For defects that do not require comprehensive judgment, the defect classification can be performed according to the set threshold through location information and Blob analysis. For defects that require comprehensive judgment, these defects are further classified according to the location information, one is divided into defects with overlapping parts, and the other is divided into defects with non-overlapping parts. For defects with non-overlapping parts, the defect classification is performed according to the set threshold. For defects with overlapping parts, the defect areas in the area are locally aligned and their contour information is compared. Finally, the defects are classified according to the set threshold in combination with Blob analysis.

Claims

1. A method for detecting an image of a solar cell after coating, characterized in that: The solar cell product to be inspected is horizontally transported to the image inspection device through the motion track (12) in the horizontal motion direction (13) for inspection, and the positioning is determined according to the camera comprehensive detection method. When the sensor a (14) senses the product, the light source (3) is turned on, and the linear array camera a (5) starts to collect images. When the sensor b (15) senses the product, the near-infrared laser (9) light source is turned on, and the linear array camera b (6) starts to collect images. The interval between the front and rear of the solar cell product should be greater than the distance between the scanning lines of the linear array camera a (5) and the linear array camera b (6). The image detection device comprises a solar cell, a light source (3), an infrared blocking lens (4), a line array camera a (5), a line array camera b (6), an infrared lens amplification lens (7), a high-pass filter (8), a near-infrared laser (9), a sensor a (14), and a sensor b (15); The light source (3) and the near-infrared laser (9) are arranged relative to each other at a certain angle inwards, the line array camera a (5) and the line array camera b (6) both collect images vertically downwards, an infrared blocking lens (4) is arranged below the line array camera a (5), an infrared amplification lens (7) is arranged below the line array camera b (6), and a high-pass filter (8) is arranged at the bottom of the infrared amplification lens (7) where the line array camera b (6) is located. The light source (3) and the near-infrared laser (9), the line array camera a (5) and the line array camera b (6) are located above the motion track (12), and the sensor a (14) and the sensor b (15) are located below the motion track (12); the light source (3) and the line array camera a (5) point to the sensor a (14) below, and the near-infrared laser (9) and the line array camera b (6) point to the sensor b (15) below. The specific process of positioning judgment based on the camera comprehensive detection method is as follows: Step 1: Linear camera a (5) acquires an image and transmits it to the computer software for processing. First, a unique ID is bound to the image to find the corner positioning points. Linear camera b (6) acquires a photoluminescent image and transmits it to the computer for processing. First, the same ID is bound to the image to find the corner positioning points. Step 2: Perform secondary positioning on the image with abnormal positioning; Step 3: The line array camera a (5) collects images for appearance defect detection and color classification, and the line array camera b (6) collects photoluminescence images for defect detection; Step 4: Calculate the coordinate transformation matrix between the two images by locating the points; Step 5: Finally, make a comprehensive judgment on the defective areas detected in the two images, give the detection results, and count the process sections where the defects occur.

2. The method for detecting an image of a solar cell after coating according to claim 1, characterized in that: Line array camera b (6) is an infrared camera, a phase scan camera with a resolution of 1024*1; line array camera a (5) is a visible light camera with a resolution of 2048*3 and 4096*3.

3. The method of the image detection device of a solar cell after coating according to claim 1, characterized in that: The step 2 comprises the following steps: Step 2.1: Determine whether there is a positioning failure; Step 2.2: Determine whether there is a lack of plating: that is, when a cell with no coating on the surface flows into the test, the visible light camera image shows that the entire surface of the cell is overexposed, and the infrared camera image shows that the entire surface is too dark. Since the background of the visible light image is white and the background of the infrared image is black, it will lead to failure in finding the corner point. In this case, the product is directly judged as "reworkable"; Step 2.3: Determine whether there are black pieces and concentric circles: The visible light camera image is a normal image, and the infrared camera image is a black piece if the entire surface is too dark. The infrared camera image has concentric circles, which is a concentric circle defect. This will cause the infrared image to fail to find the corner point, and the product will be directly judged as "not reworkable"; Step 2.4: Determine whether there is edge overexposure and dark corners: In the visible light camera image, uneven coating thickness at the edges of the product causes the color to be too light, which will lead to edge overexposure. In the infrared camera image, there are dark corners, that is, the corners are too dark. Both situations will lead to failure in finding corner points or deviations. In this case, it is necessary to find the four sides and fit the straight lines, and use the intersection of the straight lines as the corner point for positioning.

4. The method of the image detection device of a solar cell after coating according to claim 1, characterized in that: The step 4 comprises the following steps: Step 4.1: Use multi-point positioning affine transformation between the imaging of line array camera a (5) and line array camera b (6) to find the sub-pixel coordinates of multiple corner points in the four corners of the solar cell, and set corresponding IDs for the corner points in the two images; Step 4.2: Considering the factors of point finding failure, it is necessary to discard the corner points with the same ID at the same time to ensure that the calculated points are all corresponding points; Step 4.3: Since there is a conversion error from low pixels to high pixels, the transformation matrix of the camera coordinates of high pixels to low pixels is used for coordinate transformation, and finally the affine transformation matrix of the two images is calculated.

5. The method of the image detection device of the solar cell after coating according to claim 1, characterized in that: The step 5 comprises the following steps: Step 5.1: Black pieces, concentric circles, and cracks are irreparable defects and cannot be reworked, so no comprehensive judgment is required; Step 5.2: The appearance defect needs to refer to whether there is a defect at the same position inside. If there is no defect inside, it is classified according to the normal appearance defect threshold. If there is an internal defect at the same position, it is necessary to judge whether the contrast and area of ​​the internal defect meet the threshold setting, and make a comprehensive judgment to classify; Internal defects also need to refer to whether there are defects at the same location. If there are no defects at the same location, it is determined whether the threshold setting is met and classified; Step 5.3: The comprehensive algorithm needs to set three types of thresholds for judging three situations, namely, defects in both the appearance and the interior at the same position, defects in the appearance and no defects in the interior, and defects in the appearance and no defects in the interior; the process sections where the defects occur are counted and statistics are made based on these three situations and defect types; if there are mechanical defects in both the appearance and the interior, it is counted as a problem with the mechanical part of the coating section, and it is relatively serious; If there are mechanical defects on the exterior and no defects on the interior, then there are relatively minor mechanical problems in the mechanical part of the statistical coating section; If there are no defects in appearance but mechanical defects inside, the process section where the problem occurs is determined and counted based on the defect type and location; If there are process defects on the exterior and interior, they will be counted as serious coating abnormalities; If there are process defects on the exterior but no defects on the interior, it will be counted as a minor coating process defect; If there are no defects on the outside but there are defects inside, the process section where the problem occurs can be determined based on the type of defect.

Images