CN114821025A - Meter identification method and system based on deep learning - Google Patents

Meter identification method and system based on deep learning Download PDF

Info

Publication number
CN114821025A
CN114821025A CN202210261718.0A CN202210261718A CN114821025A CN 114821025 A CN114821025 A CN 114821025A CN 202210261718 A CN202210261718 A CN 202210261718A CN 114821025 A CN114821025 A CN 114821025A
Authority
CN
China
Prior art keywords
module
meter
image analysis
model
analysis module
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.)
Withdrawn
Application number
CN202210261718.0A
Other languages
Chinese (zh)
Inventor
王坤
甄超
李坚林
季坤
王刘芳
黄杰
谢佳
柯艳国
王翀
赵常威
钱宇骋
丁国成
王署东
杨海涛
张晨晨
李森林
胡啸宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Super High Voltage Branch Of State Grid Anhui Electric Power Co ltd
Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd
State Grid Anhui Electric Power Co Ltd
Huoqiu Power Supply Co of State Grid Anhui Electric Power Co Ltd
Original Assignee
Super High Voltage Branch Of State Grid Anhui Electric Power Co ltd
Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd
State Grid Anhui Electric Power Co Ltd
Huoqiu Power Supply Co of State Grid Anhui Electric Power Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Super High Voltage Branch Of State Grid Anhui Electric Power Co ltd, Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd, State Grid Anhui Electric Power Co Ltd, Huoqiu Power Supply Co of State Grid Anhui Electric Power Co Ltd filed Critical Super High Voltage Branch Of State Grid Anhui Electric Power Co ltd
Priority to CN202210261718.0A priority Critical patent/CN114821025A/en
Publication of CN114821025A publication Critical patent/CN114821025A/en
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10Character recognition
    • G06V30/14Image acquisition
    • G06V30/148Segmentation of character regions
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00Arrangements for image or video recognition or understanding
    • G06V10/20Image preprocessing
    • G06V10/22Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00Arrangements for image or video recognition or understanding
    • G06V10/70Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a meter identification method and system based on deep learning, which relate to the technical field of image identification and comprise an image analysis module, an information recording module, an information sorting module and a signal verification module; the image analysis module is used for carrying out digital target recognition on the received meter image by adopting a yolov3 detection model to obtain a digital recognition result and finally obtain a meter reading; the information recording module is used for recording the identification record of the image analysis module; the information arrangement module is used for arranging the identification records, constructing a parameter detection training sample, obtaining a parameter compensation model and feeding the parameter compensation model back to the model construction module to correct the yolov3 detection model, so that the identification accuracy is improved; the signal verification module is used for verifying the communication state of the image analysis module in real time, and after a communication early warning instruction is detected, the image analysis module enters an active standby mode, so that the influence of interference signals is effectively reduced, and the identification efficiency and accuracy are improved.

Description

Meter identification method and system based on deep learning
Technical Field
The invention relates to the technical field of image recognition, in particular to a meter recognition method and system based on deep learning.
Background
In the power grid operation and maintenance inspection, the pointer instrument has the characteristics of strong anti-electromagnetic interference capability, high precision, low price and the like, and is still a main measuring instrument for industrial production for a quite long time; because the pointer instrument can not output digital signals, reading and checking of the current meter are manually checked, an operator is required to stay in a transformer substation all the time or often, the workload of the operator is increased, and the effect of unmanned inspection can not be achieved; and the instruments installed in the high-temperature high-pressure environment are inconvenient to observe;
with the rapid development of the robot technology, the inspection robot can replace manpower, capture a meter reading image through a camera or a thermal infrared imager, and finally perform image processing on the obtained image; however, the image recognition process is often influenced by the external environment, which greatly influences the recognition precision; based on the defects, the invention provides a meter identification method and system based on deep learning.
Disclosure of Invention
The present invention is directed to solving at least one of the problems of the prior art. Therefore, the invention provides a meter identification method and system based on deep learning.
In order to achieve the above object, an embodiment according to a first aspect of the present invention provides a meter recognition system based on deep learning, including an image acquisition module, a model construction module, an information recording module, a parameter compensation module, and a signal verification module;
the image acquisition module is used for acquiring meter images and transmitting the acquired meter images to the image analysis module for learning and identification; the image analysis module is connected with the model construction module and is used for acquiring a yolov3 detection model which is constructed by the model construction module and used for training and identifying images;
the image analysis module is used for carrying out digital target recognition on the received meter image by adopting a yolov3 detection model to obtain a digital recognition result and finally obtain a meter reading;
the information recording module is used for recording the identification record of the image analysis module and transmitting the identification record to the information sorting module; the information sorting module is used for sorting the identification records, constructing a parameter detection training sample, and training based on a machine learning method to obtain a parameter compensation model; the parameter compensation module receives the parameter compensation model and then corrects the yolov3 detection model;
the signal verification module is used for verifying the communication state of the image analysis module in real time and calculating to obtain an interference coefficient Cy; and if the Cy is more than or equal to the interference threshold, judging that the signal interference is serious and the communication state is abnormal, and generating a communication early warning instruction to remind a manager to process as soon as possible.
Further, the yolov3 test model is constructed as follows:
acquiring a meter image, and constructing a data set required by training; the concrete expression is as follows: collecting n meter pictures under different shooting angles and different lighting conditions, and labeling the meter pictures to obtain a data set required by subsequent training of a yolov3 model; dividing the training set, the test set and the check set according to a set proportion; the set ratio comprises 2:1:1, 3:1:1 and 4:3: 1;
constructing a fusion model: the fusion model is constructed by combining at least two fusion modes of a support vector machine, a deep convolutional neural network and an RBF neural network, wherein the fusion modes comprise a linear weighted fusion method, a cross fusion method, a waterfall fusion method, a feature fusion method and a prediction fusion method;
and (3) carrying out data normalization on the training set, the test set and the check set, then training, testing and checking the fusion model, and marking the trained fusion model as a yolov3 detection model.
Further, the specific correction steps of the parameter compensation module are as follows:
SS 1: acquiring an identification record of an image analysis module at the current moment, and inputting meter reading obtained by current identification, actual meter reading and corresponding various environmental parameter values into a parameter compensation model to obtain an attribute parameter compensation coefficient;
SS 2: and correcting the yolov3 detection model according to the attribute parameter compensation coefficient.
Further, the specific verification steps of the signal verification module are as follows:
the signal verification module sends a verification configuration message to an FPGA main control of the image analysis module according to a preset verification period, wherein the verification configuration message comprises a first signal quality threshold;
responding to the received verification configuration message, and sending a second synchronous signal to the signal verification module by the FPGA main control; the signal quality of the second synchronous signal is determined by the signal verification module and is compared with the first signal quality threshold, and a corresponding quality difference value Z1 is obtained; setting the response time length as XT;
calculating a signal loss index SH by using a formula SH (Z1 × a1+ XT × a 2), wherein a1 and a2 are coefficient factors; the interference coefficient Cy is evaluated according to the trend of the signal loss index SH.
Further, the specific evaluation process of the interference coefficient Cy is as follows:
establishing a curve graph of the change of the signal loss index SH along with time, and comparing the signal loss index SH with a loss threshold value; if SH is larger than or equal to the loss threshold, intercepting a corresponding curve segment in a corresponding curve graph, marking the curve segment as yellow and marking the curve segment as an interference curve segment;
counting the number of the interference curve segments to be L1 in a preset time period, and integrating all the interference curve segments with time to obtain interference reference energy L2; the interference coefficient Cy is calculated by using Cy-L1 × a3+ L2 × a4, where a3 and a4 are coefficient factors.
Further, the identification record comprises meter reading obtained by the image analysis module at each time of identification, actual meter reading and corresponding environmental parameter values; the environmental parameter values comprise temperature, humidity, wind pressure, wind speed and interference signals.
Further, after detecting a communication early warning instruction, the image analysis module enters an active standby mode, namely, the FPGA master control of abnormal communication is not used for identifying meter images; and after the signal verification module judges that the communication state is normal, continuing the communication between the signal verification module and the communication module.
Further, a meter identification method based on deep learning comprises the following steps:
the method comprises the following steps: a yolov3 detection model which can train and identify images is constructed through a model construction module, and the constructed yolov3 detection model is transmitted to an image analysis module;
step two: performing digital target recognition on the meter image acquired by the image acquisition module by using a yolov3 detection model through an image analysis module to obtain a digital recognition result, and finally obtaining meter reading;
step three: recording the identification records of the image analysis module through the information recording module, and sorting the identification records by calling the information sorting module to construct a parameter detection training sample; training a machine learning-based method to obtain a parameter compensation model;
step four: feeding the parameter compensation model back to the model construction module through the parameter compensation module so as to correct the yolov3 detection model;
step five: verifying the communication state of the image analysis module in real time through a signal verification module, and calculating to obtain an interference coefficient Cy; if the Cy is larger than or equal to the interference threshold, judging that the signal interference is serious and the communication state is abnormal, and generating a communication early warning instruction to remind a manager to process as soon as possible;
step six: and after the image analysis module detects a communication early warning instruction, the image analysis module enters an active standby mode, and after the signal verification module judges that the communication state is normal, the communication between the image analysis module and the signal verification module is continued.
Compared with the prior art, the invention has the beneficial effects that:
1. the camera ball in the image acquisition module can rotate in the shell to acquire meter images at different shooting angles, so that firm data support is provided for subsequent image identification, and the image analysis module adopts a yolov3 detection model to analyze the acquired meter images and identify the reading of the meter, so that the aim of unmanned inspection is fulfilled, and the identification efficiency is effectively improved;
2. the information recording module is used for recording the identification record of the image analysis module; the information sorting module is used for sorting the identification records, constructing a parameter detection training sample, and training based on a machine learning method to obtain a parameter compensation model; the parameter compensation module receives the parameter compensation model and then corrects the yolov3 detection model, so that the yolov3 detection model is continuously corrected and perfected, the accuracy of identification is continuously improved, the whole process is automatically carried out, the work can be widely carried out, and the continuity and the effectiveness of the work are guaranteed;
3. the signal verification module is used for verifying the communication state of the image analysis module in real time, firstly, a verification configuration message is sent to an FPGA main control of the image analysis module according to a preset verification period, and a signal loss index SH is obtained by combining a corresponding quality difference value Z1 and a response time length XT; calculating to obtain an interference coefficient Cy according to the change trend of the signal loss index SH, if the Cy is more than or equal to an interference threshold, judging that the signal interference is serious and the communication state is abnormal, and generating a communication early warning instruction to prompt a manager to process as soon as possible; the influence of interference signals is effectively reduced, and therefore the identification efficiency and the accuracy of the image analysis module are improved.
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 system block diagram of a deep learning-based meter identification system according to the present invention.
FIG. 2 is a flowchart of a table meter identification method based on deep learning according to the present invention.
Detailed Description
The technical solutions of the present invention will be described clearly and completely with reference to the following embodiments, and it should be understood 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 obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, belong to the protection scope of the present invention.
As shown in fig. 1 to 2, a meter recognition system based on deep learning includes an image acquisition module, an image analysis module, a model construction module, a controller, a display module, an information recording module, an information sorting module, a parameter compensation module, a signal verification module, and an alarm module;
the image acquisition module is used for acquiring meter images and transmitting the acquired meter images to the image analysis module for learning and identification;
the image analysis module is connected with the model construction module, the model construction module is used for constructing a yolov3 detection model capable of training and identifying images, and the yolov3 detection model is used for analyzing the acquired meter images and identifying the reading of the meter; the yolov3 detection model is constructed as follows:
acquiring a meter image, and constructing a data set required by training; the concrete expression is as follows: collecting n meter pictures under different shooting angles and different lighting conditions, and labeling the meter pictures to obtain a data set required by subsequent training of a yolov3 model; dividing the training set, the test set and the check set according to a set proportion; the set ratio comprises 2:1:1, 3:1:1 and 4:3: 1;
constructing a fusion model: the fusion model is constructed by combining at least two fusion modes of a support vector machine, a deep convolutional neural network and an RBF neural network, wherein the fusion modes comprise a linear weighted fusion method, a cross fusion method, a waterfall fusion method, a feature fusion method and a prediction fusion method;
training, testing and verifying the fusion model after the training set, the testing set and the check set are subjected to data normalization, and marking the trained fusion model as a yolov3 detection model;
wherein, the yolov3 detection model uses the characteristic extraction network dark net-53, which uses the form of 3 × 3 and 1 × 1 convolution layer connected in turn, and has 53 convolution layers;
the image analysis module is used for carrying out digital target recognition on the received meter image by adopting a yolov3 detection model to obtain a digital recognition result and finally obtain a meter reading; the image analysis module is used for transmitting the finally obtained meter reading to the display module through the controller for real-time display;
the image acquisition module comprises a shell and a camera ball movably arranged in the shell, wherein a camera port is arranged on the camera ball, and a camera is arranged on the camera port;
in the scheme, the camera ball can rotate in the shell, and a user can adjust the camera angle of the camera by adjusting the camera ball, so that the reading image of the meter can be better acquired, and meanwhile, the reading of the meter in the transformer substation is identified through the image, and the purpose of unmanned inspection is achieved;
the information recording module is connected with the display module and used for recording the identification records of the image analysis module and transmitting the identification records to the information sorting module, and the identification records comprise meter readings, actual meter readings and corresponding various environmental parameter values, which are obtained by the image analysis module in each identification; various environmental parameter values comprise temperature, humidity, wind pressure, wind speed, interference signals and the like;
the information arrangement module receives the identification records and arranges the identification records, a parameter detection training sample is constructed, training is based on a machine learning method, a parameter compensation model is obtained and is transmitted to the parameter compensation module, the parameter compensation module receives the parameter compensation model and then corrects the yolov3 detection model, and the specific compensation steps are as follows:
SS 1: acquiring an identification record of an image analysis module at the current moment, and inputting meter reading obtained by current identification, actual meter reading and corresponding various environmental parameter values into a parameter compensation model to obtain an attribute parameter compensation coefficient;
SS 2: correcting the yolov3 detection model according to the attribute parameter compensation coefficient, so that the yolov3 detection model is continuously corrected and perfected, and the identification accuracy is improved;
according to the method, the results of each link of the image analysis module are used as feedback factors, and then the yolov3 detection model in the previous training and recognition processes is further verified, attribute parameters are compensated, so that the yolov3 detection model is continuously corrected and perfected, the recognition accuracy is continuously improved, the whole process is automatically carried out, the work can be widely carried out, and the continuity and the effectiveness of the work are guaranteed;
in the embodiment, in order to reduce the influence of the interference signal, the accuracy of image recognition is ensured; the signal verification module is used for verifying the communication state of the image analysis module in real time, and the specific verification steps are as follows:
the signal verification module sends a verification configuration message to an FPGA main control of the image analysis module according to a preset verification period, wherein the verification configuration message comprises a first signal quality threshold; responding to the received verification configuration message sent by the signal verification module, and sending a second synchronous signal to the signal verification module by the FPGA main control;
in response to the second synchronization signal being monitored, the signal verification module determines the signal quality of the second synchronization signal, and compares the signal quality of the second synchronization signal with the first signal quality threshold to obtain a corresponding quality difference value Z1; wherein any metric known in the art can be used to characterize signal quality, such as RSRQ, RSRP, RSSI, and so forth, as will be appreciated by those skilled in the art; the quality difference can reflect the attenuation of the signal in the transmission process;
calculating the time difference between the moment when the signal verification module sends the verification configuration message and the moment when the signal verification module monitors the second synchronous signal again to obtain a response time length XT; calculating a signal loss index SH by using a formula SH (Z1 × a1+ XT × a 2), wherein a1 and a2 are coefficient factors;
establishing a curve graph of the change of the signal loss index SH along with time, and comparing the signal loss index SH with a loss threshold value; if SH is larger than or equal to the loss threshold, intercepting a corresponding curve segment in a corresponding curve graph, marking the curve segment as yellow and marking the curve segment as an interference curve segment;
counting the number of the interference curve segments to be L1 in a preset time period, and integrating all the interference curve segments with time to obtain interference reference energy L2; calculating an interference coefficient Cy by using a Cy-L1 × a3+ L2 × a4, wherein a3 and a4 are coefficient factors;
comparing the interference coefficient Cy with an interference threshold value; if the Cy is larger than or equal to the interference threshold, judging that the signal interference is serious and the communication state is abnormal, and generating a communication early warning instruction;
the signal verification module is used for transmitting the communication early warning instruction to the controller, and the controller automatically drives the alarm module to give an alarm after receiving the communication early warning instruction so as to remind a manager to process as soon as possible; therefore, the recognition efficiency and the accuracy of the image analysis module are improved;
after detecting the communication early warning instruction, the image analysis module enters an active standby mode, namely, the FPGA master control of abnormal communication is not used for identifying the meter image, and after the signal verification module judges that the communication state is normal, the communication between the image analysis module and the meter image is continued.
A meter recognition method based on deep learning is applied to the meter recognition system and comprises the following steps:
the method comprises the following steps: a yolov3 detection model which can train and identify images is constructed through a model construction module, and the constructed yolov3 detection model is transmitted to an image analysis module;
step two: the meter images collected by the image collection module are learned and identified through the image analysis module, and the collected meter images are analyzed and the meter reading is identified through the yolov3 detection model; the finally obtained meter reading is transmitted to a display module through a controller to be displayed in real time;
step three: recording the identification record of the image analysis module through the information recording module and transmitting the identification record to the information sorting module; the information sorting module receives the identification records and sorts the identification records, a parameter detection training sample is constructed, and training is based on a machine learning method to obtain a parameter compensation model;
step four: the yolov3 detection model is corrected by a parameter compensation module through a parameter compensation model, so that the yolov3 detection model is continuously corrected and perfected, and the identification accuracy is improved;
step five: verifying the communication state of the image analysis module in real time through a signal verification module, and calculating to obtain an interference coefficient Cy; if the Cy is larger than or equal to the interference threshold, judging that the signal interference is serious and the communication state is abnormal, and generating a communication early warning instruction to remind a manager to process as soon as possible;
step six: after detecting the communication early warning instruction, the image analysis module enters an active standby mode, namely, the FPGA master control of abnormal communication is not used for identifying the meter image, and after the signal verification module judges that the communication state is normal, the communication between the image analysis module and the meter image is continued.
The above formulas are all calculated by removing dimensions and taking numerical values thereof, the formula is a formula which is obtained by acquiring a large amount of data and performing software simulation to obtain the closest real situation, and the preset parameters and the preset threshold value in the formula are set by the technical personnel in the field according to the actual situation or obtained by simulating a large amount of data.
The working principle of the invention is as follows:
during working, firstly, a model construction module is used for constructing a yolov3 detection model capable of training and identifying images; the image acquisition module is used for acquiring meter images and transmitting the acquired meter images to the image analysis module for learning and identification; the image analysis module adopts a yolov3 detection model to analyze the acquired meter image and identify the meter reading, and transmits the finally obtained meter reading to the display module through the controller for real-time display, thereby achieving the purpose of unmanned inspection;
the information recording module is connected with the display module and is used for recording the identification record of the image analysis module and transmitting the identification record to the information sorting module; the information sorting module receives the identification records and sorts the identification records, a parameter detection training sample is constructed, and training is based on a machine learning method to obtain a parameter compensation model; the parameter compensation module receives the parameter compensation model and then corrects the yolov3 detection model, so that the yolov3 detection model is continuously corrected and perfected, the accuracy of identification is continuously improved, the whole process is automatically carried out, the work can be widely carried out, and the continuity and the effectiveness of the work are guaranteed;
the signal verification module is used for verifying the communication state of the image analysis module in real time, firstly, a verification configuration message is sent to an FPGA main control of the image analysis module according to a preset verification period, and a second synchronous signal is sent to the signal verification module by the FPGA main control in response to the fact that the verification configuration message sent by the signal verification module is received; calculating to obtain a signal loss index SH by combining the corresponding quality difference Z1 and the response time length XT; calculating to obtain an interference coefficient Cy according to the change trend of the signal loss index SH, if the Cy is more than or equal to an interference threshold, judging that the signal interference is serious and the communication state is abnormal, and generating a communication early warning instruction to prompt a manager to process as soon as possible; the influence of interference signals is effectively reduced, and therefore the identification efficiency and the accuracy of the image analysis module are improved.
In the description herein, references to the description of "one embodiment," "an example," "a specific example" or the like are intended to mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
The preferred embodiments of the invention disclosed above are intended to be illustrative only. The preferred embodiments are not intended to be exhaustive or to limit the invention to the precise forms disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims (8)

1. A meter identification system based on deep learning is characterized by comprising an image acquisition module, a model construction module, an information recording module, a parameter compensation module and a signal verification module;
the image acquisition module is used for acquiring meter images and transmitting the acquired meter images to the image analysis module for learning and identification; the image analysis module is connected with the model construction module and is used for acquiring a yolov3 detection model which is constructed by the model construction module and used for training and identifying images;
the image analysis module is used for carrying out digital target recognition on the received meter image by adopting a yolov3 detection model to obtain a digital recognition result and finally obtain a meter reading;
the information recording module is used for recording the identification record of the image analysis module and transmitting the identification record to the information sorting module; the information sorting module is used for sorting the identification records, constructing a parameter detection training sample, and training based on a machine learning method to obtain a parameter compensation model; the parameter compensation module receives the parameter compensation model and then corrects the yolov3 detection model;
the signal verification module is used for verifying the communication state of the image analysis module in real time and calculating to obtain an interference coefficient Cy; and if the Cy is more than or equal to the interference threshold, judging that the signal interference is serious and the communication state is abnormal, and generating a communication early warning instruction to remind a manager to process as soon as possible.
2. The deep learning-based meter identification system of claim 1, wherein the yolov3 test model is constructed as follows:
acquiring a meter image, and constructing a data set required by training; the concrete expression is as follows: collecting n meter pictures under different shooting angles and different lighting conditions, and labeling the meter pictures to obtain a data set required by subsequent training of a yolov3 model; dividing the training set, the test set and the check set according to a set proportion; the set ratio comprises 2:1:1, 3:1:1 and 4:3: 1;
constructing a fusion model: the fusion model is constructed by combining at least two fusion modes of a support vector machine, a deep convolutional neural network and an RBF neural network, wherein the fusion modes comprise a linear weighted fusion method, a cross fusion method, a waterfall fusion method, a feature fusion method and a prediction fusion method;
and (3) carrying out data normalization on the training set, the test set and the check set, then training, testing and checking the fusion model, and marking the trained fusion model as a yolov3 detection model.
3. The deep learning-based meter identification system according to claim 1, wherein the parameter compensation module specifically modifies the steps of:
SS 1: acquiring an identification record of an image analysis module at the current moment, and inputting meter reading obtained by current identification, actual meter reading and corresponding various environmental parameter values into a parameter compensation model to obtain an attribute parameter compensation coefficient;
SS 2: and correcting the yolov3 detection model according to the attribute parameter compensation coefficient.
4. The deep learning-based meter identification system according to claim 1, wherein the signal verification module specifically verifies that the meter identification system is as follows:
the signal verification module sends a verification configuration message to an FPGA main control of the image analysis module according to a preset verification period, wherein the verification configuration message comprises a first signal quality threshold;
responding to the received verification configuration message, and sending a second synchronous signal to the signal verification module by the FPGA main control; the signal quality of the second synchronous signal is determined by the signal verification module and is compared with the first signal quality threshold, and a corresponding quality difference value Z1 is obtained; setting response time length as XT;
calculating a signal loss index SH by using a formula SH (Z1 × a1+ XT × a 2), wherein a1 and a2 are coefficient factors; the interference coefficient Cy is evaluated according to the trend of the signal loss index SH.
5. The deep learning-based meter identification system according to claim 4, wherein the interference coefficient Cy is evaluated as follows:
establishing a curve graph of the change of the signal loss index SH along with time, and comparing the signal loss index SH with a loss threshold value; if SH is larger than or equal to the loss threshold, intercepting a corresponding curve segment in a corresponding curve graph, marking the curve segment as yellow and marking the curve segment as an interference curve segment;
counting the number of the interference curve segments to be L1 in a preset time period, and integrating all the interference curve segments with time to obtain interference reference energy L2; the interference coefficient Cy is calculated by using Cy-L1 × a3+ L2 × a4, where a3 and a4 are coefficient factors.
6. The deep learning-based meter recognition system of claim 1, wherein the recognition records comprise meter readings obtained by the image analysis module at each recognition, actual meter readings and corresponding environmental parameter values; the environmental parameter values comprise temperature, humidity, wind pressure, wind speed and interference signals.
7. The meter recognition system based on deep learning of claim 1, wherein the image analysis module enters an active standby mode after detecting a communication early warning instruction, that is, the meter image is no longer recognized by the FPGA master control of abnormal communication; and after the signal verification module judges that the communication state is normal, continuing the communication between the signal verification module and the communication module.
8. A meter recognition method based on deep learning, which is applied to the meter recognition system based on deep learning according to any one of claims 1 to 7, and is characterized by comprising the following steps:
the method comprises the following steps: a yolov3 detection model which can train and identify images is constructed through a model construction module, and the constructed yolov3 detection model is transmitted to an image analysis module;
step two: performing digital target recognition on the meter image acquired by the image acquisition module by using a yolov3 detection model through an image analysis module to obtain a digital recognition result, and finally obtaining meter reading;
step three: recording the identification records of the image analysis module through the information recording module, and sorting the identification records by calling the information sorting module to construct a parameter detection training sample; training a machine learning-based method to obtain a parameter compensation model;
step four: feeding back the parameter compensation model to the model construction module through the parameter compensation module so as to correct the yolov3 detection model;
step five: verifying the communication state of the image analysis module in real time through a signal verification module, and calculating to obtain an interference coefficient Cy; if the Cy is larger than or equal to the interference threshold, judging that the signal interference is serious and the communication state is abnormal, and generating a communication early warning instruction to remind a manager to process as soon as possible;
step six: and after the image analysis module detects a communication early warning instruction, the image analysis module enters an active standby mode, and after the signal verification module judges that the communication state is normal, the communication between the image analysis module and the signal verification module is continued.
CN202210261718.0A 2022-03-16 2022-03-16 Meter identification method and system based on deep learning Withdrawn CN114821025A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210261718.0A CN114821025A (en) 2022-03-16 2022-03-16 Meter identification method and system based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210261718.0A CN114821025A (en) 2022-03-16 2022-03-16 Meter identification method and system based on deep learning

Publications (1)

Publication Number Publication Date
CN114821025A true CN114821025A (en) 2022-07-29

Family

ID=82528368

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210261718.0A Withdrawn CN114821025A (en) 2022-03-16 2022-03-16 Meter identification method and system based on deep learning

Country Status (1)

Country Link
CN (1) CN114821025A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115587978A (en) * 2022-10-08 2023-01-10 盐城工学院 On-line floor leather laminating embossing detection system based on deep learning

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115587978A (en) * 2022-10-08 2023-01-10 盐城工学院 On-line floor leather laminating embossing detection system based on deep learning
CN115587978B (en) * 2022-10-08 2023-04-21 盐城工学院 Floor leather laminating embossing on-line measuring system based on degree of depth study

Similar Documents

Publication Publication Date Title
CN111798407A (en) Electrified equipment fault diagnosis method based on neural network model
CN103399248A (en) PCB (printed circuit board) defect mass detection system and method
CN117269731B (en) PCBA performance automatic test system based on Internet of things
CN115015286B (en) Chip detection method and system based on machine vision
CN116054417B (en) Monitoring system and method for transformer substation
CN115171361B (en) Dangerous behavior intelligent detection and early warning method based on computer vision
CN111127445A (en) Distribution network line high-temperature area detection method and system based on deep learning
CN114460439A (en) Digital integrated circuit test system
CN114821025A (en) Meter identification method and system based on deep learning
CN116152202A (en) Equipment appearance detection system based on image recognition technology and infrared thermal imaging technology
CN117698288A (en) Printing quality detection method, system, equipment and storage medium
CN116634487A (en) Real-time detection control system of 5G communication base station
CN114608741A (en) Pressure sensor acquisition system based on big data
CN117190887B (en) Aerogel thickness automatic detection system for mobile phone production
CN117314829A (en) Industrial part quality inspection method and system based on computer vision
CN116758060B (en) Vertical basket of flowers visual detection system of battery piece
CN117853295A (en) Safety environmental protection emergency system based on industry interconnection and digital panorama
CN115700644A (en) Vision-based automatic inspection method and device for wind power station
US20210108618A1 (en) System and method for determining an operating condition of a wind turbine
KR102660626B1 (en) System and method for vision inspection based on AI that can be verified on site
CN118376916B (en) Automatic-based daily inspection system and method for enterprise equipment
CN117746345B (en) Material state monitoring system and method based on big data
CN118075511B (en) Intelligent camera monitoring system based on Internet of things
CN114418941B (en) Defect diagnosis method and system based on detection data of power inspection equipment
CN112504137B (en) Multi-target digital image detection method based on cloud computing

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
WW01 Invention patent application withdrawn after publication

Application publication date: 20220729