CN111340748A - Battery defect identification method and device, computer equipment and storage medium - Google Patents
Battery defect identification method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN111340748A CN111340748A CN201811544225.8A CN201811544225A CN111340748A CN 111340748 A CN111340748 A CN 111340748A CN 201811544225 A CN201811544225 A CN 201811544225A CN 111340748 A CN111340748 A CN 111340748A
- Authority
- CN
- China
- Prior art keywords
- image
- battery
- defect
- recognized
- training
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 241
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000012549 training Methods 0.000 claims description 137
- 238000005070 sampling Methods 0.000 claims description 50
- 238000012360 testing method Methods 0.000 claims description 47
- 210000002569 neuron Anatomy 0.000 claims description 32
- 239000011159 matrix material Substances 0.000 claims description 30
- 238000004590 computer program Methods 0.000 claims description 19
- 230000009467 reduction Effects 0.000 claims description 17
- 239000013598 vector Substances 0.000 claims description 17
- 238000000605 extraction Methods 0.000 claims description 8
- 210000004027 cell Anatomy 0.000 claims description 5
- 238000005401 electroluminescence Methods 0.000 abstract description 4
- 238000001514 detection method Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 8
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000000513 principal component analysis Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000012216 screening Methods 0.000 description 5
- 229910000679 solder Inorganic materials 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 230000002950 deficient Effects 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 210000004205 output neuron Anatomy 0.000 description 3
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000003068 static effect Effects 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
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The application relates to a battery defect identification method, a battery defect identification device, a computer device and a storage medium, wherein the method comprises the following steps: inputting an image to be recognized containing the battery panel characteristics into a trained battery defect recognition model, wherein the image to be recognized is an electroluminescence image, extracting the battery panel characteristics of the image to be recognized through the trained battery defect recognition model, and classifying the image to be recognized according to the battery panel characteristics to obtain the recognition result of the image to be recognized. The battery defect images are identified through the trained identification model, so that the identification speed and the identification accuracy are improved.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to a battery defect identification method and apparatus, a computer device, and a storage medium.
Background
During the manufacturing process of the battery piece, defects such as cracks, unfilled corners, hidden cracks, fragments or broken grids can occur. The subtlety of battery surface defects makes detection extremely difficult. Mainly manual detection and infrared image detection. The manual detection belongs to contact detection, not only brings secondary damage in the detection process, but also reduces the detection precision due to thought errors caused by fatigue, experience and the like. The infrared image detection may be affected by the surrounding environment, resulting in a decrease in detection accuracy.
Disclosure of Invention
In order to solve the technical problem, the application provides a battery defect identification method, a battery defect identification device, a computer device and a storage medium.
A battery defect identification method, comprising:
inputting an image to be recognized containing the characteristics of the battery panel into a trained battery defect recognition model, wherein the image to be recognized is an electroluminescent image;
extracting the battery panel characteristics of the image to be recognized through the trained battery defect recognition model;
and classifying the image to be recognized according to the characteristics of the battery panel to obtain the recognition result of the image to be recognized.
A battery defect identifying apparatus comprising:
the characteristic extraction module is used for inputting the image to be recognized containing the battery panel characteristics into the trained battery defect recognition model, extracting the battery panel characteristics of the image to be recognized through the trained battery defect recognition model, wherein the image to be recognized is an electroluminescent image;
and the identification module is used for classifying the image to be identified according to the characteristics of the battery panel to obtain an identification result of the image to be identified.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:
inputting an image to be recognized containing the characteristics of the battery panel into a trained battery defect recognition model, wherein the image to be recognized is an electroluminescent image;
extracting the battery panel characteristics of the image to be recognized through the trained battery defect recognition model;
and classifying the image to be recognized according to the characteristics of the battery panel to obtain the recognition result of the image to be recognized.
A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:
inputting an image to be recognized containing the characteristics of the battery panel into a trained battery defect recognition model, wherein the image to be recognized is an electroluminescent image;
extracting the battery panel characteristics of the image to be recognized through the trained battery defect recognition model;
and classifying the image to be recognized according to the characteristics of the battery panel to obtain the recognition result of the image to be recognized.
The battery defect identification method, the battery defect identification device, the computer equipment and the storage medium comprise the following steps: inputting an image to be recognized containing the battery panel characteristics into a trained battery defect recognition model, wherein the image to be recognized is an electroluminescence image, extracting the battery panel characteristics of the image to be recognized through the trained battery defect recognition model, and classifying the image to be recognized according to the battery panel characteristics to obtain the recognition result of the image to be recognized. The battery defect images are identified through the trained identification model, so that the identification speed and the identification accuracy are improved.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention.
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive exercise.
FIG. 1 is a diagram illustrating an exemplary application of a battery defect identification method;
FIG. 2 is a schematic flow chart diagram of a battery defect identification method according to an embodiment;
FIG. 3 is a schematic diagram of an embodiment of a battery identification model;
FIG. 4 is a block diagram of an embodiment of a battery defect identifying apparatus;
FIG. 5 is a diagram of the internal structure of one embodiment of a computer device.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all 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 application.
Fig. 1 is a diagram illustrating an application environment of the battery defect identification method according to an embodiment. Referring to fig. 1, the battery defect identifying method is applied to a battery defect identifying system. The battery defect recognition system includes a terminal 110 and a server 120. The terminal 110 and the server 120 are connected through a network. The terminal or the server inputs the image to be recognized containing the battery panel characteristics into the trained battery defect recognition model, extracts the battery panel characteristics of the image to be recognized through the trained battery defect recognition model, and classifies the image to be recognized according to the battery panel characteristics to obtain the recognition result of the image to be recognized. The terminal 110 may specifically be a desktop terminal or a mobile terminal, and the mobile terminal may specifically be at least one of a mobile phone, a tablet computer, a notebook computer, and the like. The server 120 may be implemented as a stand-alone server or a server cluster composed of a plurality of servers.
In one embodiment, as shown in FIG. 2, a battery defect identification method is provided. The embodiment is mainly illustrated by applying the method to the terminal 110 (or the server 120) in fig. 1. Referring to fig. 2, the battery defect identification method specifically includes the following steps:
step S201, inputting the image to be recognized containing the battery plate characteristics into the trained battery defect recognition model.
In particular, the panel characteristics are characteristic data for describing the panel. The image to be recognized refers to an image to be recognized, the image is obtained by shooting through shooting equipment, and each image to be recognized can comprise a finished battery plate, or a part of battery plates, or a plurality of battery plates. The trained battery defect identification model is obtained by training a large number of images carrying defect labels, wherein the trained battery defect identification model can be selected in a self-defined manner, such as a convolutional neural network model, a deep Learning neural network model or an LVQ (Learning vector quantization) neural network.
In one embodiment, the image to be recognized contains the panel features of a plurality of panels, and before inputting the image to be recognized into the trained battery defect recognition model, the method further includes: and segmenting the image to be recognized so as to enable the image to be recognized to only contain the panel feature of one panel.
In one embodiment, the image is an EL (ElectroLuminescence) image of the panel, the image quality of the EL image is high, the quality of the acquired image can be ensured, the image quality seriously affects the identification result of the image, and the type of the image can be identified more accurately only if the image quality is high.
And S202, extracting the battery panel characteristics of the image to be recognized through the trained battery defect recognition model.
And S203, classifying the image to be recognized according to the characteristics of the battery panel to obtain the recognition result of the image to be recognized.
Specifically, the battery plate characteristics of the image to be recognized are extracted through the algorithm of the trained battery defect recognition model, wherein the recognition algorithm can be customized. The extracted panel features include the locations and relative positions of elements on the panel or specific shapes in the panel, etc. And identifying the image to be identified according to the extracted features to obtain a corresponding identification result, wherein the identification result is one of identification types defined by the trained battery defect model, the identification types defined by the trained battery defect model comprise a plurality of types, and the specific types are self-defined according to requirements, for example, the trained battery defect identification model comprises a cold joint defect image, a microcrack defect image, a finger breakage defect image, a non-defective image and the like.
In one embodiment, the trained battery defect recognition model includes an input layer, a competition layer, and an output layer, wherein an output of the input layer is an input to the competition layer and an output of the competition layer is an input to the output layer.
Specifically, the image to be recognized is input according to the input rule of the input layer to obtain output data of the input layer, the output data is competed through the competition winning rule of the competition layer to obtain the cell panel characteristics extracted by the neurons winning the competition layer, and the recognition result of the image to be recognized is output on the output layer according to the cell panel characteristics extracted by the neurons winning.
The input layer is a network layer used for inputting the image to be detected to the image to be recognized according to the input rule. The input rule refers to a predefined rule, and the input rule can be self-defined, if the input rule is defined according to the application scene of the model. And inputting the image to be recognized according to the input rule to obtain output data. And inputting the output data of the input layer into a competition layer, wherein the competition layer is used for carrying out feature extraction and feature screening on the output data of the input layer and comprises a plurality of neurons. A group of neurons in the competition layer is connected with one neuron of the output layer, the neuron of each output layer corresponds to the neurons of the group of competition layers, and the neurons of the competition layers are connected with only one neuron of the output layers. And comparing the similarity of the weight vectors corresponding to all the neurons in the competition layer with the output data, and judging the most similar weight vector as a competition winning neuron, wherein the winning neuron allows the output to be 1, and the output of other neurons is 0. The output of the output neuron connected to the group in which the winning neuron is located is 1, and the output of the other output neurons is 0, thereby giving the recognition result of the image to be recognized.
The battery defect identification method, the battery defect identification device, the computer equipment and the storage medium comprise the following steps: inputting an image to be recognized containing the battery panel characteristics into a trained battery defect recognition model, wherein the image to be recognized is an electroluminescence image, extracting the battery panel characteristics of the image to be recognized through the trained battery defect recognition model, and classifying the image to be recognized according to the battery panel characteristics to obtain the recognition result of the image to be recognized. The battery defect images are identified through the trained identification model, so that the identification speed and the identification accuracy are improved.
In one embodiment, the step of generating a trained battery defect recognition model comprises:
step S301, a training set containing a plurality of training images and a test set containing a plurality of test images are obtained.
The training image and the test image carry corresponding defect labels, the training image is an image related to a training model, the test image is an image used for testing the performance of the model, the training set is a set formed by a plurality of training images, and the test set is a set formed by the test images. The defect label is label data used for describing defects of the battery in the image, and the label data can be customized according to requirements and comprises a cold joint defect image, a crack defect image, a finger breakage defect image, a non-defective image and the like. The type number of the label data is the output neuron number of the trained battery defect model.
Step S302, an initial battery defect identification model is constructed.
Step S303, inputting each training image and the corresponding defect label into an initial battery defect identification model, taking the defect label corresponding to each training image as an expected output result corresponding to each training image, and training each training image to obtain a candidate battery defect identification model.
Step S304, inputting each test image and the corresponding defect label into a candidate battery defect identification model, and classifying and identifying each test image through the candidate battery defect identification model.
Step S305, judging whether the candidate battery defect identification model meets the preset convergence condition according to the identification result of each test image and the corresponding defect label.
And step S306, obtaining a trained battery defect identification model when the condition is met.
Specifically, the model is selected as an initial battery defect identification model according to business requirements in a self-defined mode, wherein model parameters of the initial battery defect identification model can be set according to a training set. Inputting the training image and a corresponding defect label into an initial battery defect identification model, wherein the defect label is data which is calibrated in advance and used for describing the real state of the battery panel in the image, and for example, a cold joint defect image refers to the problem that the shot battery panel has cold joint. The expected output result refers to an expected output result obtained after the training image is trained through the initial battery defect identification model, if the training image is a cold joint defect image, the result obtained through the expected identification of the initial battery defect identification model is a cold joint defect image, and when the identification result is different from the actual result, the model parameters of the initial battery defect identification model are updated until the training condition is met, so that the candidate battery defect identification model is obtained. The candidate battery defect identification model is obtained by training a training image, and whether the preset convergence condition requirement is met or not needs to be tested.
And inputting the test image into the candidate battery defect identification model to obtain a corresponding identification result, and judging whether the identification result of the test image is matched with the corresponding defect label. The preset convergence condition is a preset condition for judging whether the model converges, and the condition can be self-defined, for example, the preset convergence condition is set as a preset correct recognition rate. And when the preset convergence condition is the preset correct recognition rate, counting the number of the images matched with the corresponding defect labels for the recognition of the test images, calculating to obtain the real correct recognition rate according to the counted number and the number of the test images of the test set, and when the real correct recognition rate is greater than or equal to the preset correct recognition rate, enabling the candidate battery defect recognition model to meet the preset convergence condition, so as to obtain the trained battery defect recognition model.
In one embodiment, when the battery defect identification model is not satisfied, the model parameters of the candidate battery defect identification model are updated, and the candidate battery defect identification model is trained again until the candidate battery defect identification model satisfies a preset convergence condition, so that the trained battery defect identification model is obtained.
Specifically, when the candidate battery defect identification model does not meet the preset convergence condition, updating model parameters of the candidate battery defect identification model, training the training image of the training set by using the candidate battery defect identification model with the updated model parameters to obtain the updated candidate battery defect identification model, judging the updated candidate battery defect identification model to meet the preset convergence condition again until the updated candidate battery defect identification model meets the preset convergence condition, and obtaining the trained battery defect identification model. The model is trained through the training set, and the model is tested through the testing set, so that the model can keep balance on the accuracy of feature extraction and the accuracy of model identification, and the problem of poor model applicability caused by over-fitting and under-fitting of the model is avoided.
In one embodiment, the battery defect identification method further includes:
step S401, a sampling training set is obtained, wherein the sampling training set comprises a plurality of sampling images.
And S402, performing dimensionality reduction on the sampling training set to obtain a training set.
Specifically, the sampling training set refers to an image set composed of sampling images obtained by sampling original images captured by the capturing device. The dimensionality reduction is used for screening data of the sampling training set, a data screening algorithm can be self-defined, and adopted images can be screened according to one or more of the correlation among the data, the quality of the images and the like during screening. For example, a PCA (principal component analysis) algorithm may be selected when the data is screened by using the correlation of the data.
In one embodiment, before performing dimension reduction on the sampled training set to obtain the training set, the method further includes: and judging whether the correlation between the sampling images in the sampling training set meets a preset correlation threshold, if so, reducing the dimension of the sampling training set to obtain a training set, otherwise, not reducing the dimension of the sampling training set. Wherein the correlation between the sample images may be determined according to a corresponding calculation algorithm, e.g. the sample covariance may be used to determine the correlation between the sample images. The dimension reduction is carried out on the sampling training set, the training of the images with strong correlation can be avoided, the training speed is improved, the images with high correlation indicate that the image repetition degree is high, and the images with high repetition degree are not completely reserved.
In one embodiment, step S402 includes:
step S4021, calculating a sample mean value of the sampling training set.
Step S4022, calculating a sample matrix of the sampling training set according to each sampling image and the sample mean value.
Step S4023, a covariance value of the sample matrix is calculated.
Step S4024, when the covariance value is greater than the preset threshold, calculating the eigenvalue of the sample matrix and the corresponding eigenvector.
Step S4025, selecting a target characteristic value from the characteristic values according to a preset rule, taking the characteristic vector corresponding to the target characteristic value as a training image, and forming a training set by the training image.
Specifically, the sample mean is a mean of images of each sample in the sampling training set, and the calculation method of the sample mean may be customized, for example, corresponding weights are set for different samples, and the samples are weighted and summed according to the weights corresponding to each sample to obtain the corresponding sample mean. The sample matrix is obtained by weighting the mean value of the samples by using each sampling image of the sampling training set. E.g., each sample in the sample matrix is equal to each sampled image minus the sample mean. The covariance value is a value for measuring the correlation of samples in the sample matrix, and the covariance value of the sample matrix is calculated. When the covariance value is greater than the preset threshold, the existence of the correlation of the samples of the sample matrix is represented, and the correlation meets the preset threshold, so that the dimension reduction of the sample matrix can be performed. Calculating the characteristic values and corresponding characteristic vectors of the sample matrix, selecting target characteristic values from the characteristic values according to a preset rule, taking the characteristic vectors corresponding to the target characteristic values as training images, and forming a training set by the training images. The preset rule is a preset rule for screening the target characteristic value, and if the selected characteristic value is larger than the preset characteristic value and serves as the target characteristic value, a preset number of characteristic values can be selected and serve as the target characteristic value, and the specific rule can be customized. And reducing the dimension of the sample training set according to the correlation of the samples, and removing images with high repetition degree, thereby improving the training speed of the model.
In a specific embodiment, the battery defect identification method includes:
and collecting the EL image, and respectively collecting a false solder defect image, a microcrack defect image, a finger breakage defect image and a non-defective image. When the input image of the LVQ is expressed in the form of a vector, the dimension of the vector is too large to be beneficial for calculation, so the PCA algorithm is adopted to perform dimension reduction. Wherein the EL image may be a solar panel image.
PCA projects sample data into a new space based on a covariance matrix, and therefore only the spatial coordinates corresponding to the eigenvalues of one linearly independent group where the sample data is largest are needed. Suppose a data matrix Xn×pComposed of sample images, n is the number of samples, and p is the size of the sample image. If X isn×pEach line of (b) represents a sample image, then Xn×pThe PCA dimension reduction matrix solving steps are as follows:
finding Xn×pAverage value of (a). Let the sample images be X respectively1、X2、...、XnAnd calculating the average value of the sample image, wherein the specific calculation formula is shown as (1):
from X1、X2、...、XnMinus
To obtain phi1、φ2、...、φnForm matrix M ═ phi1φ2...φnCalculate the covariance cov (M) MM of the matrix MT，MTIs the transpose of M. Solving the eigenvalue lambda and the eigenvector U, wherein M-lambda I is 0, (M-lambda I)iI)UiI is a unit vector, the eigenvalues lambda are sorted from large to small, the eigenvectors corresponding to the eigenvalues arranged at the top k are reserved to form Y,
k≤n，Y＝(y1...yk)。
and the matrix Y after the PCA dimension reduction is an input matrix of the LVQ neural network. As shown in fig. 3, the LVQ neural network includes an input layer, a competition layer, and an output layer. The input layer comprises k neurons for receiving input vectors, the competition layer comprises m neurons, the m neurons are divided into a plurality of groups and are arranged in a one-dimensional linear array, each neuron in the output layer is only connected with one group of neurons in the competition layer, namely the neuron in each output layer is connected with the neuron in at least one competition layer, each neuron in the competition layer can only be connected with the neuron in one output layer, the connection weight is fixed to be 1, the 4 types of outputs are respectively a virtual solder defect image, a micro-crack defect image, a finger breakage defect image and a non-defective image. With 80 neurons in the competition layer. The output layer result is 4 types, and the number of output layer nodes is set to 4. Selecting 1200 EL pictures of the solar cell panel, wherein 300 samples of a cold solder defect, a microcrack defect, a finger breakage defect and a defect-free sample are respectively used as training samples, then carrying out LVQ neural network training, randomly generating a matrix with 4 rows and 80 columns by a computer according to the weight of W, and correcting the W according to the LVQ neural network learning algorithm until the classification recognition accuracy of the LVQ neural network on the training samples meets the preset accuracy, thus obtaining a trained cell defect recognition model. And inputting the image to be recognized containing the battery panel characteristics into the trained battery defect recognition model, and classifying and recognizing the image to be recognized through the trained battery defect recognition model to obtain a corresponding recognition result, wherein if the image to be recognized containing the battery panel characteristics and being defect-free is input into the trained battery defect recognition model, the corresponding recognition result is defect-free. The battery defect images are identified through the trained identification model, so that the identification speed and the identification accuracy are improved. The trained model can rapidly and accurately extract the battery plate characteristics in the image and rapidly identify the type of the battery plate according to the extracted characteristics.
In one embodiment, as shown in fig. 4, there is provided a battery defect identifying apparatus 200 including:
the feature extraction module 201 is configured to input an image to be identified, which includes a battery panel feature, into the trained battery defect identification model, and extract the battery panel feature of the image to be identified through the trained battery defect identification model, where the image to be identified is an electroluminescent image.
And the identification module 202 is configured to classify the image to be identified according to the characteristics of the battery panel to obtain an identification result of the image to be identified.
In one embodiment, the battery defect recognition apparatus 200 further includes:
the data acquisition module is used for acquiring a training set containing a plurality of training images and a test set containing a plurality of test images, and the training images and the test images carry corresponding defect labels.
And the model construction module is used for constructing an initial battery defect identification model.
And the candidate model determining module is used for inputting each training image and the corresponding defect label into the initial battery defect identification model, taking the defect label corresponding to each training image as an expected output result corresponding to each training image, and training each training image to obtain a candidate battery defect identification model.
The test module is used for inputting each test image and the corresponding defect label into a candidate battery defect identification model and classifying and identifying each test image through the candidate battery defect identification model;
and the model determining module is used for judging whether the candidate battery defect identification model meets a preset convergence condition or not according to the identification result of each test image and the corresponding defect label, and obtaining a trained battery defect identification model when the candidate battery defect identification model meets the preset convergence condition.
In one embodiment, the battery defect recognition apparatus 200 further includes:
and the model updating module is used for updating model parameters of the candidate battery defect identification model when the model parameters are not met, and retraining the candidate battery defect identification model until the candidate battery defect identification model meets the preset convergence condition to obtain the trained battery defect identification model.
In one embodiment, the battery defect recognition apparatus 200 further includes:
and the sampling data acquisition module is used for acquiring a sampling training set, and the sampling training set comprises a plurality of sampling images.
And the data dimension reduction module is used for reducing the dimension of the sampling training set to obtain a training set.
In one embodiment, the data dimension reduction module includes:
the mean value calculating unit is used for calculating the sample mean value of the sampling training set;
and the covariance calculation unit is used for calculating a sample matrix of the sampling training set according to each sampling image and the sample mean value.
And the dimensionality reduction unit is used for calculating a covariance value of the sample matrix, calculating a characteristic value and a corresponding characteristic vector of the sample matrix when the covariance value is greater than a preset threshold value, selecting a target characteristic value from the characteristic value according to a preset rule, taking the characteristic vector corresponding to the target characteristic value as a training image, and forming a training set by the training image.
In one embodiment, the feature extraction module 201 is further configured to input the image to be recognized according to an input rule of the input layer to obtain output data of the input layer, and compete the output data according to a competition winning rule of the competition layer to obtain the panel features extracted by a winning neuron of the competition layer.
The recognition module 202 is further configured to output, on an output layer, a recognition result of an image to be recognized according to the panel features extracted by the winning neurons, where the trained battery defect recognition model includes an input layer, a competition layer, and an output layer, an output of the input layer is used as an input of the competition layer, and an output of the competition layer is used as an input of the output layer.
FIG. 5 is a diagram illustrating an internal structure of a computer device in one embodiment. The computer device may specifically be the terminal 110 (or the server 120) in fig. 1. As shown in fig. 5, the computer apparatus includes a processor, a memory, a network interface, an input device, and a display screen connected through a system bus. Wherein the memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program that, when executed by the processor, causes the processor to implement the battery defect identification method. The internal memory may also have a computer program stored therein, which when executed by the processor, causes the processor to perform the battery defect identification method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 5 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, the battery defect identifying apparatus provided in the present application may be implemented in the form of a computer program, and the computer program may be run on a computer device as shown in fig. 5. The memory of the computer device may store various program modules constituting the battery defect identifying apparatus, such as the feature extraction module 201 and the identification module 202 shown in fig. 4. The computer program constituted by the respective program modules causes the processor to execute the steps in the battery defect identifying method of the respective embodiments of the present application described in the present specification.
For example, the computer device shown in fig. 5 may extract the panel features of the image to be recognized through the trained battery defect recognition model by inputting the image to be recognized, which includes the panel executed by the feature extraction module 201, into the trained battery defect recognition model in the battery defect recognition apparatus shown in fig. 4. The computer device can classify the image to be recognized according to the characteristics of the battery plate through the recognition module 202 to obtain the recognition result of the image to be recognized.
In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program: inputting an image to be recognized containing the characteristics of the battery panel into a trained battery defect recognition model, wherein the image to be recognized is an electroluminescent image, extracting the characteristics of the battery panel of the image to be recognized through the trained battery defect recognition model, and classifying the image to be recognized according to the characteristics of the battery panel to obtain a recognition result of the image to be recognized.
In one embodiment, the step of generating a trained battery defect recognition model comprises: acquiring a training set comprising a plurality of training images, a test set comprising a plurality of test images, training images of the training set and corresponding defect labels carried by the test images, constructing an initial battery defect identification model, inputting each training image and corresponding defect label into the initial battery defect identification model, training each training image by taking the defect label corresponding to each training image as an expected output result corresponding to each training image to obtain a candidate battery defect identification model, inputting each test image and corresponding defect label into the candidate battery defect identification model, classifying and identifying each test image through the candidate battery defect identification model, and judging whether the candidate battery defect identification model meets a preset convergence condition or not according to the identification result of each test image and the corresponding defect label, and obtaining a trained battery defect identification model when the candidate battery defect identification model meets the preset convergence condition.
The processor in one embodiment when executing the computer program further performs the steps of: and when the battery defect identification model is not met, updating model parameters of the candidate battery defect identification model, and training the candidate battery defect identification model again until the candidate battery defect identification model meets a preset convergence condition to obtain the trained battery defect identification model.
In one embodiment the following steps are also implemented when the computer program is executed by a processor before the training set comprising a plurality of training images and the test set comprising a plurality of test images are acquired: and acquiring a sampling training set, wherein the sampling training set comprises a plurality of sampling images, and performing dimensionality reduction on the sampling training set to obtain a training set.
In one embodiment, the performing the dimensionality reduction on the sampling training set to obtain the training set comprises: calculating a sample mean value of a sampling training set, calculating a sample matrix of the sampling training set according to each sampling image and the sample mean value, calculating a covariance value of the sample matrix, calculating a characteristic value and a corresponding characteristic vector of the sample matrix when the covariance value is greater than a preset threshold value, selecting a target characteristic value from the characteristic values according to a preset rule, and forming the training set by using the characteristic vector corresponding to the target characteristic value as a training image.
In one embodiment, the trained battery defect recognition model input layer, the competition layer and the output layer, wherein the output of the input layer is used as the output of the input competition layer of the competition layer, and the image to be recognized containing the panel features is input into the trained battery defect recognition model, and the method comprises the following steps: the method comprises the following steps of inputting an image to be recognized according to an input rule of an input layer to obtain output data of the input layer, and extracting the battery panel characteristics of the image to be recognized through a trained battery defect recognition model, wherein the battery panel characteristics comprise: competition winning rules through the competition layer compete for output data, obtain the panel characteristic that the winning neuron of competition layer drawed, treat the discernment image according to the panel characteristic and classify, obtain the recognition result of treating the discernment image, include: and outputting the recognition result of the image to be recognized on an output layer according to the battery panel characteristics extracted by the winning neurons.
In one embodiment, the identification results include a cold solder defect, a microcrack defect, a finger break defect, and a defect free.
In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of: inputting an image to be recognized containing the characteristics of the battery panel into a trained battery defect recognition model, wherein the image to be recognized is an electroluminescent image, extracting the characteristics of the battery panel of the image to be recognized through the trained battery defect recognition model, and classifying the image to be recognized according to the characteristics of the battery panel to obtain a recognition result of the image to be recognized.
In one embodiment, the step of generating a trained battery defect recognition model comprises: acquiring a training set comprising a plurality of training images, a test set comprising a plurality of test images, training images of the training set and corresponding defect labels carried by the test images, constructing an initial battery defect identification model, inputting each training image and corresponding defect label into the initial battery defect identification model, training each training image by taking the defect label corresponding to each training image as an expected output result corresponding to each training image to obtain a candidate battery defect identification model, inputting each test image and corresponding defect label into the candidate battery defect identification model, classifying and identifying each test image through the candidate battery defect identification model, and judging whether the candidate battery defect identification model meets a preset convergence condition or not according to the identification result of each test image and the corresponding defect label, and obtaining a trained battery defect identification model when the candidate battery defect identification model meets the preset convergence condition.
The processor in one embodiment when executing the computer program further performs the steps of: and when the battery defect identification model is not met, updating model parameters of the candidate battery defect identification model, and training the candidate battery defect identification model again until the candidate battery defect identification model meets a preset convergence condition to obtain the trained battery defect identification model.
In one embodiment the following steps are also implemented when the computer program is executed by a processor before the training set comprising a plurality of training images and the test set comprising a plurality of test images are acquired: and acquiring a sampling training set, wherein the sampling training set comprises a plurality of sampling images, and performing dimensionality reduction on the sampling training set to obtain a training set.
In one embodiment, the performing the dimensionality reduction on the sampling training set to obtain the training set comprises: calculating a sample mean value of a sampling training set, calculating a sample matrix of the sampling training set according to each sampling image and the sample mean value, calculating a covariance value of the sample matrix, calculating a characteristic value and a corresponding characteristic vector of the sample matrix when the covariance value is greater than a preset threshold value, selecting a target characteristic value from the characteristic values according to a preset rule, and forming the training set by using the characteristic vector corresponding to the target characteristic value as a training image.
In one embodiment, the trained battery defect recognition model input layer, the competition layer and the output layer, wherein the output of the input layer is used as the output of the input competition layer of the competition layer, and the image to be recognized containing the panel features is input into the trained battery defect recognition model, and the method comprises the following steps: the method comprises the following steps of inputting an image to be recognized according to an input rule of an input layer to obtain output data of the input layer, and extracting the battery panel characteristics of the image to be recognized through a trained battery defect recognition model, wherein the battery panel characteristics comprise: competition winning rules through the competition layer compete for output data, obtain the panel characteristic that the winning neuron of competition layer drawed, treat the discernment image according to the panel characteristic and classify, obtain the recognition result of treating the discernment image, include: and outputting the recognition result of the image to be recognized on an output layer according to the battery panel characteristics extracted by the winning neurons. In one embodiment, the identification results include a cold solder defect, a microcrack defect, a finger break defect, and a defect free.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a non-volatile computer-readable storage medium, and can include the processes of the embodiments of the methods described above when the program is executed. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile 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 DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
It is noted that, in this document, relational terms such as "first" and "second," and the like, may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The foregoing are merely exemplary embodiments of the present invention, which enable those skilled in the art to understand or practice the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A battery defect identification method, the method comprising:
inputting an image to be recognized containing the characteristics of a battery panel into a trained battery defect recognition model, wherein the image to be recognized is an electroluminescent image;
extracting the battery panel characteristics of the image to be recognized through the trained battery defect recognition model;
and classifying the images to be recognized according to the characteristics of the battery plates to obtain the recognition result of the images to be recognized.
2. The method of claim 1, wherein the step of generating the trained battery defect recognition model comprises:
acquiring a training set containing a plurality of training images and a test set containing a plurality of test images, wherein the training images and the test images carry corresponding defect labels;
constructing an initial battery defect identification model;
inputting each training image and the corresponding defect label into the initial battery defect identification model, taking the defect label corresponding to each training image as an expected output result corresponding to each training image, and training each training image to obtain a candidate battery defect identification model;
inputting each test image and the corresponding defect label into the candidate battery defect identification model, and classifying and identifying each test image through the candidate battery defect identification model;
judging whether the candidate battery defect identification model meets a preset convergence condition or not according to the identification result of each test image and the corresponding defect label;
and when the condition is met, obtaining the trained battery defect identification model.
3. The method of claim 2, further comprising:
and when the battery defect identification model is not met, updating model parameters of the candidate battery defect identification model, and re-training the candidate battery defect identification model until the candidate battery defect identification model meets the preset convergence condition to obtain the trained battery defect identification model.
4. The method of claim 2, wherein prior to obtaining the training set comprising the plurality of training images and the test set comprising the plurality of test images, further comprising:
acquiring a sampling training set, wherein the sampling training set comprises a plurality of sampling images;
and performing dimensionality reduction on the sampling training set to obtain the training set.
5. The method of claim 4, wherein the dimensionality reduction of the sampled training set to obtain the training set comprises:
calculating a sample mean value of the sampling training set;
calculating a sample matrix of the sampling training set according to each sampling image and the sample mean value;
calculating covariance values of the sample matrix;
when the covariance value is larger than a preset threshold value, calculating an eigenvalue of the sample matrix and a corresponding eigenvector;
and selecting a target characteristic value from the characteristic values according to a preset rule, taking a characteristic vector corresponding to the target characteristic value as the training image, and forming the training set by the training image.
6. The method according to any one of claims 1 to 5, wherein the trained battery defect recognition model comprises an input layer, a competition layer and an output layer, wherein the output of the input layer is used as the input of the competition layer, the output of the competition layer is used as the input of the output layer, and inputting the image to be recognized containing the panel features into the trained battery defect recognition model comprises the following steps:
inputting the image to be recognized according to the input rule of the input layer to obtain output data of the input layer;
the extracting the battery panel characteristics of the image to be recognized through the trained battery defect recognition model comprises the following steps:
and competing the output data through the competition winning rule of the competition layer to obtain the cell panel characteristics extracted by the winning neurons of the competition layer.
The classifying the image to be recognized according to the battery plate characteristics to obtain the recognition result of the image to be recognized comprises the following steps:
and outputting the recognition result of the image to be recognized on the output layer according to the battery panel characteristics extracted by the winning neuron.
7. The method of claim 1, wherein the identification comprises a cold joint defect, a microcrack defect, a finger break defect, and a defect free.
8. A battery defect identifying apparatus, the apparatus comprising:
the characteristic extraction module is used for inputting an image to be recognized containing battery panel characteristics into a trained battery defect recognition model, and extracting the battery panel characteristics of the image to be recognized through the trained battery defect recognition model, wherein the image to be recognized is an electroluminescent image;
and the identification module is used for classifying the image to be identified according to the characteristics of the battery panel to obtain an identification result of the image to be identified.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method of any of claims 1 to 7 are implemented when the computer program is executed by the processor.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811544225.8A CN111340748A (en) | 2018-12-17 | 2018-12-17 | Battery defect identification method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811544225.8A CN111340748A (en) | 2018-12-17 | 2018-12-17 | Battery defect identification method and device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111340748A true CN111340748A (en) | 2020-06-26 |
Family
ID=71181288
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811544225.8A Pending CN111340748A (en) | 2018-12-17 | 2018-12-17 | Battery defect identification method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111340748A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114255540A (en) * | 2022-01-25 | 2022-03-29 | 中国农业银行股份有限公司 | Method, device, equipment and storage medium for identifying stained paper money |
| CN114897901A (en) * | 2022-07-13 | 2022-08-12 | 东声(苏州)智能科技有限公司 | Battery quality detection method and device based on sample expansion and electronic equipment |
| CN115239644A (en) * | 2022-07-05 | 2022-10-25 | 港珠澳大桥管理局 | Concrete defect identification method and device, computer equipment and storage medium |
-
2018
- 2018-12-17 CN CN201811544225.8A patent/CN111340748A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114255540A (en) * | 2022-01-25 | 2022-03-29 | 中国农业银行股份有限公司 | Method, device, equipment and storage medium for identifying stained paper money |
| CN115239644A (en) * | 2022-07-05 | 2022-10-25 | 港珠澳大桥管理局 | Concrete defect identification method and device, computer equipment and storage medium |
| CN115239644B (en) * | 2022-07-05 | 2024-04-05 | 港珠澳大桥管理局 | Concrete defect identification method, device, computer equipment and storage medium |
| CN114897901A (en) * | 2022-07-13 | 2022-08-12 | 东声(苏州)智能科技有限公司 | Battery quality detection method and device based on sample expansion and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113486981B (en) | RGB image classification method based on multi-scale feature attention fusion network | |
| CN109583489A (en) | Defect classifying identification method, device, computer equipment and storage medium | |
| CN109215034B (en) | Weak supervision image semantic segmentation method based on spatial pyramid covering pooling | |
| CN111126482B (en) | Remote sensing image automatic classification method based on multi-classifier cascade model | |
| CN111783532B (en) | Cross-age face recognition method based on online learning | |
| CN111340748A (en) | Battery defect identification method and device, computer equipment and storage medium | |
| CN109993221B (en) | Image classification method and device | |
| CN111832581B (en) | Lung feature recognition method and device, computer equipment and storage medium | |
| CN112926654A (en) | Pre-labeling model training and certificate pre-labeling method, device, equipment and medium | |
| CN106599864A (en) | Deep face recognition method based on extreme value theory | |
| CN116597384B (en) | Space target identification method and device based on small sample training and computer equipment | |
| CN112364974B (en) | YOLOv3 algorithm based on activation function improvement | |
| CN113095156B (en) | Double-current network signature identification method and device based on inverse gray scale mode | |
| CN111144398A (en) | Target detection method, target detection device, computer equipment and storage medium | |
| CN117355038B (en) | X-shaped hole processing method and system for circuit board soft board | |
| CN110135505A (en) | Image classification method, device, computer equipment and computer readable storage medium | |
| CN111340213B (en) | Neural network training method, electronic device, and storage medium | |
| CN112668462A (en) | Vehicle loss detection model training method, vehicle loss detection device, vehicle loss detection equipment and vehicle loss detection medium | |
| CN109101984B (en) | Image identification method and device based on convolutional neural network | |
| CN111915595A (en) | Image quality evaluation method, and training method and device of image quality evaluation model | |
| CN117152625A (en) | Remote sensing small target identification method, system, equipment and medium based on CoordConv and Yolov5 | |
| CN110599665B (en) | Paper pattern recognition method and device, computer equipment and storage medium | |
| CN114387524B (en) | Image identification method and system for small sample learning based on multilevel second-order representation | |
| CN110675382A (en) | Aluminum electrolysis superheat degree identification method based on CNN-LapseLM | |
| CN107679528A (en) | A kind of pedestrian detection method based on AdaBoost SVM Ensemble Learning Algorithms |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20210730 Address after: 518000 Room 201, building A, No. 1, Qian Wan Road, Qianhai Shenzhen Hong Kong cooperation zone, Shenzhen, Guangdong (Shenzhen Qianhai business secretary Co., Ltd.) Applicant after: Hongyi Technology Co.,Ltd. Address before: Room 107, building 2, Olympic Village street, Chaoyang District, Beijing Applicant before: HANERGY MOBILE ENERGY HOLDING GROUP Co.,Ltd. |
|
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |