CN111680717A - Product classification method and system on intelligent manufacturing production line based on deep learning - Google Patents
Product classification method and system on intelligent manufacturing production line based on deep learning Download PDFInfo
- Publication number
- CN111680717A CN111680717A CN202010398375.3A CN202010398375A CN111680717A CN 111680717 A CN111680717 A CN 111680717A CN 202010398375 A CN202010398375 A CN 202010398375A CN 111680717 A CN111680717 A CN 111680717A
- Authority
- CN
- China
- Prior art keywords
- neural network
- network model
- deep neural
- training
- intelligent manufacturing
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/045—Combinations of networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a product classification method and a product classification system on an intelligent manufacturing production line based on deep learning, wherein the method comprises the following steps: acquiring monitoring data of a sample to be trained on a historical monitoring database on an intelligent manufacturing production line; inputting monitoring data of a sample to be trained into a preset deep neural network model for training and learning; model updating is carried out on the deep neural network model based on the network node parameters of the deep neural network model after training and learning; inputting current product monitoring data acquired on an intelligent manufacturing production line into the updated deep neural network model for product classification prediction, and outputting a product classification prediction result; and sending the product classification prediction result to a classification controller at the tail end of the intelligent manufacturing production line, and performing control classification by the classification controller based on the product classification prediction result. In the embodiment of the invention, the products on the intelligent manufacturing production line can be accurately and automatically classified, and the labor cost is reduced.
Description
Technical Field
The invention relates to the technical field of intelligent manufacturing, in particular to a product classification method and system on an intelligent manufacturing production line based on deep learning.
Background
Industrial production automation is a process in which various automatic control, automatic detection and automatic adjustment devices are widely used in industrial production to automatically measure, check, calculate, control, monitor, etc. a production process instead of a human being to operate a machine. Automation is a higher stage of production mechanization and is one of the fundamental directions in the modernization of industrial technology. The method is divided into the following development stages: the method is semi-automatic. Namely, part adopts manual operation, and the other part adopts automatic control to carry out production; and secondly, full-scale automation, also called an automatic production line. The whole process is automated; and thirdly, comprehensive automation.
The existing industrial automation realizes the automatic production process of products, but the existing automatic production line can simultaneously output a plurality of lines, for example, products of different models are output through one output port, or products of different models are output through one output port, and the existing automatic product classification method or device can not realize accurate and rapid classification in complex multi-input and multi-output ports and also needs manual auxiliary classification.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, and provides a product classification method and system on an intelligent manufacturing production line based on deep learning, which can realize accurate automatic classification of products on the intelligent manufacturing production line, do not need manual complexity, and reduce labor cost.
In order to solve the technical problem, an embodiment of the present invention provides a product classification method on an intelligent manufacturing line based on deep learning, where the method includes:
acquiring monitoring data of a sample to be trained on a historical monitoring database on an intelligent manufacturing production line, wherein the monitoring data of the sample to be trained is monitoring data of the intelligent manufacturing production line within a preset period of time;
inputting the monitoring data of the sample to be trained into a preset deep neural network model for training and learning to obtain network node parameters of the deep neural network model after training and learning;
carrying out model updating on the deep neural network model based on the network node parameters of the deep neural network model after training and learning;
inputting the current product monitoring data collected on the intelligent manufacturing production line into the updated deep neural network model for product classification prediction, and outputting a product classification prediction result;
and sending the product classification prediction result to a classification controller at the tail end of the intelligent manufacturing production line, and performing control classification by the classification controller based on the product classification prediction result.
Optionally, the obtaining of the monitoring data of the sample to be trained on the historical monitoring database on the intelligent manufacturing production line includes:
extracting and processing samples to be trained in a historical monitoring database on the intelligent manufacturing production line according to a time sequence and a preset time length to obtain monitoring data of the samples to be trained;
the monitoring data of the sample to be trained is data which is not subjected to extracted training labels in the historical monitoring database;
and after the monitoring data of the sample to be trained is obtained, carrying out extracted training annotation in the historical monitoring database.
Optionally, the building of the history monitoring database includes:
acquiring data on the basis of set binocular camera equipment on an intelligent manufacturing production line to obtain the product acquisition data;
and performing classification labeling on the product data based on the intelligent manufacturing production line control classification, and storing the product data in a historical monitoring database according to the collected time sequence after the classification labeling is completed.
Optionally, the training loss function of the deep neural network model is as follows:
introducing regularization among parameter vectors into a loss function, and carrying out regularization processing on parameters of all layers in the deep neural network model;
after the regularization process, a new loss function is constructed as a training loss function of the deep neural network.
Optionally, the training loss function is as follows:
L＝Lcross-entropy+λR；
in a corresponding manner, the first and second optical fibers are,
wherein R is a regularization term of all layer parameter vectors, NlIs the number of l-th layer parameter vectors, Nl1,2,3, …, wherein M is the number of layers of the deep neural network model, M is 1,2,3, …, l is 1,2,3, …, M, W is the parameter of the neural network model to be trained, W is the parameter of the deep neural network model, M is the parameter ofliIs the ith parameter vector of the l layer in the model parameters of the deep neural network, WljIs the jth parameter vector of the L layer in the model parameters of the deep neural network, Lcross-entropyIs a cross entropy loss function, lambda is a hyperparameter, and L is a training loss function.
Optionally, the inputting the monitoring data of the sample to be trained into a preset deep neural network model for training and learning to obtain network node parameters of the deep neural network model after training and learning includes:
resetting coefficients of all layer parameter vectors of the preset deep neural network model by using a back propagation algorithm before training and learning of the preset deep neural network model;
and inputting the monitoring data of the sample to be trained into the deep neural network model after the coefficients of all the layer parameter vectors are reset for training and learning, and obtaining the network node parameters of the deep neural network model after training and learning.
Optionally, in the process of resetting the coefficients of all layer parameter vectors of the preset deep neural network model by using the back propagation algorithm, a formula for resetting the coefficients of all layer parameter vectors of the preset deep neural network model is as follows:
wherein the content of the first and second substances,
for updated coefficients of the i-th and j-th parameter vectors in the trained neural network model, WijThe coefficients of the ith parameter vector and the jth parameter vector in the trained neural network model are represented, lr is the learning rate, weight _ decay is the weight attenuation coefficient, momentum is the impulse, and L is the updated loss function.
Optionally, the classifying controller performs control classification based on the product classification prediction result, including:
after receiving the product classification prediction result, the classification controller calls the video on the intelligent production line to monitor the current product for real-time positioning and tracking processing based on the current product monitoring data corresponding to the product classification prediction result, so as to obtain a real-time positioning and tracking result;
and when the current product is transmitted to a preset position on the intelligent manufacturing production line, the classification controller controls the classification robot to perform control classification according to the product classification prediction result.
In addition, the embodiment of the invention also provides a product classification system on the intelligent manufacturing production line based on deep learning, which comprises the following components:
a sample data obtaining module: the system comprises a historical monitoring database, a training database and a training database, wherein the historical monitoring database is used for acquiring to-be-trained sample monitoring data on an intelligent manufacturing production line, and the to-be-trained sample monitoring data is monitoring data of the intelligent manufacturing production line within a preset period of time;
training a learning module: the system comprises a deep neural network model, a network node parameter acquisition module, a data acquisition module and a data processing module, wherein the deep neural network model is used for acquiring a deep neural network model to be trained;
an update module: the deep neural network model is updated based on the network node parameters of the deep neural network model after training and learning;
a product classification prediction module: the system comprises a deep neural network model, a product classification prediction module and a product classification prediction module, wherein the deep neural network model is used for inputting current product monitoring data acquired on the intelligent manufacturing production line into the updated deep neural network model for product classification prediction and outputting a product classification prediction result;
a control classification module: the classification controller is used for sending the product classification prediction result to the classification controller at the tail end of the intelligent manufacturing production line, and the classification controller carries out control classification based on the product classification prediction result.
In the embodiment of the invention, the products on the intelligent manufacturing production line can be accurately and automatically classified by utilizing the deep neural network model to realize the classified prediction of the products on the intelligent manufacturing production line and then sending the classified prediction result of the products to the classification controller, so that the manual complexity is not required, and the labor cost is reduced.
Drawings
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, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for product classification in an intelligent manufacturing line based on deep learning according to an embodiment of the present invention;
fig. 2 is a schematic structural composition diagram of a product classification system on an intelligent manufacturing line based on deep learning in an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Examples
Referring to fig. 1, fig. 1 is a flowchart illustrating a product classification method in an intelligent manufacturing line based on deep learning according to an embodiment of the present invention.
As shown in fig. 1, a method for classifying products in an intelligent manufacturing line based on deep learning, the method comprising:
s11: acquiring monitoring data of a sample to be trained on a historical monitoring database on an intelligent manufacturing production line, wherein the monitoring data of the sample to be trained is monitoring data of the intelligent manufacturing production line within a preset period of time;
in the specific implementation process of the invention, the obtaining of the monitoring data of the sample to be trained on the historical monitoring database on the intelligent manufacturing production line comprises the following steps: extracting and processing samples to be trained in a historical monitoring database on the intelligent manufacturing production line according to a time sequence and a preset time length to obtain monitoring data of the samples to be trained; the monitoring data of the sample to be trained is data which is not subjected to extracted training labels in the historical monitoring database; and after the monitoring data of the sample to be trained is obtained, carrying out extracted training annotation in the historical monitoring database.
Further, the construction of the history monitoring database comprises: acquiring data on the basis of set binocular camera equipment on an intelligent manufacturing production line to obtain the product acquisition data; and controlling and classifying the product collected data based on the intelligent manufacturing production line, and storing the product collected data in a historical monitoring database according to the collected time sequence after the classification is finished.
Specifically, a to-be-trained sample is extracted from a history monitoring database on an intelligent manufacturing production line, wherein the training sample is extracted according to a time sequence and a preset interval time length, so that to-be-trained sample monitoring data is obtained, the to-be-trained monitoring data is data which is not subjected to extracted training labels in the history monitoring database, and after the to-be-trained sample monitoring data is extracted, the extracted training labels need to be carried out in the history monitoring database.
The intelligent manufacturing production line is provided with monitoring equipment which can be binocular camera equipment generally, after the binocular camera equipment on the intelligent manufacturing production line collects product collection data, the product collection data are classified and labeled, and after the classification and labeling are completed, the product collection data are stored in a road monitoring database according to the collected time sequence.
S12: inputting the monitoring data of the sample to be trained into a preset deep neural network model for training and learning to obtain network node parameters of the deep neural network model after training and learning;
in the specific implementation process of the present invention, the training loss function of the deep neural network model is: introducing regularization among parameter vectors into a loss function, and carrying out regularization processing on parameters of all layers in the deep neural network model; after the regularization process, a new loss function is constructed as a training loss function of the deep neural network.
Further, the training loss function is as follows:
L＝Lcross-entropy+λR；
in a corresponding manner, the first and second optical fibers are,
wherein R is a regularization term of all layer parameter vectors, NlIs the number of l-th layer parameter vectors, Nl1,2,3, …, wherein M is the number of layers of the deep neural network model, M is 1,2,3, …, l is 1,2,3, …, M, W is the parameter of the neural network model to be trained, W is the parameter of the deep neural network model, M is the parameter ofliIs the ith parameter vector of the l layer in the model parameters of the deep neural network, WliIs the jth parameter vector of the L layer in the model parameters of the deep neural network, Lcross-entropyIs a cross entropy loss function, lambda is a hyperparameter, and L is a training loss function.
Further, the inputting the monitoring data of the sample to be trained into a preset deep neural network model for training and learning to obtain the network node parameters of the deep neural network model after training and learning includes: resetting coefficients of all layer parameter vectors of the preset deep neural network model by using a back propagation algorithm before training and learning of the preset deep neural network model; and inputting the monitoring data of the sample to be trained into the deep neural network model after the coefficients of all the layer parameter vectors are reset for training and learning, and obtaining the network node parameters of the deep neural network model after training and learning.
Further, in the process of resetting the coefficients of all layer parameter vectors of the preset deep neural network model by using the back propagation algorithm, a formula for resetting the coefficients of all layer parameter vectors of the preset deep neural network model is as follows:
wherein the content of the first and second substances,
for updated coefficients of the i-th and j-th parameter vectors in the trained neural network model, WijThe coefficients of the ith parameter vector and the jth parameter vector in the trained neural network model are represented, lr is the learning rate, weight _ decay is the weight attenuation coefficient, momentum is the impulse, and L is the updated loss function.
Specifically, the loss function in the deep neural network model is obtained by firstly introducing regularization among parameter vectors into the loss function and regularizing parameters of all layers in the deep neural network model; after the regularization process, a new loss function is constructed as a training loss function of the deep neural network.
And acquiring regularization items of all layer parameter vectors through regularization, and regularizing all layer parameter vectors, wherein the formula for acquiring the regularization items of all layer parameter vectors is as follows:
the existing training loss function is then as follows:
L＝Lcross-entropy+λR；
in a corresponding manner, the first and second optical fibers are,
wherein R is a regularization term of all layer parameter vectors, NlIs the number of l-th layer parameter vectors, Nl1,2,3, …, wherein M is the number of layers of the deep neural network model, M is 1,2,3, …, l is 1,2,3, …, M, W is the parameter of the neural network model to be trained, W is the parameter of the deep neural network model, M is the parameter ofliIs the ith parameter vector of the l layer in the model parameters of the deep neural network, WljIs the jth parameter vector of the L layer in the model parameters of the deep neural network, Lcross-entropyIs a cross entropy loss function, lambda is a hyperparameter, and L is a training loss function.
After the corresponding loss function is obtained, the loss function is fused in the preset deep neural network model, then the monitoring data of the sample to be trained is input into the preset deep neural network model for training and learning, and the network node parameters of the deep neural network model after training and learning are obtained.
Because the preset deep neural network model is trained correspondingly at intervals, coefficients of all layer parameter vectors need to be reset by using a back propagation algorithm before retraining; after resetting, inputting the sample monitoring data to be trained into the deep neural network model after the coefficients of all layer parameter vectors are reset for training and learning, and obtaining the network node parameters of the deep neural network model after training and learning.
In addition, in the process of resetting the coefficients of all layer parameter vectors of the preset deep neural network model by using a back propagation algorithm, a formula for resetting the coefficients of all layer parameter vectors of the preset deep neural network model is as follows:
wherein the content of the first and second substances,
for updated coefficients of the i-th and j-th parameter vectors in the trained neural network model, WijThe coefficients of the ith parameter vector and the jth parameter vector in the trained neural network model are represented, lr is the learning rate, weight _ decay is the weight attenuation coefficient, momentum is the impulse, and L is the updated loss function.
S13: carrying out model updating on the deep neural network model based on the network node parameters of the deep neural network model after training and learning;
in the specific implementation process of the invention, the network node parameters of the deep neural network model are obtained after training and learning, and the parameters of the original preset deep neural network model are updated by using the parameters, so that the updated deep neural network model is obtained.
S14: inputting the current product monitoring data collected on the intelligent manufacturing production line into the updated deep neural network model for product classification prediction, and outputting a product classification prediction result;
in the specific implementation process of the invention, after the current product monitoring data is collected on the intelligent manufacturing production line, the current product monitoring data is input into the updated deep neural network model for product classification prediction, and a product classification prediction result is output.
S15: and sending the product classification prediction result to a classification controller at the tail end of the intelligent manufacturing production line, and performing control classification by the classification controller based on the product classification prediction result.
In the specific implementation process of the present invention, the classification controller performs control classification based on the product classification prediction result, including: after receiving the product classification prediction result, the classification controller calls the video on the intelligent production line to monitor the current product for real-time positioning and tracking processing based on the current product monitoring data corresponding to the product classification prediction result, so as to obtain a real-time positioning and tracking result; and when the current product is transmitted to a preset position on the intelligent manufacturing production line, the classification controller controls the classification robot to perform control classification according to the product classification prediction result.
Specifically, after a product classification prediction result is obtained, the product classification prediction result needs to be transmitted to a classification controller at the tail end of an intelligent manufacturing production line in a wireless transmission mode, and after the classification controller receives the product classification prediction result, the current product on the intelligent production line is called to perform real-time positioning and tracking processing on the current product through current product monitoring data corresponding to the product classification prediction result, so that a real-time positioning and tracking result is obtained; and when the current product is tracked and transmitted to a classification preset position by the intelligent manufacturing production line, the classification controller controls the classification robot to perform control classification according to the product classification prediction result.
In the embodiment of the invention, the products on the intelligent manufacturing production line can be accurately and automatically classified by utilizing the deep neural network model to realize the classified prediction of the products on the intelligent manufacturing production line and then sending the classified prediction result of the products to the classification controller, so that the manual complexity is not required, and the labor cost is reduced.
Examples
Referring to fig. 2, fig. 2 is a schematic structural component diagram of a product classification system in an intelligent manufacturing line based on deep learning according to an embodiment of the present invention.
As shown in fig. 2, a product classification system for an intelligent manufacturing line based on deep learning, the system includes:
the sample data obtaining module 21: the system comprises a historical monitoring database, a training database and a training database, wherein the historical monitoring database is used for acquiring to-be-trained sample monitoring data on an intelligent manufacturing production line, and the to-be-trained sample monitoring data is monitoring data of the intelligent manufacturing production line within a preset period of time;
in the specific implementation process of the invention, the obtaining of the monitoring data of the sample to be trained on the historical monitoring database on the intelligent manufacturing production line comprises the following steps: extracting and processing samples to be trained in a historical monitoring database on the intelligent manufacturing production line according to a time sequence and a preset time length to obtain monitoring data of the samples to be trained; the monitoring data of the sample to be trained is data which is not subjected to extracted training labels in the historical monitoring database; and after the monitoring data of the sample to be trained is obtained, carrying out extracted training annotation in the historical monitoring database.
Further, the construction of the history monitoring database comprises: acquiring data on the basis of set binocular camera equipment on an intelligent manufacturing production line to obtain the product acquisition data; and controlling and classifying the product collected data based on the intelligent manufacturing production line, and storing the product collected data in a historical monitoring database according to the collected time sequence after the classification is finished.
Specifically, a to-be-trained sample is extracted from a history monitoring database on an intelligent manufacturing production line, wherein the training sample is extracted according to a time sequence and a preset interval time length, so that to-be-trained sample monitoring data is obtained, the to-be-trained monitoring data is data which is not subjected to extracted training labels in the history monitoring database, and after the to-be-trained sample monitoring data is extracted, the extracted training labels need to be carried out in the history monitoring database.
The intelligent manufacturing production line is provided with monitoring equipment which can be binocular camera equipment generally, after the binocular camera equipment on the intelligent manufacturing production line collects product collection data, the product collection data are classified and labeled, and after the classification and labeling are completed, the product collection data are stored in a road monitoring database according to the collected time sequence.
Training the learning module 22: the system comprises a deep neural network model, a network node parameter acquisition module, a data acquisition module and a data processing module, wherein the deep neural network model is used for acquiring a deep neural network model to be trained;
in the specific implementation process of the present invention, the training loss function of the deep neural network model is: introducing regularization among parameter vectors into a loss function, and carrying out regularization processing on parameters of all layers in the deep neural network model; after the regularization process, a new loss function is constructed as a training loss function of the deep neural network.
Further, the training loss function is as follows:
L＝Lcross-entropy+λR；
in a corresponding manner, the first and second optical fibers are,
wherein R is a regularization term of all layer parameter vectors, NlIs the number of l-th layer parameter vectors, Nl1,2,3, …, wherein M is the number of layers of the deep neural network model, M is 1,2,3, …, l is 1,2,3, …, M, W is the parameter of the neural network model to be trained, W is the parameter of the deep neural network model, M is the parameter ofliIs the ith parameter vector of the l layer in the model parameters of the deep neural network, WljIs the jth parameter vector of the L layer in the model parameters of the deep neural network, Lcross-entropyIs a cross entropy loss function, lambda is a hyperparameter, and L is a training loss function.
Further, the inputting the monitoring data of the sample to be trained into a preset deep neural network model for training and learning to obtain the network node parameters of the deep neural network model after training and learning includes: resetting coefficients of all layer parameter vectors of the preset deep neural network model by using a back propagation algorithm before training and learning of the preset deep neural network model; and inputting the monitoring data of the sample to be trained into the deep neural network model after the coefficients of all the layer parameter vectors are reset for training and learning, and obtaining the network node parameters of the deep neural network model after training and learning.
Further, in the process of resetting the coefficients of all layer parameter vectors of the preset deep neural network model by using the back propagation algorithm, a formula for resetting the coefficients of all layer parameter vectors of the preset deep neural network model is as follows:
wherein the content of the first and second substances,
for updated coefficients of the i-th and j-th parameter vectors in the trained neural network model, WijThe coefficients of the ith parameter vector and the jth parameter vector in the trained neural network model are represented, lr is the learning rate, weight _ decay is the weight attenuation coefficient, momentum is the impulse, and L is the updated loss function.
Specifically, the loss function in the deep neural network model is obtained by firstly introducing regularization among parameter vectors into the loss function and regularizing parameters of all layers in the deep neural network model; after the regularization process, a new loss function is constructed as a training loss function of the deep neural network.
And acquiring regularization items of all layer parameter vectors through regularization, and regularizing all layer parameter vectors, wherein the formula for acquiring the regularization items of all layer parameter vectors is as follows:
the existing training loss function is then as follows:
L＝Lcross-entropy+λR；
in a corresponding manner, the first and second optical fibers are,
wherein R is a regularization term of all layer parameter vectors, NlIs the number of l-th layer parameter vectors, Nl1,2,3, …, wherein M is the number of layers of the deep neural network model, M is 1,2,3, …, l is 1,2,3, …, M, W is the parameter of the neural network model to be trained, W is the parameter of the deep neural network model, M is the parameter ofliIs the ith parameter vector of the l layer in the model parameters of the deep neural network, WljIs the jth parameter vector of the L layer in the model parameters of the deep neural network, Lcross-entropyIs a cross entropy loss function, lambda is a hyperparameter, and L is a training loss function.
After the corresponding loss function is obtained, the loss function is fused in the preset deep neural network model, then the monitoring data of the sample to be trained is input into the preset deep neural network model for training and learning, and the network node parameters of the deep neural network model after training and learning are obtained.
Because the preset deep neural network model is trained correspondingly at intervals, coefficients of all layer parameter vectors need to be reset by using a back propagation algorithm before retraining; after resetting, inputting the sample monitoring data to be trained into the deep neural network model after the coefficients of all layer parameter vectors are reset for training and learning, and obtaining the network node parameters of the deep neural network model after training and learning.
In addition, in the process of resetting the coefficients of all layer parameter vectors of the preset deep neural network model by using a back propagation algorithm, a formula for resetting the coefficients of all layer parameter vectors of the preset deep neural network model is as follows:
wherein the content of the first and second substances,
for updated coefficients of the i-th and j-th parameter vectors in the trained neural network model, WijThe coefficients of the ith parameter vector and the jth parameter vector in the trained neural network model are represented, lr is the learning rate, weight _ decay is the weight attenuation coefficient, momentum is the impulse, and L is the updated loss function.
The update module 23: the deep neural network model is updated based on the network node parameters of the deep neural network model after training and learning;
in the specific implementation process of the invention, the network node parameters of the deep neural network model are obtained after training and learning, and the parameters of the original preset deep neural network model are updated by using the parameters, so that the updated deep neural network model is obtained.
The product classification prediction module 24: the system comprises a deep neural network model, a product classification prediction module and a product classification prediction module, wherein the deep neural network model is used for inputting current product monitoring data acquired on the intelligent manufacturing production line into the updated deep neural network model for product classification prediction and outputting a product classification prediction result;
in the specific implementation process of the invention, after the current product monitoring data is collected on the intelligent manufacturing production line, the current product monitoring data is input into the updated deep neural network model for product classification prediction, and a product classification prediction result is output.
The control classification module 25: the classification controller is used for sending the product classification prediction result to the classification controller at the tail end of the intelligent manufacturing production line, and the classification controller carries out control classification based on the product classification prediction result.
In the specific implementation process of the present invention, the classification controller performs control classification based on the product classification prediction result, including: after receiving the product classification prediction result, the classification controller calls the video on the intelligent production line to monitor the current product for real-time positioning and tracking processing based on the current product monitoring data corresponding to the product classification prediction result, so as to obtain a real-time positioning and tracking result; and when the current product is transmitted to a preset position on the intelligent manufacturing production line, the classification controller controls the classification robot to perform control classification according to the product classification prediction result.
Specifically, after a product classification prediction result is obtained, the product classification prediction result needs to be transmitted to a classification controller at the tail end of an intelligent manufacturing production line in a wireless transmission mode, and after the classification controller receives the product classification prediction result, the current product on the intelligent production line is called to perform real-time positioning and tracking processing on the current product through current product monitoring data corresponding to the product classification prediction result, so that a real-time positioning and tracking result is obtained; and when the current product is tracked and transmitted to a classification preset position by the intelligent manufacturing production line, the classification controller controls the classification robot to perform control classification according to the product classification prediction result.
In the embodiment of the invention, the products on the intelligent manufacturing production line can be accurately and automatically classified by utilizing the deep neural network model to realize the classified prediction of the products on the intelligent manufacturing production line and then sending the classified prediction result of the products to the classification controller, so that the manual complexity is not required, and the labor cost is reduced.
Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable storage medium, and the storage medium may include: a Read Only Memory (ROM), a Random Access Memory (RAM), a magnetic or optical disk, or the like.
In addition, the method and the system for classifying products on an intelligent manufacturing line based on deep learning according to the embodiments of the present invention are described in detail, and a specific embodiment is used herein to explain the principle and the implementation of the present invention, and the description of the embodiments is only used to help understanding the method and the core idea of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (9)
1. A product classification method on an intelligent manufacturing line based on deep learning is characterized by comprising the following steps:
acquiring monitoring data of a sample to be trained on a historical monitoring database on an intelligent manufacturing production line, wherein the monitoring data of the sample to be trained is monitoring data of the intelligent manufacturing production line within a preset period of time;
inputting the monitoring data of the sample to be trained into a preset deep neural network model for training and learning to obtain network node parameters of the deep neural network model after training and learning;
carrying out model updating on the deep neural network model based on the network node parameters of the deep neural network model after training and learning;
inputting the current product monitoring data collected on the intelligent manufacturing production line into the updated deep neural network model for product classification prediction, and outputting a product classification prediction result;
and sending the product classification prediction result to a classification controller at the tail end of the intelligent manufacturing production line, and performing control classification by the classification controller based on the product classification prediction result.
2. The method for classifying products on an intelligent manufacturing line according to claim 1, wherein the obtaining of the sample monitoring data to be trained on the historical monitoring database on the intelligent manufacturing line comprises:
extracting and processing samples to be trained in a historical monitoring database on the intelligent manufacturing production line according to a time sequence and a preset time length to obtain monitoring data of the samples to be trained;
the monitoring data of the sample to be trained is data which is not subjected to extracted training labels in the historical monitoring database;
and after the monitoring data of the sample to be trained is obtained, carrying out extracted training annotation in the historical monitoring database.
3. The method for product classification on an intelligent manufacturing line according to claim 1, wherein the construction of the history monitoring database comprises:
acquiring data on the basis of set binocular camera equipment on an intelligent manufacturing production line to obtain the product acquisition data;
and controlling and classifying the product collected data based on the intelligent manufacturing production line, and storing the product collected data in a historical monitoring database according to the collected time sequence after the classification is finished.
4. The method of claim 1, wherein the training loss function of the deep neural network model is:
introducing regularization among parameter vectors into a loss function, and carrying out regularization processing on parameters of all layers in the deep neural network model;
after the regularization process, a new loss function is constructed as a training loss function of the deep neural network.
5. The method of claim 4, wherein the training loss function is as follows:
L＝Lcross-entropy+λR；
in a corresponding manner, the first and second optical fibers are,
wherein R is a regularization term of all layer parameter vectors, NlIs the number of l-th layer parameter vectors, Nl1,2,3, …, wherein M is the number of layers of the deep neural network model, M is 1,2,3, …, l is 1,2,3, …, M, W is the parameter of the neural network model to be trained, W is the parameter of the deep neural network model, M is the parameter ofliIs the ith parameter vector of the l layer in the model parameters of the deep neural network, WljIs the jth parameter vector of the L layer in the model parameters of the deep neural network, Lcross-entropyIs a cross entropy loss function, lambda is a hyperparameter, and L is a training loss function.
6. The method for classifying products on an intelligent manufacturing line according to claim 1, wherein the step of inputting the monitoring data of the sample to be trained into a preset deep neural network model for training and learning to obtain the network node parameters of the deep neural network model after training and learning comprises the steps of:
resetting coefficients of all layer parameter vectors of the preset deep neural network model by using a back propagation algorithm before training and learning of the preset deep neural network model;
and inputting the monitoring data of the sample to be trained into the deep neural network model after the coefficients of all the layer parameter vectors are reset for training and learning, and obtaining the network node parameters of the deep neural network model after training and learning.
7. The method for classifying products on an intelligent manufacturing line according to claim 6, wherein in the process of resetting the coefficients of all layer parameter vectors of the preset deep neural network model by using the back propagation algorithm, the formula for resetting the coefficients of all layer parameter vectors of the preset deep neural network model is as follows:
wherein the content of the first and second substances,
for updated coefficients of the i-th and j-th parameter vectors in the trained neural network model, WijThe coefficients of the ith parameter vector and the jth parameter vector in the trained neural network model are represented, lr is the learning rate, weight _ decay is the weight attenuation coefficient, momentum is the impulse, and L is the updated loss function.
8. The method of claim 1, wherein the classification controller performs the controlled classification based on the product classification prediction result, and comprises:
after receiving the product classification prediction result, the classification controller calls the video on the intelligent production line to monitor the current product for real-time positioning and tracking processing based on the current product monitoring data corresponding to the product classification prediction result, so as to obtain a real-time positioning and tracking result;
and when the current product is transmitted to a preset position on the intelligent manufacturing production line, the classification controller controls the classification robot to perform control classification according to the product classification prediction result.
9. An intelligent in-line manufacturing product classification system based on deep learning, the system comprising:
a sample data obtaining module: the system comprises a historical monitoring database, a training database and a training database, wherein the historical monitoring database is used for acquiring to-be-trained sample monitoring data on an intelligent manufacturing production line, and the to-be-trained sample monitoring data is monitoring data of the intelligent manufacturing production line within a preset period of time;
training a learning module: the system comprises a deep neural network model, a network node parameter acquisition module, a data acquisition module and a data processing module, wherein the deep neural network model is used for acquiring a deep neural network model to be trained;
an update module: the deep neural network model is updated based on the network node parameters of the deep neural network model after training and learning;
a product classification prediction module: the system comprises a deep neural network model, a product classification prediction module and a product classification prediction module, wherein the deep neural network model is used for inputting current product monitoring data acquired on the intelligent manufacturing production line into the updated deep neural network model for product classification prediction and outputting a product classification prediction result;
a control classification module: the classification controller is used for sending the product classification prediction result to the classification controller at the tail end of the intelligent manufacturing production line, and the classification controller carries out control classification based on the product classification prediction result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010398375.3A CN111680717A (en) | 2020-05-12 | 2020-05-12 | Product classification method and system on intelligent manufacturing production line based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010398375.3A CN111680717A (en) | 2020-05-12 | 2020-05-12 | Product classification method and system on intelligent manufacturing production line based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111680717A true CN111680717A (en) | 2020-09-18 |
Family
ID=72433922
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010398375.3A Pending CN111680717A (en) | 2020-05-12 | 2020-05-12 | Product classification method and system on intelligent manufacturing production line based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111680717A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101782976A (en) * | 2010-01-15 | 2010-07-21 | 南京邮电大学 | Automatic selection method for machine learning in cloud computing environment |
| CN102393908A (en) * | 2011-06-29 | 2012-03-28 | 湖南大学 | Method for identifying three bottles on mixed production line based on machine vision detection |
| CN108805259A (en) * | 2018-05-23 | 2018-11-13 | 北京达佳互联信息技术有限公司 | neural network model training method, device, storage medium and terminal device |
-
2020
- 2020-05-12 CN CN202010398375.3A patent/CN111680717A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101782976A (en) * | 2010-01-15 | 2010-07-21 | 南京邮电大学 | Automatic selection method for machine learning in cloud computing environment |
| CN102393908A (en) * | 2011-06-29 | 2012-03-28 | 湖南大学 | Method for identifying three bottles on mixed production line based on machine vision detection |
| CN108805259A (en) * | 2018-05-23 | 2018-11-13 | 北京达佳互联信息技术有限公司 | neural network model training method, device, storage medium and terminal device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108805258A (en) | A kind of neural network training method and its device, computer server | |
| CN110443969B (en) | Fire detection method and device, electronic equipment and storage medium | |
| CN109990790B (en) | Unmanned aerial vehicle path planning method and device | |
| CN110751318B (en) | Ultra-short-term power load prediction method based on IPSO-LSTM | |
| CN112685950B (en) | Method, system and equipment for detecting abnormality of ocean time sequence observation data | |
| CN114169638A (en) | Water quality prediction method and device | |
| CN109978870A (en) | Method and apparatus for output information | |
| CN112464718B (en) | Target detection method based on YOLO-Terse network and storage medium | |
| CN114297907A (en) | Greenhouse environment spatial distribution prediction method and device | |
| CN110188936A (en) | Short-time Traffic Flow Forecasting Methods based on multifactor spatial choice deep learning algorithm | |
| CN115829024B (en) | Model training method, device, equipment and storage medium | |
| CN107945534A (en) | A kind of special bus method for predicting based on GMDH neutral nets | |
| CN111582395A (en) | Product quality classification system based on convolutional neural network | |
| CN114821154A (en) | Grain depot ventilation window state detection algorithm based on deep learning | |
| CN114565149A (en) | CGA fusion model-based time series data prediction method and device and computer equipment | |
| CN112509392B (en) | Robot behavior teaching method based on meta-learning | |
| CN111832479B (en) | Video target detection method based on improved self-adaptive anchor point R-CNN | |
| CN111680717A (en) | Product classification method and system on intelligent manufacturing production line based on deep learning | |
| CN111723742A (en) | Crowd density analysis method, system and device and computer readable storage medium | |
| CN115800274A (en) | Automatic self-adaption method and device for feeder of 5G power distribution network and storage medium | |
| CN112884160B (en) | Meta learning method and related device | |
| CN112532429B (en) | Multivariable QoS prediction method based on position information | |
| CN115052154A (en) | Model training and video coding method, device, equipment and storage medium | |
| CN115062759A (en) | Fault diagnosis method based on improved long and short memory neural network | |
| Zhang et al. | Image recognition of supermarket shopping robot based on CNN |
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 |