CN117094964A - Battery piece spacing detection method and device, computer equipment and storage medium - Google Patents
Battery piece spacing detection method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN117094964A CN117094964A CN202311041520.2A CN202311041520A CN117094964A CN 117094964 A CN117094964 A CN 117094964A CN 202311041520 A CN202311041520 A CN 202311041520A CN 117094964 A CN117094964 A CN 117094964A
- Authority
- CN
- China
- Prior art keywords
- spacing
- image
- target
- training
- battery piece
- 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 168
- 238000000034 method Methods 0.000 claims abstract description 30
- 238000012549 training Methods 0.000 claims description 165
- 238000002372 labelling Methods 0.000 claims description 32
- 238000004590 computer program Methods 0.000 claims description 16
- 230000004913 activation Effects 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 abstract description 6
- 210000004027 cell Anatomy 0.000 description 135
- 230000006870 function Effects 0.000 description 11
- 210000002287 horizontal cell Anatomy 0.000 description 11
- 238000010586 diagram Methods 0.000 description 8
- 238000012545 processing Methods 0.000 description 7
- 230000006978 adaptation Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- 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
- G06T2207/10004—Still image; Photographic 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
- G06T2207/20081—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]
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a battery piece spacing detection method, a device, computer equipment and a storage medium, wherein the method comprises the following steps: acquiring a target cell spacing image of a photovoltaic module to be detected; acquiring a trained target distance detection model; and inputting the target cell spacing image into a target spacing detection model to obtain a target spacing value corresponding to the target cell spacing image output by the target spacing detection model. The trained interval detection model is adopted to finish the interval detection of the battery pieces, and the learning capacity of the neural network is utilized, so that the problem of poor robustness of the traditional mode is effectively solved, various interval images of the battery pieces can be well adapted, and the accuracy of the interval detection of the battery pieces is finally improved.
Description
Technical Field
The present invention relates to the field of image processing technologies, and in particular, to a method and apparatus for detecting a distance between battery slices, a computer device, and a storage medium.
Background
The photovoltaic module appearance detection comprises distance detection between the battery pieces, including upper and lower distances and left and right distances.
Most of the existing methods adopt traditional images to analyze the spacing of the battery pieces, wherein the defects are obvious, the robustness is poor in the prior art, and the defects are very influenced by the image quality and the position diversity of the battery pieces.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a battery piece spacing detection method, a device, computer equipment and a storage medium.
In a first aspect, in one embodiment, the present invention provides a method for detecting a spacing between battery cells, including:
acquiring a target cell spacing image of a photovoltaic module to be detected;
acquiring a trained target distance detection model;
and inputting the target cell spacing image into a target spacing detection model to obtain a target spacing value corresponding to the target cell spacing image output by the target spacing detection model.
In one embodiment, before the step of obtaining the trained target pitch detection model, the method for detecting a cell pitch further includes:
determining target shape characteristics of a cell spacing image in a photovoltaic module; the target shape feature characterizes one of a height of the battery cell pitch image being less than a width thereof or a height of the battery cell pitch image being greater than a width thereof;
Determining a target convolution kernel according to the target shape characteristics;
constructing and obtaining an initial interval detection model according to the target convolution kernel;
training the initial distance detection model to obtain a target distance detection model.
In one embodiment, obtaining a target cell pitch image in a photovoltaic module to be detected includes:
acquiring an initial cell spacing image in a photovoltaic module to be detected;
detecting actual shape characteristics of an initial cell spacing image;
if the actual shape characteristic of the initial cell spacing image is matched with the target shape characteristic, the initial cell spacing image is used as the target cell spacing image;
and if the actual shape characteristic of the initial battery piece spacing image is not matched with the target shape characteristic, rotating the initial battery piece spacing image to obtain a target battery piece spacing image with the actual shape characteristic matched with the target shape characteristic.
In one embodiment, the height of the target shape feature characterizing the cell pitch image is less than its width; determining a target convolution kernel from the target shape feature, comprising:
determining a convolution kernel 3*8, a convolution kernel 5*8, a convolution kernel 7*8 and a convolution kernel 1*1 according to the target shape characteristic;
The convolution combinations of convolution kernels 3*8, 5*8, 7*8, and 1*1 are referred to as target convolution kernels.
In one embodiment, constructing an initial pitch detection model from the target convolution kernel includes:
constructing a plurality of convolution modules, a full connection module and an output module which are sequentially connected; the full connection module comprises a full connection layer and an activation layer;
and establishing the jump connection between at least two convolution modules in the plurality of convolution modules and the full connection module.
In one embodiment, training the initial pitch detection model to obtain the target pitch detection model includes:
acquiring a training sample set; the training sample set comprises a plurality of training battery piece interval images and labeling interval values corresponding to each training battery piece interval image;
inputting each training battery piece spacing image in the training sample set into an initial spacing detection model to obtain a predicted spacing value corresponding to each training battery piece spacing image output by the initial spacing detection model;
determining training loss according to the predicted interval value and the marked interval value corresponding to each training battery piece interval image;
and training the initial distance detection model according to the training loss to obtain a trained target distance detection model.
In one embodiment, determining the training loss based on the predicted pitch value and the labeled pitch value corresponding to each training battery cell pitch image comprises:
aiming at the predicted interval value and the marked interval value corresponding to each training battery piece interval image,
if the actual spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image is smaller than the preset difference value, determining a first target spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image in a first mode; the first target pitch difference is not greater than the actual pitch difference;
if the actual spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image is not smaller than the preset difference value, determining a second target spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image in a second mode; the second target pitch difference is greater than the actual pitch difference;
and obtaining training loss according to the first target interval difference value or the second target interval difference value corresponding to each training battery piece interval image.
In one embodiment, obtaining a training sample set includes:
Acquiring an initial training battery piece interval image set;
performing image enhancement on each training battery piece spacing image in the initial training battery piece spacing image set to obtain an enhanced training battery piece spacing image set; ways of image enhancement include randomly adding occlusions, randomly rotating, randomly reversing, randomly adding blur, and/or randomly shifting;
labeling each training battery piece interval image in the initial training battery piece interval image set and the enhancement training battery piece interval image set to obtain an initial labeling interval value set corresponding to the initial training battery piece interval image set and an enhancement labeling interval value set corresponding to the enhancement training battery piece image set;
and obtaining a training sample set according to the initial training battery piece spacing image set, each training battery piece spacing image in the enhanced training battery piece spacing image set, and each labeling spacing value in the initial labeling spacing value set and the enhanced labeling spacing value set.
In a second aspect, in one embodiment, the present invention provides a device for detecting a distance between battery cells, including:
the image acquisition module is used for acquiring a target cell spacing image of the photovoltaic module to be detected;
the model acquisition module is used for acquiring a trained target interval detection model;
The numerical value detection module is used for inputting the target cell spacing image into the target spacing detection model to obtain a target spacing numerical value corresponding to the target cell spacing image output by the target spacing detection model.
In a third aspect, in one embodiment, the invention provides a computer device comprising a memory and a processor; the memory stores a computer program, and the processor is configured to execute the computer program in the memory to perform the steps in the method for detecting a pitch of a battery cell according to any one of the embodiments described above.
In a fourth aspect, in one embodiment, the present invention provides a storage medium storing a computer program, the computer program being loaded by a processor to perform the steps in the battery cell pitch detection method in any one of the embodiments described above.
Through the battery piece spacing detection method, the device, the computer equipment and the storage medium, the trained spacing detection model is adopted to finish battery piece spacing detection, and the learning capacity of the neural network is utilized, so that the problem of poor robustness in the traditional mode is effectively solved, various battery piece spacing images can be well adapted, and finally the accuracy of battery piece spacing detection is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic diagram of an application scenario of a method for detecting a distance between battery slices according to an embodiment of the invention;
FIG. 2 is a flowchart of a method for detecting a distance between battery cells according to an embodiment of the invention;
FIG. 3 is a schematic view of a cell pitch area in an image of a photovoltaic module according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a cell pitch image according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of detecting a battery cell image and a battery cell pitch image according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a middle distance detection model according to an embodiment of the present invention;
FIG. 7 is a schematic diagram of training a battery cell pitch image and a binarized image in an embodiment of the invention;
FIG. 8 is a schematic diagram of a device for detecting a distance between battery cells according to an embodiment of the invention;
Fig. 9 is a schematic diagram of a computer device according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to fall within the scope of the application.
In the description of the present application, it should be understood that the terms "first," "second," and the like are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more of the described features. In the description of the present application, the meaning of "a plurality" is two or more, unless explicitly defined otherwise. In the present application, the term "exemplary" is used to mean "serving as an example, instance, or illustration. Any embodiment described as "exemplary" in this disclosure is not necessarily to be construed as preferred or advantageous over other embodiments. The following description is presented to enable any person skilled in the art to make and use the application. In the following description, details are set forth for purposes of explanation. It will be apparent to one of ordinary skill in the art that the present application may be practiced without these specific details. In other instances, well-known structures and processes have not been described in detail so as not to obscure the description of the application with unnecessary detail. Thus, the present application is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.
The battery piece spacing detection method is applied to a battery piece spacing detection device which is arranged in computer equipment; the computer device may be a terminal, for example, a mobile phone or a tablet computer, and the computer device may also be a server, or a service cluster formed by a plurality of servers.
As shown in fig. 1, fig. 1 is a schematic diagram of an application scenario of a battery spacing detection method according to an embodiment of the present invention, where the application scenario of the battery spacing detection method according to the embodiment of the present invention includes a computer device 100 (a battery spacing detection device is integrated in the computer device 100), and a computer readable storage medium corresponding to the battery spacing detection method is run in the computer device 100 to execute steps of the battery spacing detection method.
It can be understood that the computer device in the application scenario of the battery spacing detection method shown in fig. 1, or the apparatus included in the computer device, is not limited to the embodiment of the present invention, that is, the number of devices and the type of devices included in the application scenario of the battery spacing detection method, or the number of apparatuses and the type of apparatuses included in each device do not affect the overall implementation of the technical solution in the embodiment of the present invention, and all the methods can be calculated as equivalent substitutions or derivatives of the technical solution claimed in the embodiment of the present invention.
The computer device 100 in the embodiment of the present invention may be an independent device, or may be a device network or a device cluster formed by devices, for example, the computer device 100 described in the embodiment of the present invention includes, but is not limited to, a computer, a network host, a single network device, a plurality of network device sets, or a cloud device formed by a plurality of devices. Wherein, cloud equipment is composed of a large number of computers or network equipment based on Cloud Computing (Cloud Computing).
It will be understood by those skilled in the art that the application scenario shown in fig. 1 is only one application scenario corresponding to the technical solution of the present invention, and does not limit the application scenario of the technical solution of the present invention, other application scenarios may also include more or fewer computer devices than those shown in fig. 1, or a network connection relationship of the computer devices, for example, only 1 computer device is shown in fig. 1, and it is understood that the scenario of the battery inter-chip distance detection method may also include one or more other computer devices, which is not limited herein in particular; the computer device 100 may further include a memory for storing information related to the battery cell pitch detection method.
In addition, in the application scenario of the battery cell pitch detection method in the embodiment of the present invention, the computer device 100 may be provided with a display device, or the computer device 100 is not provided with a display device and is connected to the external display device 200 in a communication manner, where the display device 200 is configured to output a result of executing the battery cell pitch detection method in the computer device. The computer device 100 may access a background database 300 (the background database 300 may be a local memory of the computer device 100, and the background database 300 may also be disposed in the cloud), where information related to the battery spacing detection method is stored in the background database 300.
It should be noted that, the application scenario of the battery piece spacing detection method shown in fig. 1 is only an example, and the application scenario of the battery piece spacing detection method described in the embodiment of the present invention is for more clearly describing the technical solution of the embodiment of the present invention, and does not constitute a limitation to the technical solution provided by the embodiment of the present invention.
Based on the application scenario of the battery piece spacing detection method, an embodiment of the battery piece spacing detection method is provided.
In a first aspect, as shown in fig. 2, in an embodiment, the present invention provides a method for detecting a spacing between battery cells, including:
step 201, obtaining a target cell spacing image in a photovoltaic module to be detected;
the photovoltaic modules generally comprise a plurality of battery pieces which are arranged in rows and columns, namely, a battery piece spacing image in each photovoltaic module comprises a vertical battery piece spacing image between two adjacent battery pieces in the row direction and a horizontal battery piece spacing image between two adjacent batteries in the column direction; specifically, as shown in fig. 3, in the left part in fig. 3, the upper and lower circled areas are corresponding horizontal cell pitch images, and in the right part in fig. 3, the middle circled area is corresponding numerical cell pitch image; the cell spacing images obtained by cutting from the photovoltaic module image are shown in fig. 4, and the cell spacing images with different spacing values are shown in fig. 4, wherein the cut cell spacing images need to be slightly larger than a cell spacing region so as to ensure the subsequent detection effect;
The battery piece spacing image is cut out from the photovoltaic module image, and can be cut out manually or in a network model cutting mode; for manual cutting, an operator can introduce the photovoltaic module image into the image processing software through corresponding image processing software, and the image corresponding to the cell spacing area is circled in the photovoltaic module image according to the cutting requirement, so that the required cell spacing image is obtained; for network model cutting, an image cutting model can be obtained through training in advance, and a photovoltaic module image is input into the trained image cutting model, so that a battery piece interval image output by the image cutting model is obtained;
in this embodiment, since the interval value corresponding to the battery piece interval image is finally required to be obtained, the cut battery piece interval image is required to contain corresponding coordinate information or size information, so that when the interval detection is performed on the battery piece interval image subsequently, corresponding conversion calculation can be performed based on the coordinate signal or the size information of the battery piece interval image as a reference; specifically, whether the clipping is performed manually or by a network model, the coordinate information or the size information of the battery piece image in the photovoltaic module image can be determined according to the coordinate information or the size information of the photovoltaic module image, as shown in fig. 5, the determined coordinate information or the size information of the battery piece image in the photovoltaic module image can be represented as a large rectangular frame (including a red frame and a green frame) in fig. 5, the actual size of the large rectangular frame is the size information of the battery piece image in the photovoltaic module image, the coordinate information of four corners of the large rectangular frame is the coordinate information of the battery piece image in the photovoltaic module image, and then the coordinate information or the size information of the battery piece spacing image in the photovoltaic module image is determined according to the coordinate information or the size information of the battery piece image in the photovoltaic module image, as shown in fig. 5, the determined coordinate information or the size information of the battery piece spacing image in the photovoltaic module image can be represented as a small rectangular frame (including a blue frame), the actual size of the small rectangular frame is the size information of the battery piece spacing image in the photovoltaic module image, and the four corners of the small rectangular frame are the coordinate information of the battery piece spacing image in the photovoltaic module image;
Step 202, obtaining a trained target distance detection model;
the trained target interval detection model can be packaged and encapsulated into a computer executable file, and is deployed in an execution main body (namely computer equipment) of the battery piece interval detection method in the embodiment, when interval detection is required to be carried out on a target battery piece interval image obtained by cutting, the target interval detection model on the encapsulated computer equipment is directly started;
step 203, inputting the target cell pitch image into a target pitch detection model to obtain a target pitch value corresponding to the target cell pitch image output by the target pitch detection model;
in this step, after the target cell pitch image is input to the target pitch detection model, the target pitch detection model can perform pitch detection on the target cell pitch image according to the learned detection capability, and finally output the target pitch value corresponding to the target cell pitch image.
According to the battery piece spacing detection method, the trained spacing detection model is adopted to finish battery piece spacing detection, and the learning capacity of the neural network is utilized, so that the problem of poor robustness of a traditional mode is effectively solved, various battery piece spacing images can be well adapted, and finally the accuracy of battery piece spacing detection is improved.
In one embodiment, before the step of obtaining the trained target pitch detection model, the method for detecting a cell pitch further includes:
determining target shape characteristics of a cell spacing image in a photovoltaic module;
wherein the target shape feature characterizes one of a height of the battery cell pitch image being less than a width thereof or a height of the battery cell pitch image being greater than a width thereof;
the target distance detection model in the embodiment adopts a convolutional neural network, and convolution kernels with different specifications have different performances when processing the same image; for the target cell pitch image, in order to improve the effect of the convolution kernel, it is also necessary to select the convolution kernel that matches the target shape feature of the target cell pitch image; as shown in fig. 4 and 5, the cell pitch image is long, its width is narrow, its length is relatively long, and considering the placement position of the cell pitch image, it may include the above-mentioned vertical cell pitch image and horizontal cell pitch image, if the shape features of the cell pitch image are described in the vertical direction and the horizontal direction (i.e., the dimension in the vertical direction is collectively referred to as "high" and the dimension in the horizontal direction is referred to as "wide"), the shape features of the vertical cell pitch image represent that the width of the cell pitch image is smaller than the height thereof, and conversely, the shape features of the horizontal cell pitch image represent that the height of the cell pitch image is smaller than the width thereof; for the same cell spacing image, the same cell spacing image can be provided with different shape characteristics through rotation, for example, for a horizontal cell spacing image, the cell spacing image is rotated for ninety degrees, so that the cell spacing image can be converted into a vertical cell spacing image, the shape characteristics are converted from height smaller than width to width smaller than height, in the process, the cell spacing image is rotated, the image content is not changed essentially, and the performance of a convolution kernel is affected; therefore, in the case that the cell pitch image in the photovoltaic module image includes both the vertical cell pitch image and the horizontal cell pitch image, one of the horizontal or vertical shape needs to be determined as the target shape, so that the shape feature corresponding to the target shape is taken as the target shape feature, which is convenient for determining the corresponding convolution kernel according to the target shape feature later;
Determining a target convolution kernel according to the target shape characteristics;
after the target shape feature is determined, determining a convolution kernel matched with the target shape feature, so as to obtain a target convolution kernel; if the determined target shape feature is a horizontal shape feature, the determined target convolution kernel may be understood as a horizontal convolution kernel, and if the determined target shape feature is a vertical shape feature, the determined target convolution kernel may be understood as a vertical convolution kernel; if the determined target convolution kernel is a horizontal convolution kernel, the trained target interval detection model has better performance on the horizontal battery piece interval image when the interval detection is carried out, so that the interval detection accuracy of the horizontal battery piece interval image can be improved, and similarly, if the determined target convolution kernel is a vertical convolution kernel, the trained target interval detection model has better performance on the vertical battery piece interval image when the interval detection is carried out, so that the interval detection accuracy of the vertical battery piece interval image can be improved;
constructing and obtaining an initial interval detection model according to the target convolution kernel;
training the initial interval detection model to obtain a target interval detection model;
The target interval detection model constructed and trained based on the horizontal convolution kernel can be understood as a horizontal interval detection model, has good performance on a horizontal battery piece interval image, can improve the detection accuracy, but has relatively poor performance on a vertical battery piece interval image; therefore, in this case, in the present embodiment, it is also possible to construct and train to obtain a horizontal pitch detection model and a vertical pitch detection model based on the horizontal convolution kernel and the vertical convolution kernel, respectively; when the target cell spacing image is a horizontal cell spacing image, the target cell spacing image is input into a horizontal spacing detection model for detection, and when the target cell spacing image is a vertical cell spacing image, the target cell spacing image is input into a vertical spacing detection model for detection, so that the accuracy of the horizontal cell spacing image and the vertical cell spacing image can be improved simultaneously.
In one embodiment, obtaining a target cell pitch image in a photovoltaic module to be detected includes:
acquiring an initial cell spacing image in a photovoltaic module to be detected;
detecting actual shape characteristics of an initial cell spacing image;
If the actual shape characteristic of the initial cell spacing image is matched with the target shape characteristic, the initial cell spacing image is used as the target cell spacing image;
if the actual shape feature of the initial battery piece spacing image is not matched with the target shape feature, rotating the initial battery piece spacing image to obtain a target battery piece spacing image with the actual shape feature matched with the target shape feature;
the above embodiment has mentioned that the horizontal space detection model and the vertical space detection model can be respectively constructed and trained based on the horizontal convolution kernel and the vertical convolution kernel, so that the accuracy of the horizontal cell space image and the vertical cell space image can be improved simultaneously; however, additional construction and training of a network model are more costly and increase the storage burden of computer equipment, so in this embodiment, the obtained cell pitch image of the photovoltaic module to be detected is used as an initial cell pitch image, so as to detect the actual shape feature thereof, that is, whether the cell pitch image is a horizontal cell pitch image or a vertical cell pitch image is judged, if the actual shape feature of the initial cell pitch image is matched with the target shape feature, the initial cell pitch image is used as the target cell pitch image and then is input into the target pitch detection model for pitch detection, otherwise, if the actual shape feature of the initial cell pitch image is not matched with the target shape feature, the initial cell pitch image can be preprocessed, so that the initial cell pitch image is converted into the target cell pitch image with the shape feature matched with the target shape feature and then is input into the target pitch detection model for pitch detection; it has been mentioned in the above embodiments that the shape characteristics of the cell pitch image are changed by rotating it, so that in this embodiment, the specific way of preprocessing can be characterized by rotating the initial cell pitch image, more specifically, the rotation angle can be clockwise (90 ° +180 °. N, where n is 0 or an integer);
If the detected actual shape characteristic of the initial cell pitch image is characterized as the width of the cell pitch image is smaller than the height of the cell pitch image, the initial cell pitch image is rotated by 90 degrees clockwise, and the rotated cell pitch image is used as the target cell pitch image; for example, if the determined target shape characteristic is characterized in that the width of the cell spacing image is smaller than the height of the cell spacing image, if the detected actual shape characteristic of the initial cell spacing image is characterized in that the height of the cell spacing image is smaller than the width of the cell spacing image, the initial cell spacing image is rotated by 90 degrees clockwise, and the rotated cell spacing image is used as the target cell spacing image;
in addition, in other embodiments, when it is detected that the actual shape feature of the initial cell pitch image does not match the target shape feature, if the initial cell pitch image is not subjected to preprocessing such as rotation but two pitch detection models exist, the pitch detection is not performed, and another pitch detection model is acquired again, and further, a subsequent step is performed based on the acquired pitch detection model.
Through the rotation processing of the battery piece interval image, the actual shape characteristic of the battery piece interval image can be matched with the target shape characteristic, and the high-performance detection of the battery piece interval image with different shape characteristics is realized on the basis of a space detection model.
As shown in fig. 6, in one embodiment, the height of the target shape feature characterizing the cell pitch image is less than its width; determining a target convolution kernel from the target shape feature, comprising:
determining a convolution kernel 3*8, a convolution kernel 5*8, a convolution kernel 7*8 and a convolution kernel 1*1 according to the target shape characteristic;
a convolution combination of the convolution kernel 3*8, the convolution kernel 5*8, the convolution kernel 7*8, and the convolution kernel 1*1 is used as a target convolution kernel;
the height of the target shape characteristic representation cell spacing image is smaller than the width thereof, namely the description is mainly applicable to the horizontal cell spacing image, and the analysis and research of the horizontal cell spacing image are carried out, and the target convolution kernel in the embodiment is obtained through preliminary selection, test and final selection, so that the accuracy of final spacing detection is high based on the obtained target convolution kernel.
As shown in fig. 6, in one embodiment, constructing an initial pitch detection model from a target convolution kernel includes:
Constructing a plurality of convolution modules (in fig. 6, a space detection model comprises five convolution modules, each convolution module comprises four convolution layers, and the four convolution layers sequentially adopt convolution kernels 3*8, 5*8, 7*8 and 1*1), a full connection module and an output module (namely Value in fig. 6); the fully connected module includes a fully connected layer (i.e., FC layer in fig. 6, i.e., fully Connected Layer) and an active layer (i.e., relu in fig. 6, one of the activation functions);
establishing jump connection between at least two convolution modules in the plurality of convolution modules and the full connection module;
in fig. 6, the outputs of the second convolution module, the third convolution module and the fourth convolution module are respectively spliced with the output of the fifth convolution module, and then the spliced outputs are output to the full-connection module, so that multi-scale features are enriched, and the accuracy of the subsequent interval detection is further improved; in other embodiments, other ways of jump connection may be employed, such as splicing only the output of the second convolution module with the output of the fifth convolution module; because the first convolution module has limited extracted characteristics, the detection capability of the model is not greatly assisted, the calculation amount and the training speed are comprehensively considered, and the output of the first convolution module is not spliced;
The accuracy of detection can be mentioned by splicing the active layers on the full connection layer, so that more accurate distance values are output.
In one embodiment, training the initial pitch detection model to obtain the target pitch detection model includes:
acquiring a training sample set; the training sample set comprises a plurality of training battery piece interval images and labeling interval values corresponding to each training battery piece interval image;
the specific manner of training the battery piece spacing image may refer to the step of acquiring the target battery piece spacing image in the above embodiment, which is not described herein again;
the method comprises the steps that a annotator can acquire corresponding annotation interval values through a manual measurement mode, then establish a corresponding relation with training battery piece interval images, and also acquire the corresponding annotation interval values through an algorithm mode for training battery piece interval images with higher image definition and correct positions, specifically, as shown in fig. 7, the algorithm can acquire training battery piece interval images (namely upper side images in fig. 7) firstly, then perform binarization processing on the training battery piece interval images to acquire corresponding binarized images (namely lower side images in fig. 7), and finally calculate the number of pixels with corresponding intervals in the binarized images, so that the corresponding annotation interval values are obtained; it should be noted that, for images with lower definition or uneven spacing, the labeling spacing value obtained by the algorithm may have errors, and at this time, the labeling spacing value needs to be manually selected and modified;
In this embodiment, the distance value is marked, that is, the mode is supervised training, and in other embodiments, non-supervised training may be used;
inputting each training battery piece spacing image in the training sample set into an initial spacing detection model to obtain a predicted spacing value corresponding to each training battery piece spacing image output by the initial spacing detection model;
determining training loss according to the predicted interval value and the marked interval value corresponding to each training battery piece interval image;
in the training process, the space image of the training battery piece is mainly taken as input, the output predicted space value is taken as actual output, and the labeling space value is taken as expected output, so that training loss is determined according to the actual output and the expected output;
when training loss is determined, an MSE loss function or an RMSE loss function can be adopted, so that the number of pixel points can be better returned, namely a distance value can be better determined;
specifically, determining the training loss according to the predicted interval value and the labeling interval value corresponding to each training battery piece interval image includes:
aiming at the predicted interval value and the marked interval value corresponding to each training battery piece interval image,
If the actual spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image is smaller than the preset difference value, determining a first target spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image in a first mode; the first target pitch difference is not greater than the actual pitch difference;
the first mode is used when the difference between the predicted value and the actual value is smaller, and the obtained first target distance difference is not larger than the actual distance difference, namely the difference is not amplified;
if the actual spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image is not smaller than the preset difference value, determining a second target spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image in a second mode; the second target pitch difference is greater than the actual pitch difference;
the second mode is used when the difference between the predicted value and the actual value is larger, and the obtained second target distance difference is larger than the actual distance difference, namely the difference is amplified to accelerate the convergence speed of training;
more specifically, taking the number of pixels as an example, the preset difference may be three pixels, the first target pitch difference obtained in the first manner is equal to the actual pitch difference, and the second target pitch difference obtained in the second manner is a third power of the actual pitch difference, and the following expression is referred to for specific details:
abs(YGt–Ypd)<3pixel；
abs((YGt–Ypd))^3>＝3pixel；
Wherein YGt represents a predicted pitch value (i.e., predicted value), ypd represents a labeled pitch value (i.e., actual value), and abs represents an absolute value;
obtaining training loss according to the first target spacing difference value or the second target spacing difference value corresponding to each training battery piece spacing image;
after obtaining a first target distance difference value or a second target distance difference value corresponding to each training battery piece distance image, the target distance difference values can be directly averaged to obtain training loss; in other embodiments, each target distance difference may be substituted into the MSE loss function or RMSE loss function to calculate, so as to obtain training loss;
training the initial distance detection model according to the training loss to obtain a trained target distance detection model;
determining whether a preset convergence condition is met according to the training loss, if yes, obtaining a trained target interval detection model, otherwise, adjusting model parameters of an initial interval detection model according to the training loss, then obtaining a next training sample set, and continuing training until the obtained training loss meets the preset convergence condition;
after the trained target pitch detection model is obtained, the verification sample set and the test sample set may also be used to verify the accuracy of the model and the final effect of the test model.
In one embodiment, obtaining a training sample set includes:
acquiring an initial training battery piece interval image set;
performing image enhancement on each training battery piece spacing image in the initial training battery piece spacing image set to obtain an enhanced training battery piece spacing image set; ways of image enhancement include randomly adding occlusions, randomly rotating, randomly reversing, randomly adding blur, and/or randomly shifting;
labeling each training battery piece interval image in the initial training battery piece interval image set and the enhancement training battery piece interval image set to obtain an initial labeling interval value set corresponding to the initial training battery piece interval image set and an enhancement labeling interval value set corresponding to the enhancement training battery piece image set;
obtaining a training sample set according to the initial training battery piece spacing image set, each training battery piece spacing image in the enhanced training battery piece spacing image set, and each labeling spacing value in the initial labeling spacing value set and the enhanced labeling spacing value set;
it should be noted that, the image enhancement mode of random rotation does not conflict with the shape maintaining feature, and the random rotation is only performed within a small range, such as clockwise rotation by 1 ° -5 ° or the like, which considers that the spacing area in the actually obtained cell spacing image is not necessarily completely horizontal, and a certain deviation may exist, so that the samples are enriched by the random rotation mode to improve the robustness of the model;
Wherein, adding the shielding randomly comprises adding black blocks, white blocks, mosaic blocks and the like, reversing randomly comprises reversing upside down and reversing left and right, and the maximum limit of random offset needs to ensure that a spacing area is in the image.
In a second aspect, as shown in fig. 8, in one embodiment, a device for detecting a distance between battery cells according to the present invention includes:
the image acquisition module 301 is configured to acquire a target cell pitch image of a photovoltaic module to be detected;
the model acquisition module 302 is configured to acquire a trained target distance detection model;
the numerical value detection module 303 is configured to input the target cell pitch image into the target pitch detection model, and obtain a target pitch numerical value corresponding to the target cell pitch image output by the target pitch detection model.
Through above-mentioned battery piece interval detection device, adopt trained interval detection model to accomplish battery piece interval detection, utilize neural network's learning ability to effectively solve the problem that traditional mode robustness is poor, adaptation multiple battery piece interval image that can be fine, finally improved the accuracy that battery piece interval detected.
In one embodiment, the device for detecting a distance between battery cells further includes:
The model building module is used for determining target shape characteristics of the cell spacing image in the photovoltaic module before the step of acquiring the trained target spacing detection model; the target shape feature characterizes one of a height of the battery cell pitch image being less than a width thereof or a height of the battery cell pitch image being greater than a width thereof; determining a target convolution kernel according to the target shape characteristics; constructing and obtaining an initial interval detection model according to the target convolution kernel; training the initial distance detection model to obtain a target distance detection model.
In one embodiment, the image acquisition module is specifically configured to acquire an initial cell pitch image in the photovoltaic module to be detected; detecting actual shape characteristics of an initial cell spacing image; if the actual shape characteristic of the initial cell spacing image is matched with the target shape characteristic, the initial cell spacing image is used as the target cell spacing image; and if the actual shape characteristic of the initial battery piece spacing image is not matched with the target shape characteristic, rotating the initial battery piece spacing image to obtain a target battery piece spacing image with the actual shape characteristic matched with the target shape characteristic.
In one embodiment, the height of the target shape feature characterizing the cell pitch image is less than its width; the model building module is specifically configured to determine and obtain a convolution kernel 3*8, a convolution kernel 5*8, a convolution kernel 7*8, and a convolution kernel 1*1 according to the target shape feature; the convolution combinations of convolution kernels 3*8, 5*8, 7*8, and 1*1 are referred to as target convolution kernels.
In one embodiment, the model building module is specifically configured to build a plurality of convolution modules, a full connection module, and an output module that are sequentially connected; the full connection module comprises a full connection layer and an activation layer; and establishing the jump connection between at least two convolution modules in the plurality of convolution modules and the full connection module.
In one embodiment, the model building module is specifically configured to obtain a training sample set; the training sample set comprises a plurality of training battery piece interval images and labeling interval values corresponding to each training battery piece interval image; inputting each training battery piece spacing image in the training sample set into an initial spacing detection model to obtain a predicted spacing value corresponding to each training battery piece spacing image output by the initial spacing detection model; determining training loss according to the predicted interval value and the marked interval value corresponding to each training battery piece interval image; and training the initial distance detection model according to the training loss to obtain a trained target distance detection model.
In one embodiment, the model building module is specifically configured to determine, for each of the predicted pitch value and the labeled pitch value corresponding to the pitch image of the training battery, a first target pitch difference value between the predicted pitch value and the labeled pitch value corresponding to the pitch image of the training battery in a first manner if an actual pitch difference value between the predicted pitch value and the labeled pitch value corresponding to the pitch image of the training battery is less than a preset difference value; the first target pitch difference is not greater than the actual pitch difference; if the actual spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image is not smaller than the preset difference value, determining a second target spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image in a second mode; the second target pitch difference is greater than the actual pitch difference; and obtaining training loss according to the first target interval difference value or the second target interval difference value corresponding to each training battery piece interval image.
In one embodiment, the model building module is specifically configured to obtain an initial training battery piece spacing image set; performing image enhancement on each training battery piece spacing image in the initial training battery piece spacing image set to obtain an enhanced training battery piece spacing image set; ways of image enhancement include randomly adding occlusions, randomly rotating, randomly reversing, randomly adding blur, and/or randomly shifting; labeling each training battery piece interval image in the initial training battery piece interval image set and the enhancement training battery piece interval image set to obtain an initial labeling interval value set corresponding to the initial training battery piece interval image set and an enhancement labeling interval value set corresponding to the enhancement training battery piece image set; and obtaining a training sample set according to the initial training battery piece spacing image set, each training battery piece spacing image in the enhanced training battery piece spacing image set, and each labeling spacing value in the initial labeling spacing value set and the enhanced labeling spacing value set.
In a third aspect, in one embodiment, the present invention provides a computer device, which is the model training computer device or the light leakage detection computer device in the above embodiment, as shown in fig. 9, which shows a structure of the computer device according to the present invention, in particular:
the computer device may include one or more processors 401 of a processing core, memory 402 of one or more computer readable storage media, a power supply 403, and an input unit 404, among other components. Those skilled in the art will appreciate that the architecture of the computer device shown in fig. 9 is not limiting of the computer device, and may include more or fewer components than shown, or may combine certain components, or a different arrangement of components. Wherein:
the processor 401 is a control center of the computer device, connects various parts of the entire computer device using various interfaces and lines, and performs various functions of the computer device and processes data by running or executing software programs and/or modules stored in the memory 402, and calling data stored in the memory 402, thereby performing overall monitoring of the computer device. Optionally, processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly processes an operating system, a user interface, a computer program, etc., and the modem processor mainly processes wireless communication. It will be appreciated that the modem processor described above may not be integrated into the processor 401.
The memory 402 may be used to store software programs and modules, and the processor 401 executes various functional applications and data processing by executing the software programs and modules stored in the memory 402. The memory 402 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, a computer program (such as a sound playing function, an image playing function, etc.) required for at least one function, and the like; the storage data area may store data created according to the use of the server, etc. In addition, memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.
The computer device further comprises a power supply 403 for supplying power to the various components, preferably the power supply 403 may be logically connected to the processor 401 by a power management system, so that functions of charge, discharge, and power consumption management may be performed by the power management system. The power supply 403 may also include one or more of any of a direct current or alternating current power supply, a recharging system, a power failure detection circuit, a power converter or inverter, a power status indicator, and the like.
The computer device may also include an input unit 404, which input unit 404 may be used to receive input numeric or character information and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.
Although not shown, the computer device may further include a display unit or the like, which is not described herein. In particular, in this embodiment, when the computer device trains the computer device for the model, the processor 401 in the computer device loads executable files corresponding to the processes of one or more computer programs into the memory 402 according to the following instructions, and the processor 401 executes the computer programs stored in the memory 402 to perform the following steps:
acquiring a target cell spacing image of a photovoltaic module to be detected;
acquiring a trained target distance detection model;
and inputting the target cell spacing image into a target spacing detection model to obtain a target spacing value corresponding to the target cell spacing image output by the target spacing detection model.
Through above-mentioned computer equipment, adopt trained interval detection model to accomplish battery piece interval detection, utilize neural network's learning ability to effectively solve the problem that traditional mode robustness is poor, adaptation multiple battery piece interval image that can be fine has finally improved the accuracy that battery piece interval detected.
It will be appreciated by those of ordinary skill in the art that all or part of the steps of any of the methods of the above embodiments may be performed by a computer program, or by computer program control related hardware, which may be stored in a computer readable storage medium and loaded and executed by a processor.
In a fourth aspect, in one embodiment, the present invention provides a storage medium having stored therein a plurality of computer programs, the computer programs being loadable by a processor, to perform the steps of:
acquiring a target cell spacing image of a photovoltaic module to be detected;
acquiring a trained target distance detection model;
and inputting the target cell spacing image into a target spacing detection model to obtain a target spacing value corresponding to the target cell spacing image output by the target spacing detection model.
Through above-mentioned storage medium, adopt trained interval detection model to accomplish battery piece interval detection, utilize neural network's learning ability to effectively solve the problem that traditional mode robustness is poor, adaptation multiple battery piece interval image that can be fine has finally improved the accuracy that battery piece interval detected.
It will be appreciated by those of ordinary skill in the art that any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink), DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The steps in the method for detecting the distance between the battery cells in any embodiment provided by the present invention can be executed by the computer program stored in the storage medium, so that the beneficial effects that can be achieved by the method for detecting the distance between the battery cells in any embodiment provided by the present invention can be achieved, which are detailed in the previous embodiments and are not repeated herein.
The specific implementation of each operation above may be referred to the previous embodiments, and will not be described herein.
In the foregoing embodiments, the descriptions of the embodiments are focused on, and the portions of one embodiment that are not described in detail in the foregoing embodiments may be referred to in the foregoing detailed description of other embodiments, which are not described herein again.
The above description of the method, the device, the computer equipment and the storage medium for detecting the spacing between the battery cells provided by the invention applies specific examples to illustrate the principles and the implementation of the invention, and the description of the above examples is only used for helping to understand the method and the core idea of the invention; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in light of the ideas of the present invention, the present description should not be construed as limiting the present invention.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
Claims (11)
1. The method for detecting the spacing between the battery pieces is characterized by comprising the following steps of:
acquiring a target cell spacing image of a photovoltaic module to be detected;
acquiring a trained target distance detection model;
and inputting the target cell spacing image into the target spacing detection model to obtain a target spacing value corresponding to the target cell spacing image output by the target spacing detection model.
2. The battery cell pitch detection method according to claim 1, further comprising, prior to the step of acquiring the trained target pitch detection model:
determining target shape characteristics of a cell spacing image in a photovoltaic module; the target shape feature characterizes one of a height of the battery cell pitch image being less than a width thereof or a height of the battery cell pitch image being greater than a width thereof;
determining a target convolution kernel according to the target shape characteristics;
constructing and obtaining an initial interval detection model according to the target convolution kernel;
training the initial distance detection model to obtain the target distance detection model.
3. The method for detecting a pitch of a battery cell according to claim 2, wherein the acquiring the image of the pitch of the target battery cell in the photovoltaic module to be detected includes:
Acquiring an initial cell spacing image in a photovoltaic module to be detected;
detecting the actual shape characteristics of the initial cell spacing image;
if the actual shape characteristic of the initial cell spacing image is matched with the target shape characteristic, the initial cell spacing image is used as the target cell spacing image;
and if the actual shape characteristic of the initial battery piece spacing image is not matched with the target shape characteristic, rotating the initial battery piece spacing image to obtain the target battery piece spacing image with the actual shape characteristic matched with the target shape characteristic.
4. The battery cell pitch detection method of claim 2, wherein the target shape feature characterizes a height of the battery cell pitch image as less than a width thereof; the determining a target convolution kernel according to the target shape feature comprises the following steps:
determining a convolution kernel 3*8, a convolution kernel 5*8, a convolution kernel 7*8 and a convolution kernel 1*1 according to the target shape feature;
the convolution combinations of the convolution kernel 3*8, the convolution kernel 5*8, the convolution kernel 7*8, and the convolution kernel 1*1 are taken as the target convolution kernels.
5. The method for detecting the spacing between the battery cells according to claim 2, wherein the constructing an initial spacing detection model according to the target convolution kernel includes:
Constructing a plurality of convolution modules, a full connection module and an output module which are sequentially connected; the full-connection module comprises a full-connection layer and an activation layer;
and establishing the jump connection between at least two convolution modules in the plurality of convolution modules and the full connection module.
6. The method for detecting a pitch of a battery cell according to claim 2, wherein the training the initial pitch detection model to obtain the target pitch detection model includes:
acquiring a training sample set; the training sample set comprises a plurality of training battery piece interval images and labeling interval values corresponding to each training battery piece interval image;
inputting each training battery piece spacing image in the training sample set into the initial spacing detection model to obtain a predicted spacing value corresponding to each training battery piece spacing image output by the initial spacing detection model;
determining training loss according to the predicted interval value and the marked interval value corresponding to each training battery piece interval image;
and training the initial distance detection model according to the training loss to obtain the trained target distance detection model.
7. The method of claim 6, wherein determining the training loss according to the predicted and labeled pitch values corresponding to each of the training battery pitch images comprises:
aiming at the predicted interval value and the marked interval value corresponding to each training battery piece interval image,
if the actual spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image is smaller than the preset difference value, determining a first target spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image in a first mode; the first target pitch difference is not greater than the actual pitch difference;
if the actual spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image is not smaller than the preset difference value, determining a second target spacing difference value of the predicted spacing value and the marked spacing value corresponding to the training battery piece spacing image in a second mode; the second target pitch difference is greater than the actual pitch difference;
and obtaining the training loss according to the first target distance difference value or the second target distance difference value corresponding to each training battery piece distance image.
8. The method for detecting a spacing between battery cells according to claim 6, wherein the acquiring a training sample set includes:
acquiring an initial training battery piece interval image set;
performing image enhancement on each training battery piece spacing image in the initial training battery piece spacing image set to obtain an enhanced training battery piece spacing image set; ways of image enhancement include randomly adding occlusions, randomly rotating, randomly reversing, randomly adding blur, and/or randomly shifting;
labeling each training battery piece interval image in the initial training battery piece interval image set and the enhancement training battery piece interval image set to obtain an initial labeling interval value set corresponding to the initial training battery piece interval image set and an enhancement labeling interval value set corresponding to the enhancement training battery piece image set;
and obtaining the training sample set according to the initial training battery piece spacing image set, each training battery piece spacing image in the enhanced training battery piece spacing image set, and each labeling spacing value in the initial labeling spacing value set and the enhanced labeling spacing value set.
9. The utility model provides a battery piece interval detection device which characterized in that includes:
The image acquisition module is used for acquiring a target cell spacing image of the photovoltaic module to be detected;
the model acquisition module is used for acquiring a trained target interval detection model;
the numerical value detection module is used for inputting the target cell spacing image into the target spacing detection model to obtain a target spacing numerical value corresponding to the target cell spacing image output by the target spacing detection model.
10. A computer device comprising a memory and a processor; the memory stores a computer program, and the processor is configured to execute the computer program in the memory to perform the steps in the battery inter-cell distance detection method according to any one of claims 1 to 8.
11. A storage medium storing a computer program to be loaded by a processor to perform the steps in the battery inter-cell distance detection method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311041520.2A CN117094964A (en) | 2023-08-17 | 2023-08-17 | Battery piece spacing detection method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311041520.2A CN117094964A (en) | 2023-08-17 | 2023-08-17 | Battery piece spacing detection method and device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117094964A true CN117094964A (en) | 2023-11-21 |
Family
ID=88776692
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311041520.2A Pending CN117094964A (en) | 2023-08-17 | 2023-08-17 | Battery piece spacing detection method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117094964A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113284133A (en) * | 2021-06-16 | 2021-08-20 | 北京瑞莱智慧科技有限公司 | Method, device and equipment for determining photovoltaic cell piece spacing and storage medium |
| CN114581362A (en) * | 2021-07-22 | 2022-06-03 | 正泰集团研发中心(上海)有限公司 | Photovoltaic module defect detection method and device, electronic equipment and readable storage medium |
| CN114663393A (en) * | 2022-03-22 | 2022-06-24 | 聚时科技(上海)有限公司 | Convolutional neural network-based detection method for defect of poor spacing between photovoltaic modules |
| CN115829922A (en) * | 2022-09-23 | 2023-03-21 | 正泰新能科技有限公司 | Method, device, equipment and medium for detecting space between battery pieces |
| WO2023123003A1 (en) * | 2021-12-29 | 2023-07-06 | 宁德时代新能源科技股份有限公司 | Machine vision detection method, detection apparatus and detection system thereof |
-
2023
- 2023-08-17 CN CN202311041520.2A patent/CN117094964A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113284133A (en) * | 2021-06-16 | 2021-08-20 | 北京瑞莱智慧科技有限公司 | Method, device and equipment for determining photovoltaic cell piece spacing and storage medium |
| CN114581362A (en) * | 2021-07-22 | 2022-06-03 | 正泰集团研发中心(上海)有限公司 | Photovoltaic module defect detection method and device, electronic equipment and readable storage medium |
| WO2023123003A1 (en) * | 2021-12-29 | 2023-07-06 | 宁德时代新能源科技股份有限公司 | Machine vision detection method, detection apparatus and detection system thereof |
| CN114663393A (en) * | 2022-03-22 | 2022-06-24 | 聚时科技(上海)有限公司 | Convolutional neural network-based detection method for defect of poor spacing between photovoltaic modules |
| CN115829922A (en) * | 2022-09-23 | 2023-03-21 | 正泰新能科技有限公司 | Method, device, equipment and medium for detecting space between battery pieces |
Non-Patent Citations (1)
| Title |
|---|
| 韩洋祺，等: "人工智能基础", vol. 1, 31 March 2021, 湖南大学出版社, pages: 193 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112001399B (en) | Image scene classification method and device based on local feature saliency | |
| CN115496892A (en) | Industrial defect detection method and device, electronic equipment and storage medium | |
| CN112330488B (en) | Power grid frequency situation prediction method based on transfer learning | |
| CN116313970A (en) | Wafer pre-alignment method, device, equipment and wafer detection system | |
| CN102713974B (en) | Learning device, recognition device, study recognition system and study recognition device | |
| CN116973788A (en) | Internal short circuit fault detection method, device, computer equipment and storage medium | |
| CN116298897A (en) | Method and apparatus for battery short detection | |
| CN109840353B (en) | Lithium ion battery two-factor inconsistency prediction method and device | |
| CN111340790B (en) | Bounding box determination method, device, computer equipment and storage medium | |
| CN117094964A (en) | Battery piece spacing detection method and device, computer equipment and storage medium | |
| TW202127312A (en) | Image processing method and computer readable medium thereof | |
| CN116188374A (en) | Socket detection method, device, computer equipment and storage medium | |
| CN111985471A (en) | License plate positioning method and device and storage medium | |
| CN115861769A (en) | Training method of plastic shell defect detection model, and plastic shell defect detection method and device | |
| CN113221499B (en) | Mask layout generation method and device, computer equipment and storage medium | |
| CN111551499B (en) | Method and device for measuring sugar content of fruit, computer equipment and storage medium | |
| CN117197248B (en) | Electrolytic capacitor direction judging method, device, equipment and storage medium | |
| CN115063432B (en) | Method and device for identifying alignment mark pattern, computer equipment and storage medium | |
| CN117129029B (en) | Chip detection method and system | |
| CN116736173B (en) | Energy storage battery model construction and energy storage battery state judgment method and device | |
| CN113112008B (en) | Method, apparatus and computer readable storage medium for neural network data quantization | |
| WO2024002394A2 (en) | Method and apparatus for measuring number of target objects, and electronic device and storage medium | |
| CN113033760B (en) | Method, testing device and system for testing operator precision in neural network | |
| CN114116287A (en) | Memory verification method and device for intelligent device, electronic device and storage medium | |
| Zhang et al. | A novel method of battery pack energy health estimation based on visual feature learning |
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 |