CN106952260B - Solar cell defect detection system and method based on CIS image acquisition - Google Patents

Abstract

The invention discloses a solar cell defect detection system and method based on a CIS image acquisition unit, comprising the following steps: the system comprises a transmission device, an image acquisition device and a data processing module; the image acquisition device is positioned right above the conveying device and is connected with the data processing module; the solar cell is transported through the conveying device, when the solar cell passes through the lower part of the image acquisition device, the image acquisition device acquires images of the solar cell and transmits the images to the data processing module, and the data processing module performs defect analysis and detection on the acquired images. The invention carries out image acquisition through the image acquisition device, improves the real-time performance and reduces the cost, and the defects of the solar cell detected by the invention comprise: the defects of unfilled corners, broken edges, holes and dirt are various, and the detection efficiency is ensured. The invention adopts the CIS image acquisition system to acquire images, can improve the image acquisition rate, enhance the real-time property and reduce the cost.

Description

A solar cell defect detection system and method based on CIS image acquisition

technical field

The invention belongs to the technical field of solar cells, and more particularly, relates to a solar cell defect detection system and method based on CIS image acquisition.

Background technique

As a free, clean, safe and abundant renewable energy, solar energy is more and more favored by people. As a device that converts solar radiation into photothermoelectricity or photoelectric direct conversion, solar cells are increasingly popular in manufacturing and testing.

As the core device of a solar cell, the quality of its manufacturing process directly affects the power generation efficiency, short-circuit current, open-circuit voltage, and service life of the solar cell. The manufacturing process needs to go through two parts: cutting and texturing, and then chemical etching and cleaning to remove the surface damage layer. Because solar cells are fragile, in the production process (surface corrosion, texturing, diffusion, surface film formation, screen printing, passivation, sintering, etc.), solar cells may be caused by certain process defects or the influence of the production environment. surface defects. Therefore, it is necessary to develop a solar cell inspection system and method with good real-time performance, high identification efficiency, and identification of various defects.

SUMMARY OF THE INVENTION

In view of the defects of the prior art, the purpose of the present invention is to provide a solar cell defect detection system based on CIS image acquisition, which aims to solve the detection problems of missing corners, chipped edges, holes and dirt of solar cells.

The invention provides a solar cell defect detection system based on a CIS image acquisition unit, comprising: a transmission device, an image acquisition device and a data processing module; the image acquisition device is located just above the transmission device, and the image acquisition device is connected with the data processing module ; The solar cell is transported through the conveying device, when the solar cell passes under the image acquisition device, the image acquisition device collects the image of the solar cell and transmits it to the data processing module. Image for defect analysis and detection.

Further, the image acquisition device includes: a CIS image acquisition unit and two strip light sources, the CIS image acquisition unit is arranged directly above the conveying device, and a strip light source is installed on each side to make the collected solar cells The image brightness of the film is uniform.

Further, the CIS image acquisition unit includes: an FPGA module, a CIS image acquisition module, an analog-to-digital conversion module, a storage module, an acquisition speed matching module and an image output module respectively connected to the FPGA module, and a power supply module for providing working power. The CIS image acquisition module is used for image acquisition of solar cells and outputs a series of analog data; the analog-to-digital conversion module is used to convert the analog data into digital signals; the FPGA module is used to realize the timing sequence of the entire system control; the acquisition speed matching module is used to adjust the image acquisition rate of the CIS image module according to the transmission speed of the transmission device; the storage module is used to buffer the collected digital image data; the image output module is used to collect the collected solar energy The cell slice image is sent to the data processing module.

The present invention also provides a solar cell defect detection method based on a CIS image acquisition unit, comprising the following steps:

(1) Image acquisition of solar cells by an image acquisition device;

(2) correcting the solar cell according to the collected image to obtain a corrected solar cell image;

(3) A defect detection result is obtained after the image of the solar cell is subjected to corner-cutting processing or edge chipping processing.

Further, after step (2), the method further includes: (2.0) obtaining a standardized grayscale image of the solar cell after performing the process of removing electrodes and grid lines on the solar cell image. Wherein, step (2.0) may be after step (2) and before step (3), or may be after step (3).

Further, the process of removing electrodes and grid lines is specifically:

(2.01) Grayscale processing is performed on the corrected image of the solar cell to obtain a grayscale image of the solar cell;

(2.02) On the grayscale image, use the morphological opening operation to remove the horizontal grid lines to obtain the grayscale image of the solar cell after removing the horizontal grid lines;

(2.03) On the grayscale image of the solar cell after removing the horizontal grid line, use the maximum inter-class variance method to binarize the solar cell after removing the horizontal grid line, use the horizontal scanning to locate the electrode and remove the electrode, and obtain Grayscale image of solar cell with horizontal grid lines and electrodes removed;

(2.04) On the grayscale image of the solar cell with the horizontal grid lines and electrodes removed, use horizontal scanning to locate the vertical grid lines and boundary lines and remove them to obtain a standardized grayscale image of the solar cell.

Further, after step (2.0), the method further includes: (2.1) obtaining a defect detection result after performing hole processing or dirt processing on the standardized grayscale image of the solar cell.

Further, the hole processing is specifically:

(2.11) On the standardized grayscale image of solar cells, use the method of finding contours to find the positions of holes and dirt.

(2.12) Using the position found in (2.11), calculate the average gray value of the position area, compare it with the set threshold, determine whether the defect is a hole or a dirt, and count the number of both.

Further, the waste treatment is specifically:

(2.11) On the standardized grayscale image of solar cells, use the method of finding contours to find the positions of holes and dirt.

(2.12) Using the position found in (2.11), calculate the average gray value of the position area, compare it with the set threshold, determine whether the defect is a hole or a dirt, and count the number of both.

Further, step (2) is specifically:

In S21, grayscale processing is performed on the collected solar cell image to obtain a solar cell grayscale image;

S22 obtains the pixel points representing the four boundaries of the solar cell sheet by searching on the grayscale image; wherein, searching is performed in the order of the upper boundary, the lower boundary, the left boundary and the right boundary;

S23 screen the pixel points of the four boundaries and obtain the four correction boundaries of the solar cell;

S24 calculates the intersection points between the four correction boundaries, that is, the four corner points of the solar cell;

S25 obtains a perspective transformation matrix according to the four corner points of the solar cell and the set four corner points, and performs perspective transformation on the collected image through the perspective transformation matrix to obtain a corrected image of the solar cell.

Further, the processing of missing corners described in step (3) is specifically:

(3.11) Grayscale processing is performed on the corrected image of the solar cell to obtain a grayscale image of the solar cell;

(3.12) On the grayscale image, use three step search methods to detect whether the upper left corner of the solar cell is missing corners from the horizontal to the right and the vertical to the downward;

(3.13) On the grayscale image, use three search methods of step size to detect whether the upper right corner of the solar cell is missing from the horizontal to the left and the vertical to the downward;

(3.14) On the grayscale image, use three search methods of step size to detect whether the lower left corner of the solar cell is missing from the horizontal to the right and the vertical upward;

(3.15) On the grayscale image, three step search methods are used to detect whether the lower right corner of the solar cell is missing from the horizontal to the left and the vertical to the top.

Further, the edge collapse processing described in the step (3) is specifically:

(3.11) Grayscale processing is performed on the corrected image of the solar cell to obtain a grayscale image of the solar cell;

(3.12) On the grayscale image, project the left and right boundaries, and record the number of pixels in the edge collapse area;

(3.13) On the grayscale image, project the upper and lower boundaries, and record the number of pixels in the edge collapse area;

(3.14) Compare the number of pixels in the edge collapse area with the set threshold, determine whether it is edge collapse, and count the number of edge collapses;

(3.15) Considering that the four corners of the solar cell are missing corners, after determining the starting point and the ending point of the continuous boundary of the four sides, repeat steps (3.12)-(3.14).

Through the above technical solutions conceived by the present invention, compared with the prior art, since the CIS image acquisition system is used for image acquisition, the rate of image acquisition can be increased, the real-time performance can be enhanced, and the cost can be reduced. The defect detection method can meet the detection of various defects, has high versatility, and has a high detection accuracy under the condition of ensuring the detection rate, which well meets the needs of the current industrial inspection field.

Description of drawings

1 is a schematic diagram of a solar cell defect detection system based on CIS image acquisition of the present invention;

Fig. 2 is the control flow schematic diagram of the solar cell defect detection method based on CIS image acquisition of the present invention;

Fig. 3 is the principle block diagram of CIS image acquisition unit in the present invention;

4 is a schematic flowchart of a calibration method for a solar cell in the present invention;

5 is a schematic flowchart of a method for detecting a missing angle of a solar cell in the present invention;

6 is a schematic flowchart of a method for detecting edge collapse of a solar cell in the present invention;

7 is a schematic flowchart of a method for detecting holes and dirt in a solar cell according to the present invention;

FIG. 8 is a schematic diagram of scanning a detection template composed of three pixel points in the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

The invention relates to a solar cell defect detection system and method based on a CIS (Contact Image Sensor, contact image sensor) image acquisition system. As shown in FIG. 1, the solar cell defect detection system based on the CIS image acquisition unit provided by the present invention includes: a transmission device 1, an image acquisition device 2 and a data processing module 3; the image acquisition device 2 is located directly above the transmission device 1, At the same time, the image acquisition device 2 is connected with the data processing module 3 through a data line. The solar cell is transported by the conveying device 1. When the solar cell passes under the image acquisition device 2, the image acquisition device 2 captures the image of the solar cell, and then transmits the image to the data processing module 3, and the data processing module 3 monitors the image. Carry out defect analysis and detection.

The image acquisition device 2 includes a CIS image acquisition unit 21 and two uniform strip light sources 22; a CIS image acquisition unit 21 is installed in the upper center of the image acquisition device, and a uniform strip light source is installed on both sides of the device. As shown in FIG. 2 , the image acquisition device 21 includes: a CIS image acquisition module, an analog-to-digital conversion module, an FPGA module (Field-Programmable Gate Array, Field Programmable Gate Array), a storage module, a power supply module, an acquisition speed matching module, and an image output module . When the system is working, the CIS image acquisition module collects images of the solar cells passing below and outputs a series of analog data, which are converted into digital signals by the analog-to-digital converter. The FPGA is responsible for the timing control of the entire system, and the acquisition speed matching module is based on the transmission device. The transmission speed of the CIS image module adjusts the image acquisition rate of the CIS image module. The FPGA caches the collected digital image data in the storage module, and then sends the collected solar cell image to the data processing module through the image output module. The power module supplies power to the entire system. , requires an external 5V power supply. The uniform strip light source is a strip light source with the same illumination intensity at each light-emitting point. The purpose is to make the brightness of the solar cell image uniform, so as to avoid the inconsistency of the brightness of each part of the image, which will affect the subsequent detection results.

In the present invention, the image acquisition device is used for image acquisition, which improves the real-time performance and reduces the cost. The defects of the solar cell detected by the system and the method include: missing corners, chipped edges, holes, and dirt. There are many types of defects detected and the detection is guaranteed. efficiency.

In the solar cell defect detection system based on the CIS image acquisition unit provided by the present invention, the color image of the solar cell is obtained through the scanning of the CIS image acquisition unit, and the color image is subjected to grayscale processing. Then, four boundary searches are performed on the grayscale image to determine the boundary intersection (corner point), and the corrected image of the solar cell is obtained by changing the perspective. Then, scan the four corners of the solar cell to determine whether it is missing corners and count the number of missing corners. Then, the four sides of the solar cell are scanned respectively to determine whether the edges are chipped and the number of chipped edges is counted. Then, the horizontal grid lines, vertical grid lines, electrodes, and boundary lines of the solar cell are removed by morphological methods. Finally, holes and dirt are found by the contour finding method, and the holes and dirt are judged according to the gray value and counted respectively. .

The present invention also provides a solar cell defect detection method based on the CIS image acquisition unit, the specific steps are as follows:

(1) Image acquisition of solar cells by an image acquisition device

The image acquisition device collects images through the CIS image acquisition unit. The CIS image acquisition unit scans to obtain several lines of image information (the number of lines in each scan is set according to the speed of the transmission equipment and the resolution of the collected image), thereby obtaining Image of a solar cell.

(2) Correction of solar cells

In step S21, grayscale processing is performed on the collected solar cell image to obtain a solar cell grayscale image.

S22 obtains the pixel points representing the four boundaries of the solar cell sheet by searching on the grayscale image; wherein, searching is performed in the order of the upper boundary, the lower boundary, the left boundary and the right boundary.

S23 screen the pixel points of the four boundaries and obtain the four correction boundaries of the solar cell;

S24 calculates the intersection points between the four correction boundaries, that is, the four corner points of the solar cell.

S25 obtains a perspective transformation matrix according to the four corner points of the solar cell and the set four corner points, and performs perspective transformation on the collected image through the perspective transformation matrix to obtain a corrected image of the solar cell.

(3) Detection of missing corners of solar cells

In S31 , grayscale processing is performed on the image of the corrected image of the solar cell to obtain a grayscale image of the solar cell.

S32 On the grayscale image, three step search methods are used to detect whether the upper left corner of the solar cell is missing corners from the horizontal to the right and the vertical to the downward.

S33 On the grayscale image, three step search methods are used to detect whether the upper right corner of the solar cell is missing corners from the horizontal to the left and the vertical to the downward.

S34 On the grayscale image, three step search methods are used to detect whether the lower left corner of the solar cell is missing corners from the horizontal to the right and the vertical upward.

S35 On the grayscale image, three step search methods are used to detect whether the lower right corner of the solar cell is missing corners from the horizontal to the left and the vertical to the upward.

(4) Detection of solar cell chipping

In S41, grayscale processing is performed on the image of the corrected image of the solar cell to obtain a grayscale image of the solar cell.

S42 On the grayscale image, the left and right boundaries are projected, and the number of pixels in the edge-break area is recorded respectively.

S43, on the grayscale image, project the upper and lower boundaries, and record the number of pixels in the edge chipping area respectively.

S44 compares the number of pixels in the edge-break area with the set threshold, determines whether it is edge-break, and counts the number of edge-breaks.

S45 Considering that the four corners of the solar cell sheet are missing corners, after determining the starting point and the ending point of the continuous boundary of the four sides, the above steps S42-S44 are repeated.

(5) Detection of holes and dirt in solar cells

In S51 , grayscale processing is performed on the image of the corrected image of the solar cell to obtain a grayscale image of the solar cell.

S52, on the grayscale image, removes the horizontal grid lines by using the morphological opening operation, and obtains a grayscale image of the solar cell after removing the horizontal grid lines.

S53 On the grayscale image of the solar cell after removing the horizontal grid lines, the maximum inter-class variance method (OSTU algorithm) is used to binarize the solar cells after removing the horizontal grid lines, and the electrodes are positioned and removed by horizontal scanning. , to obtain a grayscale image of the solar cell with the horizontal grid lines and electrodes removed.

S54 , on the grayscale image of the solar cell with the horizontal gridlines and electrodes removed, use horizontal scanning to locate the vertical gridlines and boundary lines and remove them to obtain a standardized grayscale image of the solar cell.

S55 uses the method of finding contours to find the positions of holes and dirt on the standardized grayscale image of solar cells.

S56 uses the position found in S55 to calculate the average gray value of the position area, compares it with the set threshold, determines whether the defect is a hole or a dirt, and counts the number of both.

The solar cell defect detection system and method based on the CIS image acquisition unit provided by the present invention can at least bring the following beneficial effects: real-time, and reduce costs. At the same time, the defect detection method provided by the present invention can meet the detection of various defects, has high versatility, and under the condition of ensuring the detection rate, the detection accuracy is relatively high, which well meets the needs of the current industrial detection field. .

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the present invention will be specifically described below with reference to the accompanying drawings and embodiments. The drawings in the following description are only some embodiments of the invention. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative effort.

As shown in FIG. 1 , the present invention provides a solar cell defect detection system based on a CIS image acquisition unit. As can be seen from the figure, the system includes: a transmission device 1 , an image acquisition device 2 and a data processing module 3 . The specific process is as follows: the solar cells are transported through the transmission device, and when passing under the device box, the CIS image acquisition tube captures the images of the solar cells, and at the same time transmits the images to the data processing module through USB3.0 for subsequent defect detection. During the image acquisition process, the transmission device transmits the current speed of the transmission device in real time to the CIS image acquisition unit through the cable, so as to adjust the acquisition frequency of the CIS image acquisition tube.

As shown in FIG. 2 , the present invention provides a solar cell defect detection method based on a CIS image acquisition unit. As can be seen from the figure, the present invention includes six processes: correcting solar cells, detecting missing corners of solar cells, detecting solar cell chipping, removing solar cell electrodes and grid lines, and detecting solar cell holes and dirt .

As shown in Figure 4, the specific steps of calibrating the solar cell include:

The solar cells are arbitrarily placed on the conveying device, so that the positions and orientations of the solar cells in the images collected by the CIS image acquisition unit are also arbitrary. In order to process the image more conveniently later, it is necessary to cut out the area of the solar cell in the image and correct it.

S21 reads the solar cell image and performs grayscale processing on it.

S22 searches for the upper, lower, left and right boundaries of the solar cell on the grayscale image respectively.

When searching for four boundaries, the search methods are divided into three methods: large-step search, small-step search, and pixel-by-pixel search. Large-step search is to search according to the specified larger number of lines, usually the specified larger number of lines is an integer multiple of 10, for example: 10 lines; small-step search is to search according to the specified smaller number of lines, usually small steps The number of long lines is half the number of large steps, for example: 5 lines; pixel-by-pixel search means searching by 1 line. The boundary is searched by three different search methods, which can improve the search efficiency and ensure the accuracy of the search.

First search the upper boundary, if the gray value of the starting pixel point of the search is less than the threshold gray value of the boundary (that is, the average gray value of the image), it is judged that the starting pixel point is in the area of the solar cell, and the pixel-by-pixel search is used to search upward until Find the pixel point whose gray value is less than the threshold gray value of the boundary, which is the upper boundary point. If the starting pixel of the search is not in the solar cell area, firstly use a large step to search until the gray value of the pixel is less than the boundary gray threshold, and the search position is already in the solar cell, and then use a small step to search Search upward until the gray value of the pixel point is greater than the threshold gray value of the boundary, at this time the search position is outside the solar cell, and finally use the pixel-by-pixel search to search downward until the pixel point whose gray value is less than the threshold gray value of the boundary is found, which is the upper boundary point. Then search in the horizontal direction to obtain a series of upper boundary points.

Calculate the gradient G(x,y)=dx(i,j)+dy(i,j) for each upper boundary point, where dx(i,j)=I(i+1,j)-I(i, j), dy(i, j)=I(i, j+1)-I(i, j), (i, j) is the coordinate of the pixel, and I is the gray value of the pixel. And arrange these boundary points in ascending order according to the gradient size, select the pixel point with the median value of the gradient as the initial centroid of the K-means clustering algorithm, and then calculate the Euclidean distance between the searched upper boundary point and the centroid, and then compare with the distance threshold ( It is used to filter out the effective upper boundary point, which is usually selected according to the actual required accuracy) for comparison. If it is less than the distance threshold, the point is an effective point. In this way, abnormal upper boundary points (noise points, isolated points) can be removed, and effective upper boundary points can be obtained. Use the valid upper boundary points to fit the straight line by the least square method, and the straight line obtained is the upper boundary.

For the lower boundary, we also use three steps to search, but first search downwards, and then search horizontally to obtain a series of lower boundary points, and then use the K-means clustering algorithm to obtain the effective lower boundary points, and then use the least squares method to fit Combine the lines to get the lower boundary.

For the left boundary, we also use three steps to search, but first search to the left, and then vertically search to obtain a series of left boundary points, and then use the K-means clustering algorithm to obtain the effective left boundary points, and then use the least squares method to fit Combine the lines to get the left boundary.

For the right boundary, we also use three steps to search, but first search to the right, and then search vertically to obtain a series of lower boundary points, and then use the K-means clustering algorithm to obtain the effective right boundary points, and then use the least squares method to fit Combine the lines to get the right boundary.

S23 corrects the four boundaries. Let the equation of the straight line of the four boundaries be y=k ₁ x+b ₁ , y=k ₂ x+b ₂ , y=k ₃ x+b ₃ , y=k ₄ x+b ₄ (k ₁ , k ₂ , k ₃ , k ₄ are the slopes of the upper, lower, left, and right boundaries, and b ₁ , b ₂ , b ₃ , and b ₄ are the intercepts of the upper, lower, left, and right boundaries).

According to the included angle formula tanθ ₁ =(k ₁ -k ₂ )/(1+k ₁ k ₂ ) (θ ₁ is the angle between the upper and lower boundaries), determine the parallelism of the upper and lower boundaries, if θ ₁ is less than the angle threshold (used for Judging the parallelism of the upper and lower boundaries, usually 5°), there is no need to correct the upper and lower boundaries; if θ ₁ is greater than the angle threshold, let km = ( _{k 3} + _{k 4} )/2 (km is the mean value of the slope of the left and right boundaries) , through the verticality of the upper and lower boundaries and the left and right boundaries, namely || _{k 1} km || and || _{k 2} km ||, if || _{k 1} km || is greater than || _{k 2} km ||, then The upper boundary is more perpendicular to the left and right boundaries, then the lower boundary equation is y=k ₁ x+b ₂ ; if || _{k 1} km || is smaller than || _{k 2} km ||, the lower boundary is more perpendicular to the left and right boundaries, The upper bound equation is then y=k ₂ x+b ₁ ; the upper and lower bounds are corrected in this way.

Similarly, the left and right boundaries can be corrected by judging the angle between the left and right boundaries and the perpendicularity to the upper and lower boundaries.

S24 calculates the intersections between the four boundaries, that is, the four corners of the solar cell.

Given any two straight lines y=k ₁ x+b ₁ and y=k ₂ x+b ₂ , and k ₁ ≠0∪k ₂ ≠0, (k ₁ , k ₂ are the slopes of the two straight lines, b ₁ and b ₂ are the intercepts of the two straight lines respectively), then the intersection of the two straight lines is:

x ₀ =(b ₂ -b ₁ )/(k ₂ -k ₁ ), y ₀ =(k ₁ b ₂ -k ₂ b ₁ )/(k ₁ -k ₂ )

In this way, the four corner points of the solar cell are obtained.

S25 uses the calculated four corner points and the corrected four corner points to obtain a perspective transformation matrix, and performs perspective transformation to obtain a corrected image of the solar cell.

The corrected four corner points are fixed, respectively (0, 0), (0, width-1), (height-1, 0), (width-1, height-1) (width, height are respectively Correct the width and height of the image).

(u，v，w为原始图像任意像素横纵垂直坐标、x′，y′，w′为透视变换后图像对应像素点的横纵垂直坐标，

为透视变换矩阵)。由于处理的图像是二维的，所以令w＝1(即垂直坐标W为1)，将变换后(x′，y′，w′)齐次化，即为

令变换后的图像坐标

则有

Perspective transformation is the projection of an image onto a new view, also known as projection mapping. The general transformation formula is:

(u, v, w are the horizontal, vertical and vertical coordinates of any pixel in the original image, x', y', w' are the horizontal, vertical and vertical coordinates of the corresponding pixel in the image after perspective transformation,

is the perspective transformation matrix). Since the processed image is two-dimensional, let w=1 (that is, the vertical coordinate W is 1), and the transformed (x', y', w') is homogeneous, which is

Let the transformed image coordinates

then there are

As shown in Figure 5, the specific steps for detecting the missing angle of the solar cell include:

S31 reads the corrected image of the solar cell, and performs grayscale processing on it.

When detecting missing corners, three search methods in S22 will be used: large-step search, small-step search, and pixel-by-pixel search.

S32 detects whether the upper left corner of the solar cell is missing.

Firstly, the detection template composed of 3 pixel points in Fig. 8 is used for vertical downward scanning detection pixel by pixel. , usually the value is greater than the average gray value of the image), then it is judged that this point is the boundary point A of the missing corner. Then, starting from the boundary point A, scan and detect vertically downward with large steps until the gray value of the pixel point is less than the threshold of missing corners. At this time, the search position is already in the solar cell, and then use small steps to search vertically upward until The gray value of the pixel point is greater than the missing corner threshold. At this time, the search position is outside the solar cell. Finally, the pixel-by-pixel search is used to search vertically downward until the pixel whose gray value is less than the set threshold is found, which is the lower border missing corner boundary. point to get the length of the missing corner. At the same time, starting from the boundary point A, scan and detect horizontally with a large step size until the gray value of the pixel point is less than the threshold of the missing corner. At this time, the search position is already in the solar cell, and then use a small step size to search horizontally to the left until the pixel The gray value of the point is greater than the missing corner threshold. At this time, the search position is outside the solar cell. Finally, the pixel-by-pixel search is used to search horizontally to the right until a pixel whose gray value is less than the missing corner threshold is found, which is the right border missing corner boundary point. , to get the width of the missing corner. Using the obtained length and width of the missing corner, the missing corner can be marked.

S33 detects whether the upper right corner of the solar cell is missing.

Similar to the method of detecting missing corners in S32, at this time, the detection template is scanned horizontally to the left from the upper right corner, and then scanned vertically downward, as shown in Figure 8.

S34 detects whether the lower left corner of the solar cell is missing.

Similar to the method of detecting missing corners in S32, at this time, the detection template is scanned horizontally to the right from the lower left corner, and then scanned vertically upward, as shown in Figure 8.

S35 detects whether the lower right corner of the solar cell is missing.

Similar to the method of detecting missing corners in S32, at this time, the detection template is scanned horizontally to the left from the lower right corner, and then scanned vertically upward, as shown in Figure 8.

As shown in Figure 6, the specific steps for detecting the edge collapse of the solar cell include:

S41 reads the corrected image of the solar cell, and performs grayscale processing on it.

S42 performs projection on the left and right boundaries, and counts the number of pixels in the edge collapse area respectively.

S43 projects the upper and lower boundaries, and counts the number of pixel points in the edge collapse area respectively.

S44 compares the number of pixels in the edge collapse area with the edge collapse threshold (this threshold is used to determine the edge collapse boundary point, usually the value is greater than the average gray level of the image), determine whether it is edge collapse, and count the edge collapse number.

S45, considering that the four corners of the solar cell are missing corners, determine the starting point and the ending point of the continuous boundary of the four sides. Project the upper, lower, left, and right boundaries, and count the number of pixels in the collapsed edge area respectively. Compare the number of pixels in the edge collapse area with the edge collapse threshold (this threshold is used to determine whether the edge collapses, the value is set according to the actual accuracy, usually 100), determine whether it is edge collapse, and count the number of edge collapses.

By projecting the boundary, we can intuitively and effectively judge whether the boundary is collapsed through the number of pixels after projection. When the boundary collapses, the number of pixels after projection will increase significantly, and then the abnormal area (that is, the area with increased pixels in the histogram) can be determined according to the statistical histogram, and the number of pixels in the abnormal area is counted, and then Compare with the edge collapse threshold to determine whether the abnormal area is an edge collapse area. Through the above practices, the detection speed can be accelerated to a certain extent, the detection time can be saved, and the detection efficiency can be improved.

As shown in Figure 7, the specific steps for detecting solar cell holes and dirt include:

S51 reads the corrected image of the solar cell, and performs grayscale processing on it.

S52 utilizes the morphological opening operation to remove the horizontal grid lines.

The morphological opening operation is to first perform the erosion operation on the image, and then perform the dilation operation. Erosion is the process of eliminating all boundary points of an object, resulting in the remaining object being one pixel smaller than the original along its inner edge. The dilation operation is the process of merging all background points in contact with an object into that object.

After removing the grid lines by morphology, it is convenient to detect the holes and dirt on the solar surface later, so as to avoid the grid lines interfering with the subsequent detection, thereby affecting the detection accuracy. After the grid lines are removed, the holes and dirt can be detected more conveniently.

S53 uses the maximum inter-class variance method (OSTU algorithm) to binarize the solar cell image after removing the horizontal grid lines, and uses horizontal scanning to locate and remove the electrodes.

The principle of maximum inter-class variance method (OSTU algorithm): for an image, when the segmentation threshold between foreground and background is set to t, the proportion of foreground points in the image is w ₀ , the mean value is u ₀ , and the proportion of background points in the image is w ₁ , the mean is u ₁ . Then the mean value of the whole image is u=w ₀ *u ₀ +w ₁ *u ₁ . Establish the objective function g(t)=w ₀ *(u ₀ -u) ² +w ₁ *(u ₁ -u) ² , and g(t) is the expression of the inter-class variance when the segmentation threshold is t. The OSTU algorithm is that g(t) obtains the global maximum value, and when g(t) is the maximum value, the corresponding t becomes the optimal threshold.

For the grayscale image after solar cell correction, the solar cell area is the foreground, and the rest is the background. When the optimal threshold g(t) is obtained, compare each pixel in the image with g(t), if it is greater than g(t), set the pixel gray value of this point to 1; Let the gray value of the pixel point be 0. In this way, the binary image of the image can be obtained.

There are 4 electrodes on the solar cell, which are located at 1/8, 3/8, 5/8, and 7/8 in the horizontal direction of the cell. During the horizontal scanning process, firstly scan to the right from 1/16 of the horizontal direction until the gray value of the pixel point is 1. At this time, the left boundary of the electrode has been reached, and continue to scan to the right until the gray value of the pixel point is reached. If it is 0, the right boundary of the electrode has been reached at this time, so the first electrode is located. Similarly, locate the remaining electrodes.

The electrode is removed by compensating the gray value of the pixel point of the electrode according to the gray value of the pixel point in the adjacent regions on the left and right sides of the electrode.

S54 locates vertical grid lines and boundary lines using horizontal scanning and removes them.

There are three vertical grid lines on the solar cell, which are respectively located at 1/4, 1/2 and 3/4 of the horizontal direction of the cell. During the horizontal scanning process, first scan to the right from 1/5 of the horizontal direction until the gray value of the pixel point is 1. At this time, the left boundary of the electrode has been reached, and continue to scan to the right until the gray value of the pixel point is reached. If it is 0, the right boundary of the electrode has been reached at this time, so the first vertical grid line is located. Similarly, locate the remaining vertical grid lines.

There are 2 boundary lines on the solar cell sheet, the left boundary and the right boundary respectively. In the process of horizontal scanning, firstly start scanning to the right from the left border of the horizontal direction until the gray value of the pixel point is 1. At this time, the left border of the electrode has been reached, and continue to scan to the right until the gray value of the pixel point is 1. 0, the right boundary of the electrode has been reached at this time, so the first boundary line is located. Similarly, locate the remaining boundary lines.

Use morphological operations to remove vertical grid lines and boundary lines.

S55 uses the method of finding contours to find holes and dirt.

Use the method of finding contours to find all the contours in the image: 1. Scan the binary image of the solar cell to find the first pixel that does not belong (that is, the pixel has a gray value of 1, and is adjacent to the surrounding eight pixels) The gray mean value of the domain is greater than 0.5), and the coordinates of the pixel point are set as (x ₀ , y ₀ ). 2. Taking (x ₀ , y ₀ ) as the center, consider the eight neighborhood pixels (x, y) around (x ₀ , y ₀ ), if (x, y) satisfies the growth criterion (that is, the gray value of the pixel point) is 1, and the gray mean value of the surrounding eight neighborhoods is greater than 0.5), merge (x, y) with (x ₀ , y ₀ ) (in the same area), and push (x, y) into the stack ( a way to store data). 3. Take a pixel from the stack and return it as (x ₀ , y ₀ ) to step 2. 4. When the stack is empty, go back to step 1. 5. Repeat steps 1-4 until every point in the image is have attribution.

Through the above steps, you can find all the contours in the binary image, and compare the contour size with the contour threshold (that is, according to the number of pixels in the contour to determine whether it is a hole or dirt, usually the value is 100), if it is larger than the contour If the threshold is set, it is determined as holes and dirt, and their position information (center, length, width) is obtained.

S56 calculates the gray value of the position area, judges whether the defect is a hole or a dirt according to the contour threshold, and counts the number.

Using the position information found in S55, calculate the gray average value in this area, and compare it with the hole contamination threshold (that is, the threshold for distinguishing holes and dirt, usually the average value of the background gray value and the gray maximum value, usually Take 125) for comparison, if it is greater than the hole contamination threshold, it is judged as a hole; if it is less than the threshold, it is judged as a dirt, and the number is counted.

The specific embodiments of the invention have been described in detail above, but the present invention is not limited to the specific embodiments described above, which are only used as examples. For those skilled in the art, any equivalent modifications and substitutions to the system are also within the scope of the present invention. Therefore, equivalent changes and modifications made without departing from the spirit and scope of the invention should be included within the scope of the present invention.

Claims (9)

1. A solar cell defect detection system based on a CIS image acquisition unit is characterized by comprising: the system comprises a transmission device (1), an image acquisition device (2) and a data processing module (3);

the image acquisition device (2) is positioned right above the conveying device (1), and the image acquisition device (2) is connected with the data processing module (3); the solar cell pieces placed at will are transported through the conveying device (1), when the solar cell pieces pass below the image acquisition device (2), the image acquisition device (2) acquires images of the solar cell pieces and transmits the images to the data processing module (3), and the data processing module (3) performs defect analysis and detection on the acquired images;

the image acquisition device (2) comprises: the device comprises a CIS image acquisition unit (21) and two strip-shaped light sources (22), wherein the CIS image acquisition unit (21) is arranged right above the conveying device (1), and two strip-shaped light sources are respectively arranged on two sides of the CIS image acquisition unit and are used for enabling the brightness of the acquired images of the solar cell to be uniform;

the CIS image acquisition unit (21) comprises: the system comprises an FPGA module, a CIS image acquisition module, an analog-to-digital conversion module, a storage module, an acquisition speed matching module, an image output module and a power supply module, wherein the CIS image acquisition module, the analog-to-digital conversion module, the storage module, the acquisition speed matching module and the image output module are respectively connected with the FPGA module;

the CIS image acquisition module is used for acquiring images of the solar cell and outputting a series of analog data; the analog-to-digital conversion module is used for converting analog data into digital signals; the FPGA module is used for realizing the time sequence control of the whole system; the acquisition speed matching module is used for adjusting the image acquisition speed of the CIS image module according to the transmission speed of the transmission device; the storage module is used for caching the acquired digital image data; the image output module is used for sending the collected solar cell images to the data processing module.

2. A solar cell defect detection method based on a CIS image acquisition unit is characterized by comprising the following steps:

(1) the solar cell pieces which are randomly placed are transported through the conveying device, and when the solar cell pieces pass through the lower part of the image acquisition device, the image acquisition device is used for acquiring images of the solar cell pieces;

(2) correcting the solar cell according to the acquired image to obtain a corrected solar cell image;

the step (2) is specifically as follows:

s21, carrying out graying processing on the collected solar cell images to obtain a grayscale image of the solar cell;

s22 obtaining pixel points representing four boundaries of the solar cell piece on the gray-scale image through searching; wherein, the searching is performed by the order of the upper boundary, the lower boundary, the left boundary and the right boundary;

s23, screening the pixel points of the four boundaries and obtaining four correction boundaries of the solar cell;

s24, calculating intersection points among the four correction boundaries, namely four corner points of the solar cell;

s25, obtaining a perspective transformation matrix according to four corners of the solar cell and the set four corners, and obtaining an image corrected by the solar cell after performing perspective transformation on the collected image through the perspective transformation matrix;

(3) and carrying out unfilled corner processing or edge breakage processing on the solar cell image to obtain a defect detection result.

3. The method for detecting defects of a solar cell slice as claimed in claim 2, further comprising after the step (2):

and (2.0) removing electrodes and grid lines from the solar cell image to obtain a standardized solar cell gray scale image.

4. The method for detecting the defects of the solar cell as claimed in claim 3, wherein the step of removing the electrodes and the grid lines comprises the following specific steps:

(2.01) carrying out gray processing on the image of the corrected image of the solar cell to obtain a gray image of the solar cell;

(2.02) removing the horizontal grid lines on the gray-scale image by using morphological open operation to obtain the gray-scale image of the solar cell after the horizontal grid lines are removed;

(2.03) on the solar cell gray-scale image with the horizontal grid lines removed, carrying out binarization processing on the solar cell with the horizontal grid lines removed by using a maximum inter-class variance method, and positioning and removing the electrodes by using horizontal scanning to obtain the solar cell gray-scale image with the horizontal grid lines and the electrodes removed;

and (2.04) on the gray-scale image of the solar cell with the horizontal grid lines and the horizontal electrodes removed, positioning the vertical grid lines and the boundary lines by using horizontal scanning, and removing to obtain a standardized gray-scale image of the solar cell.

5. The method for detecting defects of a solar cell slice as claimed in claim 3, further comprising after the step (2.0): and (2.1) carrying out hole treatment or dirt treatment on the standardized gray level image of the solar cell to obtain a defect detection result.

6. The method for detecting defects of a solar cell as claimed in claim 5, wherein the hole processing specifically comprises:

(2.11) searching positions of holes and dirt on a standardized gray scale image of the solar cell by using a contour searching method;

and (2.12) calculating the mean value of the gray values of the position areas by using the positions found in (2.11), comparing the mean value with a set threshold value, judging whether the defects are holes or dirt, and respectively counting the number of the holes or the dirt.

7. The method for detecting the defects of the solar cell as claimed in claim 5, wherein the dirt treatment is specifically as follows:

(2.11) searching positions of holes and dirt on a standardized gray scale image of the solar cell by using a contour searching method;

and (2.12) calculating the mean value of the gray values of the position areas by using the positions found in (2.11), comparing the mean value with a set threshold value, judging whether the defects are holes or dirt, and respectively counting the number of the holes or the dirt.

8. The method for detecting defects of solar cells as claimed in any one of claims 2 to 7, wherein the corner defect processing in the step (3) is specifically as follows:

(3.11) carrying out gray processing on the image of the corrected image of the solar cell to obtain a gray image of the solar cell;

(3.12) detecting whether the upper left corner of the solar cell is unfilled from the horizontal right direction and the vertical downward direction by using three step search modes on the gray-scale map;

(3.13) detecting whether the upper right corner of the solar cell is unfilled from the horizontal left direction and the vertical downward direction by using three step search modes on the gray-scale image;

(3.14) detecting whether the left lower corner of the solar cell is unfilled from the horizontal right direction and the vertical upward direction by using three step search modes on the gray-scale map;

and (3.15) detecting whether the corner of the lower right corner of the solar cell is unfilled from the horizontal left and the vertical downward by using three step search modes on the gray-scale map.

9. The method for detecting defects of a solar cell as claimed in any one of claims 2 to 7, wherein the edge breakage treatment in the step (3) is specifically:

(3.11) carrying out gray processing on the image of the corrected image of the solar cell to obtain a gray image of the solar cell;

(3.12) projecting the left and right boundaries on the gray scale map, and respectively recording the number of pixel points in the edge collapse area;

(3.13) projecting the upper and lower boundaries on the gray-scale image, and respectively recording the number of pixel points in the edge collapse area;

(3.14) comparing the number of the pixel points in the edge collapse area with a set threshold value, judging whether the edge collapse is caused, and counting the number of the edge collapse;

(3.15) considering the case that the four corners of the solar cell are unfilled corners, determining the starting point and the end point of the continuous boundary of the four sides, and repeating the steps (3.12) - (3.14).

Images