CN114187260A - Photovoltaic module defect detection method and system - Google Patents
Photovoltaic module defect detection method and system Download PDFInfo
- Publication number
- CN114187260A CN114187260A CN202111506090.8A CN202111506090A CN114187260A CN 114187260 A CN114187260 A CN 114187260A CN 202111506090 A CN202111506090 A CN 202111506090A CN 114187260 A CN114187260 A CN 114187260A
- Authority
- CN
- China
- Prior art keywords
- image
- detection image
- detection
- neural network
- convolutional neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 211
- 230000007547 defect Effects 0.000 title claims abstract description 135
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 83
- 238000012545 processing Methods 0.000 claims abstract description 57
- 238000000034 method Methods 0.000 claims abstract description 28
- 238000001914 filtration Methods 0.000 claims abstract description 21
- 230000004927 fusion Effects 0.000 claims abstract description 13
- 238000012937 correction Methods 0.000 claims abstract description 10
- 238000003709 image segmentation Methods 0.000 claims abstract description 10
- 230000007246 mechanism Effects 0.000 claims description 46
- 238000007637 random forest analysis Methods 0.000 claims description 23
- 230000005540 biological transmission Effects 0.000 claims description 15
- 238000010191 image analysis Methods 0.000 claims description 14
- 239000000126 substance Substances 0.000 claims 1
- 230000008569 process Effects 0.000 description 7
- 238000002372 labelling Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 238000004519 manufacturing process Methods 0.000 description 4
- 238000012546 transfer Methods 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000012790 confirmation Methods 0.000 description 3
- 230000002950 deficient Effects 0.000 description 3
- 238000002503 electroluminescence detection Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 229910021419 crystalline silicon Inorganic materials 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000003708 edge detection Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005401 electroluminescence Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000007599 discharging Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003707 image sharpening Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 239000002210 silicon-based material Substances 0.000 description 1
- 239000002910 solid waste Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/60—Rotation of a whole image or part thereof
- G06T3/608—Skewing or deskewing, e.g. by two-pass or three-pass rotation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
-
- G06T5/70—
-
- G06T5/73—
-
- G06T5/90—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10048—Infrared image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20076—Probabilistic image processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Abstract
The invention provides a method and a system for detecting defects of a photovoltaic module, wherein the method comprises the following steps: acquiring an EL image of a photovoltaic module to be detected; performing fusion filtering denoising processing on the EL image to obtain a first detection image; carrying out position correction processing on the first detection image to obtain a second detection image; performing ROI division processing on the second detection image to acquire a third detection image; performing image enhancement processing on the third detection image to obtain a fourth detection image; performing image segmentation processing on the fourth detection image to obtain a target detection image; and adopting a target convolutional neural network to detect and classify the defects of the target detection image and marking the positions of the defects. According to the photovoltaic module defect detection method, uncertain factors of manual detection can be eliminated, detection classification automation is realized, the detection quality and the detection efficiency are improved, meanwhile, labor force is greatly saved, and unnecessary economic loss is avoided.
Description
Technical Field
The invention relates to the technical field of photovoltaic module detection, in particular to a photovoltaic module defect detection method and a photovoltaic module defect detection system.
Background
As the environment becomes increasingly polluted, the solar industry, which is one of clean energy, is also gradually developed. The main carrier of the solar power generation is a cell panel, other harmful gas or solid waste is not generated in the energy conversion process, and the solar power generation system is a novel energy source which is environment-friendly, safe and pollution-free. At present, more than 90% of the solar panels are made of crystalline silicon materials, and due to the influence of production equipment, production raw material quality, process parameters and the like, various defects may occur in the production and processing flow of the crystalline silicon solar cells, so that the photoelectric conversion efficiency and the service life of the solar cells are seriously influenced.
In the related art, a defect detection method of a solar cell (photovoltaic module) is mainly based on an electroluminescence imaging technology and depends on manual observation for judgment. However, the subjective judgment standards of the detection personnel are different, the detection personnel have large uncertain factors, the phenomena of false detection, missed detection and the like easily occur, the detection speed is low, the efficiency is low, the requirement of fast detection of a production line is difficult to meet, and unnecessary economic loss is easily caused.
Disclosure of Invention
The invention aims to solve the technical problems and provides a photovoltaic module defect detection method, which adopts an automatic detection and classification mode to replace a manual observation detection mode, eliminates uncertainty factors of manual detection, realizes automatic detection and classification, greatly saves labor force while improving detection quality and detection efficiency, and avoids unnecessary economic loss.
The technical scheme adopted by the invention is as follows:
a photovoltaic module defect detection method comprises the following steps: acquiring an EL image of a photovoltaic module to be detected; performing fusion filtering denoising processing on the EL image to obtain a first detection image; carrying out position correction processing on the first detection image to obtain a second detection image; performing ROI division processing on the second detection image to acquire a third detection image; performing image enhancement processing on the third detection image to obtain a fourth detection image; performing image segmentation processing on the fourth detection image to obtain a target detection image; and adopting a target convolutional neural network to detect and classify the defects of the target detection image and marking the positions of the defects.
The target convolutional neural network comprises a convolutional neural network with multi-path contrast output, wherein a tail full-connection layer of the VGGNet convolutional neural network is respectively output to a softmax classifier and a Random Forest classifier to form the convolutional neural network with the multi-path contrast output, and the defect detection and classification of the target detection image by adopting the target convolutional neural network comprises the following steps: when the target detection image is input into the convolutional neural network with the multi-path contrast output, classifying the same object output by the whole connecting layer at the tail end of the VGGNet convolutional neural network through the softmax classifier and the Random Forest classifier so as to respectively obtain a first classification probability value and a second classification probability value; and comparing the first classification probability value and the second classification probability value with a first probability standard value, and confirming a defect detection result according to a comparison result.
The comparing the first classification probability value and the second classification probability value with a first probability standard value, and determining a defect detection result according to the comparison result includes: judging whether the defect categories output by the softmax classifier and the Random Forest classifier are the same or not; if not, carrying out exception prompt; if so, judging whether the first classification probability value is larger than the first probability standard value, and judging whether the second classification probability value is larger than the first probability standard value; if the first classification probability value is larger than the first probability standard value and the second classification probability value is larger than the first probability standard value, confirming that the defect class of the target detection image is the defect class output by the softmax classifier and the Random Forest classifier; confirming that the defect class of the target detection image is the defect class output by the softmax classifier if the first classification probability value is larger than the first probability standard value and the second classification probability value is smaller than or equal to the first probability standard value; if the first classification probability value is smaller than or equal to the first probability standard value and the second classification probability value is larger than the first probability standard value, confirming that the defect category of the target detection image is the defect category output by the Random Forest classifier; and if the first classification probability value is less than or equal to the first probability standard value and the second classification probability value is less than or equal to the first probability standard value, performing low-probability prompt.
The target convolutional neural network comprises a fused convolutional neural network, wherein the fused convolutional neural network is composed of a VGGNet convolutional neural network and an AlexNet convolutional neural network, and the defect detection and classification of the target detection image by adopting the target convolutional neural network comprises the following steps: outputting a third classification probability value and a fourth classification probability value through the VGGNet convolutional neural network and the AlexNet convolutional neural network, respectively, when the target detection image is input into the fused convolutional neural network; and confirming a defect detection result according to the third classification probability value and the fourth classification probability value.
The determining a defect detection result according to the third classification probability value and the fourth classification probability value includes: judging whether the third classification probability value is larger than a second probability standard value or not; if so, confirming that the defect type of the target detection image is the defect type output by the VGGNet convolutional neural network; if not, judging whether the fourth classification probability value is larger than a third probability standard value or not; if so, confirming that the defect type of the target detection image is the defect type output by the AlexNet convolutional neural network; if not, judging whether the third classification probability value is larger than the fourth classification probability value or not; if so, confirming that the defect type of the target detection image is the defect type output by the VGGNet convolutional neural network; and if not, determining that the defect type of the target detection image is the defect type output by the AlexNet convolutional neural network.
A photovoltaic module defect detection system, comprising: the photovoltaic module detection device comprises a conveying mechanism, an image shooting mechanism, an image transmission element and an image analysis element, wherein the conveying mechanism is distributed on two sides of the image shooting mechanism and used for conveying a photovoltaic module to be detected to a processing area; the image shooting mechanism is used for acquiring an EL image of the photovoltaic module to be detected; the image transmission element is used for receiving the EL image of the photovoltaic module to be detected and transmitting the EL image of the photovoltaic module to be detected to the image analysis element; the image analysis element is used for carrying out fusion filtering denoising processing on the EL image to obtain a first detection image, carrying out position correction processing on the first detection image to obtain a second detection image, carrying out ROI (region of interest) division processing on the second detection image to obtain a third detection image, carrying out image enhancement processing on the third detection image to obtain a fourth detection image, carrying out image segmentation processing on the fourth detection image to obtain a target detection image, and carrying out defect detection and classification on the target detection image by adopting a target convolutional neural network; wherein the image capturing mechanism is further configured to mark the defect location.
The invention has the beneficial effects that:
the invention adopts an automatic detection and classification mode to replace a manual observation detection mode, eliminates uncertainty factors of manual detection, realizes automatic detection and classification, greatly saves labor force while improving detection quality and detection efficiency, and avoids causing unnecessary economic loss.
Drawings
FIG. 1 is a flow chart of a method for detecting defects in a photovoltaic module according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a defect detection system for a photovoltaic module according to an embodiment of the present invention;
FIG. 3 is a schematic structural view of a delivery mechanism according to one embodiment of the present invention;
FIG. 4 is a schematic structural view of a delivery mechanism according to another embodiment of the present invention;
fig. 5 is a schematic structural diagram of an image capturing mechanism according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a flow chart of a method for detecting defects in a photovoltaic module according to an embodiment of the present invention.
As shown in fig. 1, the method for detecting defects of a photovoltaic module according to an embodiment of the present invention may include the following steps:
and S1, acquiring the EL image of the photovoltaic module to be detected.
Specifically, a forward bias voltage may be applied to a photovoltaic module (solar panel assembly) to be tested in a dark box, and then a CCD (charge coupled device) camera is used to capture and collect an EL (Electroluminescence) image of the photovoltaic module to be tested.
And S2, performing fusion filtering and denoising processing on the EL image to obtain a first detection image.
Specifically, a weighted fusion algorithm of median filtering and gaussian filtering may be adopted to perform fusion filtering denoising processing on the EL image to obtain the first detection image.
Specifically, the median filtering and the gaussian filtering are weighted and fused according to the formula (1), and the sum of the weights is 1, that is to say
g＝k×f+(1-k)×h， (1)
Wherein g is the result image, k is the weight, f and h represent median filtering and gaussian filtering. The median filtering is nonlinear filtering, so that pulse interference noise and isolated noise points can be effectively filtered; gaussian filtering is linear filtering, Gaussian noise can be effectively removed, the two filtering methods are weighted and fused, a proper weight is found in the experimental process, a better denoising effect is achieved, and the image edge and local details are effectively protected.
S3, the first detected image is subjected to position correction processing to acquire a second detected image.
Wherein the position of the first detection image can be corrected by means of perspective transformation.
Specifically, since the CCD camera is likely to have a problem such as an angular tilt due to the influence of the imaging angle, the position of the obtained image can be corrected by perspective conversion. Specifically, a threshold value is set for the obtained image (first detection image) to perform binarization, the outline is extracted, the polygon is approximated to obtain a quadrangle, a convex hull is searched to obtain a vertex, a vertex coordinate is set according to the size of the photovoltaic module to be detected, a function is called to obtain a perspective transformation matrix, and then the first detection image is subjected to perspective transformation to obtain a correction image (second detection image).
S4, the second detection image is subjected to ROI division processing to acquire a third detection image.
The ROI division processing can be performed on the second detection image by adopting a maximum connected region method and combining a Sobel edge detection mode to obtain a third detection image.
Specifically, the minimum external rectangle of the photovoltaic module to be detected is obtained by using a maximum connected domain method, the image is intercepted to perform preliminary foreground extraction, and then the battery module area is segmented by using Sobel edge detection. Wherein, the convolution template used by Sobel operator is shown in formula (2), namely
S5, image enhancement processing is performed on the third detected image to acquire a fourth detected image.
The image enhancement processing comprises contrast enhancement and image sharpening operations. In the actual process, proper contrast and sharpening degree are adjusted according to the actual image effect, and the detail features of the image are highlighted to the greatest extent so as to detect defects.
S6, an image segmentation process is performed on the fourth detection image to acquire a target detection image.
The fourth detection image may be subjected to image segmentation processing by an bisection method to obtain a target detection image, and the position of each battery cell unit in the target detection image may be marked.
Specifically, according to the specifications of the photovoltaic module to be detected and the cell units, the images are equally divided in the vertical direction and the horizontal direction, and the cell units are extracted and subjected to position marking.
And S7, adopting the target convolutional neural network to detect and classify the defects of the target detection image and marking the positions of the defects.
According to one embodiment of the invention, the target convolutional neural network comprises a convolutional neural network with multiple contrast outputs, wherein the last full connection layer of the VGGNet convolutional neural network is respectively output to a softmax classifier and a Random Forest classifier to form the convolutional neural network with the multiple contrast outputs, and the defect detection and classification of the target detection image by using the target convolutional neural network comprises the following steps: when a target detection image is input into a convolutional neural network with multipath contrast output, classifying the same object output by a full connection layer at the tail end of the VGGNet convolutional neural network through a softmax classifier and a Random Forest classifier to respectively obtain a first classification probability value and a second classification probability value; and comparing the first classification probability value and the second classification probability value with the first probability standard value, and confirming the defect detection result according to the comparison result.
In an embodiment of the present invention, comparing the first classification probability value and the second classification probability value with the first probability standard value, and confirming the defect detection result according to the comparison result, includes: judging whether the defect categories output by the softmax classifier and the Random Forest classifier are the same or not; if not, carrying out exception prompt; if so, judging whether the first classification probability value is larger than a first probability standard value or not, and judging whether the second classification probability value is larger than the first probability standard value or not; if the first classification probability value is larger than the first probability standard value and the second classification probability value is larger than the first probability standard value, confirming that the defect category of the target detection image is the defect category output by the softmax classifier and the Random Forest classifier; if the first classification probability value is larger than the first probability standard value and the second classification probability value is smaller than or equal to the first probability standard value, confirming that the defect type of the target detection image is the defect type output by the softmax classifier; if the first classification probability value is less than or equal to the first probability standard value and the second classification probability value is greater than the first probability standard value, confirming that the defect class of the target detection image is the defect class output by the Random Forest classifier; and if the first classification probability value is less than or equal to the first probability standard value and the second classification probability value is less than or equal to the first probability standard value, performing low-probability prompt.
Specifically, when a target detection image is input into a convolutional neural network having a multi-path contrast output, the same object is classified using two classifiers, namely, softmax classifier and Random Forest, based on VGGNet, and output probability values of the two classifiers are recorded, namely, the softmax classifier outputs a first classification probability value and the Random Forest classifier outputs a second classification probability value. And comparing the first classification probability value and the second classification probability value with the first probability standard value, and further confirming the defect detection result.
It should be noted that the softmax classifier and the Random Forest classifier output a classification probability value and also output a corresponding defect class.
Specifically, the first probability standard value may be x1 (the specific value may be calibrated according to the actual situation), the first classification probability value is P1, and the second classification probability value is P2. If the defect categories output by the softmax classifier and the Random Forest classifier are the same, and the first classification probability value P1 and the second classification probability value P2 are both greater than the first probability standard value x1, determining that the defect category of the target detection image is the category output by the softmax classifier and the Random Forest classifier; if the defect categories output by the softmax classifier and the Random Forest classifier are the same, one of the first classification probability value P1 and the second classification probability value P2 is larger than the first probability standard value x1, and the other one is smaller than or equal to x1, outputting a defect category result output by the classifier with the classification probability value larger than the first probability standard value x 1; if the defect categories output by the softmax classifier and the Random Forest classifier are the same, and the first classification probability value P1 and the second classification probability value P2 are both smaller than the first probability standard value x1, performing low-probability prompt; if the defect categories output by the softmax classifier and the Random Forest classifier are different, no matter what relationship the first classification probability value P1 and the second classification probability value P2 are with the first probability standard value x1, no defect category is output, and exception prompting is carried out. Wherein, the low probability prompt and the abnormal prompt need to be further confirmed by the staff, namely, the staff is required to confirm the defect type.
According to another embodiment of the invention, the target convolutional neural network comprises a fused convolutional neural network, wherein the fused convolutional neural network is composed of a VGGNet convolutional neural network and an AlexNet convolutional neural network, and the defect detection and classification of the target detection image by using the target convolutional neural network comprises the following steps: when the target detection image is input into the fusion convolutional neural network, a third classification probability value and a fourth classification probability value are respectively output through the VGGNet convolutional neural network and the AlexNet convolutional neural network; and confirming the defect detection result according to the third classification probability value and the fourth classification probability value.
In one embodiment of the present invention, confirming the defect detection result according to the third classification probability value and the fourth classification probability value includes: judging whether the third classification probability value is greater than a second probability standard value or not; if so, confirming that the defect type of the target detection image is the defect type output by the VGGNet convolutional neural network; if not, judging whether the fourth classification probability value is greater than a third probability standard value or not; if so, confirming that the defect type of the target detection image is the defect type output by the AlexNet convolutional neural network; if not, judging whether the third classification probability value is greater than the fourth classification probability value or not; if so, confirming that the defect type of the target detection image is the defect type output by the VGGNet convolutional neural network; and if not, determining that the defect type of the target detection image is the defect type output by the AlexNet convolutional neural network.
Specifically, when the target detection image is input into the fusion convolutional neural network, the VGGNet convolutional neural network performs the first-step detection, outputs the third classification probability value and the corresponding defect type, and the AlexNet convolutional neural network performs the second-step defect detection, and outputs the fourth classification probability value and the corresponding defect type. And comparing the third classification probability value with the second probability standard value, and comparing the fourth classification probability value with the third probability standard value, thereby confirming the defect detection result.
Specifically, let the third classification probability value output by the VGGNet convolutional neural network be P3, the corresponding second probability criterion value be x2, the fourth classification probability value output by the AlexNet convolutional neural network be P4, and the corresponding third probability criterion value be x 3. If the third classification probability value P3 output by the VGGNet convolutional neural network is larger than the second probability standard value x2, directly outputting a defect type confirmation result output by the VGGNet convolutional neural network; and if the third classification probability value P3 is less than or equal to the second probability standard value x2, judging by an AlexNet convolutional neural network. If the fourth classification probability value P4 output by the AlexNet convolutional neural network is larger than the third probability standard value x3, outputting a defect classification confirmation result output by the AlexNet convolutional neural network; if the fourth classification probability value P4 is less than or equal to the third probability criterion value x3, the third classification probability value P3 is compared with the fourth classification probability value P4. If the third classification probability value P3 is larger than the fourth classification probability value P4, confirming the result output according to the defect classification output by the VGGNet convolutional neural network; and if the third classification probability value P3 is less than or equal to the fourth classification probability value P4, outputting a defect type confirmation result output by the AlexNet convolutional neural network.
In practice, the number of defect samples is small, and a large number of training samples are required for training the convolutional neural network, so that the number of training samples can be increased by performing operations such as rotation and translation on the defect sample image.
In summary, according to the method for detecting defects of a photovoltaic module in the embodiments of the present invention, an EL image of the photovoltaic module to be detected is obtained, the EL image is subjected to fusion filtering and denoising processing to obtain a first detection image, the first detection image is subjected to position correction processing to obtain a second detection image, the second detection image is subjected to ROI division processing to obtain a third detection image, and the third detection image is subjected to image enhancement processing to obtain a fourth detection image. And performing image segmentation processing on the fourth detection image to obtain a target detection image, performing defect detection and classification on the target detection image by adopting a target convolutional neural network, and marking the defect position. From this, adopt automated inspection and categorised mode to replace the manual observation detection mode, got rid of the nondeterminacy factor that artifical detected, realize detecting categorised automation, when improving detection quality and detection efficiency, saved the labour greatly, avoid causing unnecessary economic loss.
Corresponding to the embodiment, the invention further provides a photovoltaic module defect detection system.
As shown in fig. 2, the photovoltaic module defect detection system may include: a conveying mechanism 1, an image capturing mechanism 2, an image transmission component 3, and an image analysis component 4. The conveying mechanisms 1 are distributed on two sides of the image shooting mechanism 2, and the conveying mechanisms 1 are used for conveying the photovoltaic module to be detected to the processing area; the image shooting mechanism 2 is used for acquiring an EL image of the photovoltaic module to be detected; the image transmission element 3 is used for receiving the EL image of the photovoltaic component to be detected and transmitting the EL image of the photovoltaic component to be detected to the image analysis element 4; the image analysis element 4 is used for performing fusion filtering denoising processing on the EL image to obtain a first detection image, performing position correction processing on the first detection image to obtain a second detection image, performing ROI segmentation processing on the second detection image to obtain a third detection image, performing image enhancement processing on the third detection image to obtain a fourth detection image, performing image segmentation processing on the fourth detection image to obtain a target detection image, and performing defect detection and classification on the target detection image by adopting a target convolutional neural network; wherein the image capturing mechanism 2 is also used for marking defect positions.
Wherein, the transmission mechanism 1 may include a transmission mechanism 11 and a transmission mechanism 12 distributed on both sides of the image capturing mechanism 2. The carrying-in mechanism 11 is on the inlet side of the image capturing mechanism 2, and the carrying-out mechanism 12 is on the outlet side of the image capturing mechanism 2.
In one embodiment of the invention, as shown in fig. 3 and 4, the feeding mechanism 11 has a one-way conveyor, and the discharging mechanism 12 has a longitudinal conveyor 121 and a transverse conveyor 122. When the conveyor belt works normally, the position of the transverse conveyor belt 122 is lower than that of the longitudinal conveyor belt 121, and the longitudinal conveyor belt 121 works to carry out longitudinal conveying; when the defective solar cell is transferred, the transverse transfer belt 122 is elevated above the longitudinal transfer belt 121, and the transverse transfer belt 122 operates to transfer the defective solar cell transversely to the processing area.
In one embodiment of the present invention, as shown in fig. 5, the image capturing mechanism 2 may be composed of a box 21, a conveyor 22, a camera 23, a labeling device 24 and a dc power supply 25, wherein the conveyor 22 is disposed at the entrance and exit of the two sides in the box 1, the camera 23 is mounted at the bottom of the box 1, the labeling device 24 is mounted at the top of the box 1, and the dc power supply 25 is mounted at one side of the box 1 to provide a forward bias voltage for the electroluminescence detection of the photovoltaic module 26 to be detected.
Specifically, the shooting devices 23 include infrared cameras, such as CCD infrared cameras, etc., and the number of the shooting devices 23 and the installation position at the bottom of the box body 1 are adjusted according to the specification of the photovoltaic module 26; when the photovoltaic module 26 reaches a designated position, the positive electrode and the negative electrode of the direct current power supply 25 are respectively connected with the positive electrode and the negative electrode of the photovoltaic module 26, and forward bias voltage is provided for electroluminescence detection of the photovoltaic module 26 to be detected; if the photovoltaic module 26 is detected to be defective, the labeling device 24 labels the defect type and the defect position at the corresponding position of the photovoltaic module to be detected.
After the image analysis element 4 receives the EL image, the defect detection, classification and position labeling of the solar cell are completed according to the defect detection method of the photovoltaic module of the above embodiment. To avoid redundancy, it will not be described in detail here.
Specifically, the photovoltaic module to be detected is firstly transmitted into the image shooting mechanism through the transmission mechanism, then the direct-current power supply is connected with the photovoltaic module to be detected to provide forward bias voltage for electroluminescence detection of the photovoltaic module, the shooting device shoots an EL image, and the EL image is transmitted to the image analysis element through the image transmission element to complete defect detection of the photovoltaic module. If no defect is detected, the photovoltaic module is transmitted to the transmission mechanism from the image shooting mechanism and is transmitted to the next procedure by the transmission mechanism; if the defects exist, the labeling device is used for labeling the types and the positions of the defects on the corresponding positions of the photovoltaic modules, the labels are conveyed to the conveying-out mechanism, the labels are conveyed to the processing area from the other conveying direction of the conveying-out mechanism, and the labels are waited for further processing by workers. From this, adopt automated inspection and categorised mode to replace the manual observation detection mode, got rid of the nondeterminacy factor that artifical detected, realize detecting categorised automation, when improving detection quality and detection efficiency, saved the labour greatly, avoid causing unnecessary economic loss.
It should be noted that, for a more specific implementation of the system for detecting defects of a photovoltaic module according to the embodiment of the present invention, reference may be made to the above-mentioned embodiment of the method for detecting defects of a photovoltaic module, and details are not described herein again.
The photovoltaic module defect detection system comprises: the photovoltaic module detection device comprises a conveying mechanism, an image shooting mechanism, an image transmission element and an image analysis element, wherein the conveying mechanism is distributed on two sides of the image shooting mechanism and used for conveying a photovoltaic module to be detected to a processing area; the image shooting mechanism is used for acquiring an EL image of the photovoltaic module to be detected; the image transmission element is used for receiving the EL image of the photovoltaic module to be detected and transmitting the EL image of the photovoltaic module to be detected to the image analysis element; the image analysis element is used for carrying out fusion filtering denoising processing on the EL image to obtain a first detection image, carrying out position correction processing on the first detection image to obtain a second detection image, carrying out ROI division processing on the second detection image to obtain a third detection image, carrying out image enhancement processing on the third detection image to obtain a fourth detection image, carrying out image segmentation processing on the fourth detection image to obtain a target detection image, and carrying out defect detection and classification on the target detection image by adopting a target convolutional neural network; wherein the image capturing mechanism is further used for marking the defect position. From this, adopt automated inspection and categorised mode to replace the manual observation detection mode, got rid of the nondeterminacy factor that artifical detected, realize detecting categorised automation, when improving detection quality and detection efficiency, saved the labour greatly, avoid causing unnecessary economic loss.
In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implying any number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. The meaning of "plurality" is two or more unless specifically limited otherwise.
In the present invention, unless otherwise expressly stated or limited, the terms "mounted," "connected," "secured," and the like are to be construed broadly and can, for example, be fixedly connected, detachably connected, or integrally formed; can be mechanically or electrically connected; either directly or indirectly through intervening media, either internally or in any other relationship. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
In the present invention, unless otherwise expressly stated or limited, the first feature "on" or "under" the second feature may be directly contacting the first and second features or indirectly contacting the first and second features through an intermediate. Also, a first feature "on," "over," and "above" a second feature may be directly or diagonally above the second feature, or may simply indicate that the first feature is at a higher level than the second feature. A first feature being "under," "below," and "beneath" a second feature may be directly under or obliquely under the first feature, or may simply mean that the first feature is at a lesser elevation than the second feature.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps of the process, and alternate implementations are included within the scope of the preferred embodiment of the present invention in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present invention.
The logic and/or steps represented in the flowcharts or otherwise described herein, e.g., an ordered listing of executable instructions that can be considered to implement logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium could even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via for instance optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.
It should be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.
It will be understood by those skilled in the art that all or part of the steps carried by the method for implementing the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer readable storage medium, and when the program is executed, the program includes one or a combination of the steps of the method embodiments.
In addition, functional units in the embodiments of the present invention may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may also be stored in a computer readable storage medium.
The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc. Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.
Claims (6)
1. A photovoltaic module defect detection method is characterized by comprising the following steps:
acquiring an EL image of a photovoltaic module to be detected;
performing fusion filtering denoising processing on the EL image to obtain a first detection image;
carrying out position correction processing on the first detection image to obtain a second detection image;
performing ROI division processing on the second detection image to acquire a third detection image;
performing image enhancement processing on the third detection image to obtain a fourth detection image;
performing image segmentation processing on the fourth detection image to obtain a target detection image;
and adopting a target convolutional neural network to detect and classify the defects of the target detection image and marking the positions of the defects.
2. The method according to claim 1, wherein the target convolutional neural network comprises a convolutional neural network with multiple contrast outputs, wherein the last fully-connected layer of the VGGNet convolutional neural network is respectively output to a softmax classifier and a Random Forest classifier to form the convolutional neural network with multiple contrast outputs, and the defect detection and classification of the target detection image by using the target convolutional neural network comprises:
when the target detection image is input into the convolutional neural network with the multi-path contrast output, classifying the same object output by the whole connecting layer at the tail end of the VGGNet convolutional neural network through the softmax classifier and the Random Forest classifier so as to respectively obtain a first classification probability value and a second classification probability value;
and comparing the first classification probability value and the second classification probability value with a first probability standard value, and confirming a defect detection result according to a comparison result.
3. The method for detecting defects in a photovoltaic module according to claim 2, wherein the comparing the first classification probability value and the second classification probability value with a first probability standard value and confirming the defect detection result according to the comparison result comprises:
judging whether the defect categories output by the softmax classifier and the Random Forest classifier are the same or not;
if not, carrying out exception prompt;
if so, judging whether the first classification probability value is larger than the first probability standard value, and judging whether the second classification probability value is larger than the first probability standard value;
if the first classification probability value is larger than the first probability standard value and the second classification probability value is larger than the first probability standard value, confirming that the defect class of the target detection image is the defect class output by the softmax classifier and the Random Forest classifier;
confirming that the defect class of the target detection image is the defect class output by the softmax classifier if the first classification probability value is larger than the first probability standard value and the second classification probability value is smaller than or equal to the first probability standard value;
if the first classification probability value is smaller than or equal to the first probability standard value and the second classification probability value is larger than the first probability standard value, confirming that the defect category of the target detection image is the defect category output by the Random Forest classifier;
and if the first classification probability value is less than or equal to the first probability standard value and the second classification probability value is less than or equal to the first probability standard value, performing low-probability prompt.
4. The method according to claim 1, wherein the target convolutional neural network comprises a fused convolutional neural network, wherein the fused convolutional neural network is composed of a VGGNet convolutional neural network and an AlexNet convolutional neural network, and the defect detection and classification of the target detection image by using the target convolutional neural network comprises:
outputting a third classification probability value and a fourth classification probability value through the VGGNet convolutional neural network and the AlexNet convolutional neural network, respectively, when the target detection image is input into the fused convolutional neural network;
and confirming a defect detection result according to the third classification probability value and the fourth classification probability value.
5. The method for detecting defects in a photovoltaic module according to claim 4, wherein the confirming the defect detection result according to the third classification probability value and the fourth classification probability value comprises:
judging whether the third classification probability value is larger than a second probability standard value or not;
if so, confirming that the defect type of the target detection image is the defect type output by the VGGNet convolutional neural network;
if not, judging whether the fourth classification probability value is larger than a third probability standard value or not;
if so, confirming that the defect type of the target detection image is the defect type output by the AlexNet convolutional neural network;
if not, judging whether the third classification probability value is larger than the fourth classification probability value or not;
if so, confirming that the defect type of the target detection image is the defect type output by the VGGNet convolutional neural network;
and if not, determining that the defect type of the target detection image is the defect type output by the AlexNet convolutional neural network.
6. A photovoltaic module defect detection system, comprising: a conveying mechanism, an image shooting mechanism, an image transmission component and an image analysis component, wherein,
the conveying mechanisms are distributed on two sides of the image shooting mechanism and are used for conveying the photovoltaic module to be detected to the processing area;
the image shooting mechanism is used for acquiring an EL image of the photovoltaic module to be detected;
the image transmission element is used for receiving the EL image of the photovoltaic module to be detected and transmitting the EL image of the photovoltaic module to be detected to the image analysis element;
the image analysis element is used for carrying out fusion filtering denoising processing on the EL image to obtain a first detection image, carrying out position correction processing on the first detection image to obtain a second detection image, carrying out ROI (region of interest) division processing on the second detection image to obtain a third detection image, carrying out image enhancement processing on the third detection image to obtain a fourth detection image, carrying out image segmentation processing on the fourth detection image to obtain a target detection image, and carrying out defect detection and classification on the target detection image by adopting a target convolutional neural network; wherein the content of the first and second substances,
the image capturing mechanism is also used for marking the defect position.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111506090.8A CN114187260A (en) | 2021-12-10 | 2021-12-10 | Photovoltaic module defect detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111506090.8A CN114187260A (en) | 2021-12-10 | 2021-12-10 | Photovoltaic module defect detection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114187260A true CN114187260A (en) | 2022-03-15 |
Family
ID=80543039
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111506090.8A Pending CN114187260A (en) | 2021-12-10 | 2021-12-10 | Photovoltaic module defect detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114187260A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117173178A (en) * | 2023-11-02 | 2023-12-05 | 南通逸飞智能科技有限公司 | Photovoltaic device processing detection method and system |
-
2021
- 2021-12-10 CN CN202111506090.8A patent/CN114187260A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117173178A (en) * | 2023-11-02 | 2023-12-05 | 南通逸飞智能科技有限公司 | Photovoltaic device processing detection method and system |
| CN117173178B (en) * | 2023-11-02 | 2024-04-05 | 南通逸飞智能科技有限公司 | Photovoltaic device processing detection method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111275679B (en) | Image-based solar cell defect detection system and method | |
| Di Tommaso et al. | A multi-stage model based on YOLOv3 for defect detection in PV panels based on IR and visible imaging by unmanned aerial vehicle | |
| CN114187260A (en) | Photovoltaic module defect detection method and system | |
| WO2023168984A1 (en) | Area-array camera-based quality inspection method and system for cathode copper | |
| CN102901735B (en) | System for carrying out automatic detections upon workpiece defect, cracking, and deformation by using computer | |
| JP2010135446A (en) | Apparatus and method for inspecting solar battery cell, and recording medium having program of the method recorded thereon | |
| Chindarkkar et al. | Deep learning based detection of cracks in electroluminescence images of fielded PV modules | |
| Nian et al. | Automatic detection of defects in solar modules: image processing in detecting | |
| CN111260617A (en) | Solar cell panel defect detection method based on deep learning | |
| Dhimish et al. | Ultrafast high-resolution solar cell cracks detection process | |
| CN114636706A (en) | Comprehensive method and device for image detection of solar cell after film coating | |
| CN115294031A (en) | Photovoltaic module fault image identification method based on infrared thermal imaging analysis | |
| Al-Mashhadani et al. | Deep learning methods for solar fault detection and classification: a review | |
| Aghamohammadi et al. | Solar cell panel crack detection using particle swarm optimization algorithm | |
| Rodriguez et al. | Automatic solar cell diagnosis and treatment | |
| US20230042106A1 (en) | System and method for the statistical analysis of images of photovoltaic panels | |
| CN116297199A (en) | Detection system for automatically detecting lithium battery parameters based on machine vision | |
| CN114359155A (en) | Film laminating method and system | |
| CN114663393A (en) | Convolutional neural network-based detection method for defect of poor spacing between photovoltaic modules | |
| WO2010041723A1 (en) | Inspection device for solar batteries, inspection method for solar batteries, program, inspection system for solar batteries | |
| CN112541478A (en) | Insulator string stain detection method and system based on binocular camera | |
| CN112561875A (en) | Photovoltaic cell panel coarse grid detection method based on artificial intelligence | |
| CN113965164B (en) | Solar cell module defect detection system and method based on convolutional neural network | |
| Ahan et al. | AI-assisted cell-level fault detection and localization in solar PV electroluminescence images | |
| CN113953208B (en) | Full-automatic sorting device and method for electronic components |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |