CN110688900A - Withdrawal meter management method based on image recognition - Google Patents
Withdrawal meter management method based on image recognition Download PDFInfo
- Publication number
- CN110688900A CN110688900A CN201910793065.9A CN201910793065A CN110688900A CN 110688900 A CN110688900 A CN 110688900A CN 201910793065 A CN201910793065 A CN 201910793065A CN 110688900 A CN110688900 A CN 110688900A
- Authority
- CN
- China
- Prior art keywords
- image
- identification
- bar code
- meter
- network
- 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
- 238000007726 management method Methods 0.000 title claims abstract description 18
- 238000010606 normalization Methods 0.000 claims abstract description 17
- 238000005516 engineering process Methods 0.000 claims abstract description 4
- 238000000034 method Methods 0.000 claims description 45
- 238000013528 artificial neural network Methods 0.000 claims description 18
- 238000001514 detection method Methods 0.000 claims description 14
- 238000011176 pooling Methods 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 12
- 238000001914 filtration Methods 0.000 claims description 9
- 238000010586 diagram Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 238000013527 convolutional neural network Methods 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000003491 array Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 238000005286 illumination Methods 0.000 claims description 3
- 230000005764 inhibitory process Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
-
- 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/103—Workflow collaboration or project management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/02—Recognising information on displays, dials, clocks
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Data Mining & Analysis (AREA)
- Entrepreneurship & Innovation (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Operations Research (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Water Supply & Treatment (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Character Discrimination (AREA)
Abstract
The invention discloses a meter withdrawal management method based on image recognition, which comprises the following steps: step one, collecting images of withdrawn electric energy meters; secondly, performing phenotype identification, bar code identification and dial plate number indication identification on the image in sequence by adopting an image identification technology; the phenotype identification is to input the collected withdrawn electric energy meter image, identify and return the model information of the withdrawn electric energy meter by building an electric energy meter type classification network; the bar code identification is to detect the area of the bar code in the image according to the characteristics of the bar code and then identify the bar code according to the coding rule and the normalization theory; the dial plate number indication identification is to segment the numbers after positioning the digital area and then to make a data set for identification; and step three, generating and recording electric meter data. The invention provides a meter withdrawal management method based on image recognition, which ensures the accuracy of electric energy meter information withdrawal and reduces the situations of wrong manual entry of meter number and incorrect balance electric quantity.
Description
Technical Field
The invention relates to the field of detection of electric power and electric energy meters, in particular to a meter withdrawal management method based on image recognition.
Background
The power marketing is a core business of a power supply enterprise, the quality of the power marketing work is related to the survival and the development of the company, and the market competitiveness of the company is determined. With the further acceleration of the reform of the power enterprises, how to adapt to market economy by using high and new technological means, how to improve efficiency, reduce cost and realize high-efficiency and high-quality services becomes an important task for realizing the modernization of power utilization marketing.
With the continuous popularization of the intelligent electric meters, the types of the meter withdrawing meters are more and more, and the entry work of the number of the withdrawn meters is more and more. The traditional manual input method cannot meet the existing service requirements.
The traditional recording work of the bottom number of the meter withdrawal meter has more defects:
firstly, manual entry is easy to make mistakes. In the process of changing the service of the client metering device, the filling of the withdrawn electric energy meter is manually input by an operator, so that errors or mistakes during the input are easy to occur;
secondly, the input information is inaccurate, which causes the loss of the company. The recorded meter number is inconsistent with the number of withdrawn electric energy meters, so that the accuracy of meter withdrawal information cannot be ensured, the balance electric quantity is inaccurate, and unnecessary loss is brought to a power supply company.
Disclosure of Invention
The invention provides a meter withdrawal management method based on image recognition to solve the technical defects at present, so that the accuracy of electric energy meter information withdrawal is ensured, and the situations of wrong manual entry of the number of meters and incorrect balance electric quantity are reduced.
Another object of the invention is to correct the last entered data, improving the accuracy of the revocation list information.
The technical scheme provided by the invention is as follows: a revocation list management method based on image recognition comprises the following steps:
step one, collecting images of withdrawn electric energy meters;
secondly, performing phenotype identification, bar code identification and dial plate number indication identification on the image in sequence by adopting an image identification technology;
the phenotype identification is to input the collected withdrawn electric energy meter image, identify and return the model information of the withdrawn electric energy meter by building an electric energy meter type classification network;
the bar code identification is to detect the area of the bar code in the image according to the characteristics of the bar code and then identify the bar code according to the coding rule and the normalization theory;
the dial plate number indication identification is to segment the numbers after positioning the digital area and then to make a data set for identification;
and step three, generating and recording electric meter data.
Preferably, the acquisition method of the first step specifically includes:
and placing the withdrawn electric energy meter in a photographing area, and photographing the withdrawn electric energy meter by controlling the digital camera through the computer to form the image.
Preferably, in the second step, the phenotype identification specifically includes:
building a neural network for classifying the types of the electric meters;
marking an ammeter type data set, and training a neural network for classifying the ammeter types;
preprocessing the image to obtain a preprocessed image, wherein the preprocessing comprises image size normalization processing;
inputting the preprocessed image into a trained neural network for classifying the types of the electric meters, extracting image features through a convolution layer and an activation function layer, continuously simplifying the image features through a pooling layer, inputting the simplified image features into a full-link layer for classification, and finally judging and outputting a specific phenotype.
Preferably, the neural network for classifying the types of the electricity meters has a specific structure as follows:
a neural network module taking a convolutional layer and a Relu activation layer as bases;
the basic neural network modules are 8 groups in total, and 1-8 groups are numbered in sequence; the convolution kernel size in each group was 3 x 3;
adding a maxporoling pooling layer behind the 2 nd, 4 th, 6 th and 8 th groups of basic modules; the sizes of the pooling layers are 2, 2 and 4 respectively;
and introducing three full-connection layers after the 8 th group of pooling layers for meter type classification.
Preferably, the method for detecting the region where the barcode is located in the second step includes:
a single-stage convolutional neural network method, a Fastser-RCNN method, or a Corner Net method.
Preferably, the single-stage convolutional neural network method specifically includes:
step a, inputting the image into a network, and extracting an image characteristic diagram through a convolution layer, a batch normalization layer, a Leaky Relu layer combined module and a residual error module used by a ResNet network to obtain a characteristic diagram of an input electric energy meter image;
b, carrying out target detection on the multi-scale characteristic images obtained at different stages by adopting an FPN method, and distinguishing a bar code region and a background region;
and c, filtering the detection result obtained by multiple scales by adopting a non-maximum value inhibition method to obtain the area where the bar code is located, wherein the filtering threshold value is 0.45.
Preferably, the barcode identification comprises:
according to the area of the detected bar code, performing binarization processing on the area of the bar code by using a binarization method adaptive to a local area to obtain a strip-shaped binarization image;
scanning the binary image line by line, and judging the line characteristics conforming to the bar code;
and calculating the average value of the space widths of the bar codes according to all the arrays meeting the identification processing conditions, and identifying the corresponding bar code characters according to the bar code coding rule and normalization.
Preferably, the dial reading identification includes:
s1, positioning the digital area of the image by adopting a single-stage convolution neural network method;
step S2, segmenting the digital area, and carrying out binarization processing to obtain a single digital area;
and step S3, performing size normalization operation on the single digital image, and inputting the single digital image into the trained basic network to obtain a final recognition result.
It is preferable that the first and second liquid crystal layers are formed of,
the method for constructing the trained basic network comprises the following steps:
different data sets are manufactured according to different phenotypes, a handwritten font recognition network is used as a basic network, a convolution layer of the handwritten font recognition network is replaced by a residual error module used in ResNet to enhance the network performance, and then the basic network is trained by the data sets to obtain a trained basic network.
Preferably, the method for creating the data set comprises:
intercepting complete number display images of a plurality of groups of 0-9 real character wheel electric meters;
counting the display number intervals of the same character wheel in the real character wheel ammeter;
making a plurality of groups of sequenced 0-9 long digital images according to digital intervals;
and randomly intercepting and adding Gaussian blur, and obtaining half-word and whole-word character wheel ammeter data according to Gaussian noise and the change of illumination conditions of the HSV color space of the image.
Preferably, in the third step, the recorded electric meter data needs to be corrected, and the correction coefficient λ satisfies:
wherein S is the area of the bar code region, N is the number of whole words, N is the number of half words, kappa is the filtering threshold, fxIs a one-dimensional gaussian function.
The invention has the following beneficial effects:
1) the efficiency is improved and the loss is reduced. The situations of wrong manual meter number entry and incorrect balance electric quantity can be reduced, the accuracy of withdrawing the information of the electric energy meter is ensured, and the loss of electric quantity and electric charge in meter withdrawing service processing is reduced; the number of withdrawn tables in the table library is automatically filled, and the image identification information replaces manual input, so that the workload of front-line personnel is reduced, and the burden is reduced and the efficiency is improved;
2) standardize the business process and promote the internal management. The method is beneficial to standardizing the electric energy meter withdrawal service processing flow and enhancing supervision and management of the electric energy meter withdrawal service.
3) And correcting the last input data to improve the information accuracy of the withdrawn list.
Drawings
FIG. 1 is a flowchart of revocation list management based on image recognition according to the present invention.
FIG. 2 is a flow chart of the image recognition system according to the present invention.
FIG. 3 is a flow chart of the detection method of YOLO V3 tiny according to the present invention.
Detailed Description
The present invention is further described in detail below with reference to the attached drawings so that those skilled in the art can implement the invention by referring to the description text.
As shown in fig. 1, the present invention provides a revocation list management method based on image recognition, which mainly includes: and withdrawing the table image identification content, withdrawing the table image identification operation process, withdrawing the table image identification information file, withdrawing the table image identification information query, and checking the table image identification effect.
The digital camera can be controlled to take photos by withdrawing the list for photographing and identifying the operation, and the on-off and photographing of the camera are directly controlled by operating software without manual adjustment. The power supply device is added in the operation table to provide power supply for the electronic meter, display electronic display number, and set infrared reading, and can read the number information of the electronic meter, and immediately identify the model, the bar code and the number in the image after the meter image is withdrawn.
And the revocation list information management function can arrange the identified revocation list information to form a revocation list information file, count the identification number, the identification rate and the identification accuracy of the revocation list, compare and analyze the revocation list information with marketing data, and finish the arrangement, comparison, statistics and analysis of the image identification information.
And the meter withdrawing image identification content mainly comprises phenotype identification, bar code identification and expression number identification. Phenotype identification, namely identifying the model of the electric meter; the method comprises the following steps of bar code identification, namely determining the position of a bar code for a given electric meter image and identifying the number represented by the bar code; and (4) dial indication identification, namely determining the indication area position of the ammeter image, cutting out a single digital image according to the phenotype, extracting features and identifying (including whole character, shielding and half character identification).
And the withdrawn form information query mainly comprises picture information (a front view, a side view and the like), identification information (an electric meter model, a bar code, a representation number and the like), marketing customer information (a customer name, an address and the like), and withdrawn information (withdrawn time, identification time, withdrawn personnel and the like).
And counting the withdrawn meter quantity, which mainly comprises counting the quantity of the withdrawn electric energy meters of each power supply branch company in each month, the type of the electric energy meters, operators, the withdrawn months, the identification rate, the identification success rate and the like. The withdrawn form identification data is compared and analyzed with the form number manually input in the marketing management information system, and the place where the withdrawn form data is different from the manually input data is analyzed, so that supervision of the withdrawn form number is realized.
Picture recognition revocation list system: the system is realized in a mode that the withdrawn electric energy meter is placed in a photographing area in a fixed operation platform, and the digital camera is controlled by a computer to photograph the withdrawn electric energy meter to form an image. The method is applied to a visual image recognition technology, and can automatically recognize information such as electric energy meter models, bar codes, display numbers and the like in images.
And (3) picture identification function design: the image recognition design is to identify the watch shape, the bar code identification and the dial indication number from the withdrawn electric meter image. The method mainly comprises phenotype identification, bar code identification and representation number identification. The flow chart of the image recognition process is shown in fig. 1:
(1) and (4) phenotype identification. Identifying the model of the electric meter;
(2) and (5) identifying the bar code. Determining the position of a bar code in the electric meter image and identifying the number represented by the bar code;
(3) and identifying dial plate readings. The position of the indication area of the electric meter image is determined, a single digital image is cut out according to the phenotype, the characteristics are extracted, and the identification (including whole character, shading and half character identification) is carried out by utilizing a digital identification neural network.
Phenotype identification protocol: and building an electric meter type classification network, inputting the shot withdrawn electric energy meter image, and identifying and returning the model information of the withdrawn electric energy meter. The specific flow is as follows.
(1) The method comprises the following steps of building a neural network for classifying the types of the electric meters, wherein the structure of the neural network is briefly described as follows:
p1, adopting a convolution layer and a Relu activation layer as a basic neural network module;
p2, using 8 groups of basic modules, and numbering the 1 st to 8 th groups of the basic modules; the convolution kernel size in each group was 3 x 3;
p3, simultaneously adding a maxpoling pooling layer behind the 2 nd, 4 th, 6 th and 8 th groups of basic modules; the pooling sizes are 2, 2 and 4 respectively;
and P4, introducing three full-connection layers after the last group of pooling layers to classify the types of the electric meters.
(2) And marking a reasonable number of ammeter type data sets, and training the constructed network.
(3) And (5) image preprocessing. Including image size normalization and the like.
(4) The preprocessed image is input into a successfully trained network, efficient image features are extracted through a convolution layer and an activation function layer, the features are simplified continuously through a pooling layer, and finally the simplified image features are input into a full-connection layer for classification. And judging and outputting a specific phenotype.
Barcode identification scheme: and searching the area where the bar code is located in the image of the withdrawal list according to the characteristics of the bar code, and then identifying the bar code according to the coding rule and the normalization theory.
(1) Preprocessing collected color images such as size normalization;
(2) and searching the area of the bar code. In the invention, a single-stage convolutional neural network YOLO V3 tiny (you Only Look one) method with high speed and good concurrency is adopted to detect the area where the bar code in the electric meter image is located; and identifying the detected bar code area; it is understood that other methods of object detection may be used to determine barcode regions, such as Fastser-RCNN, Corner Net, etc.
(3) Referring to fig. 3, the method for detecting the state of the barcode area in the electricity meter image by using the YOLO V3 tiny method includes:
s1, inputting the collected image into a network, extracting the image characteristic diagram through a DBL module, namely a combined module of the convolution layer, a batch normalization layer and a Leaky Relu layer, and a residual error module used by a ResNet network to obtain the characteristic diagram (feature map) of the input ammeter image
S2, adopting FPN (feature pyramid) idea to detect the target of the multi-scale characteristic diagram obtained at different stages and distinguish the bar code area and the background area; the method can adapt to the problem of large scale difference of the shot electric meter images, and can obtain a better detection result for the electric meter images with any shot size.
And S3, finally, filtering the detection result obtained by the multi-scale by adopting NMS Non-maximum inhibition (Non-maximum-update), wherein the threshold value of the filtering is 0.45.
(4) And obtaining the bar code area according to the detection. Carrying out binarization on the bar code region by using a local region adaptive binarization method; scanning the bar-shaped binary image line by line, and judging whether the bar-shaped binary image conforms to the characteristics of the bar code line;
(5) and (5) identifying the bar code. And calculating the average value of the space widths of the bar codes according to all the arrays meeting the identification processing conditions, and identifying the corresponding bar code characters according to the bar code coding rule and the normalization theory.
Dial plate registration identification scheme: the method for identifying the electric meter reading information in the withdrawn electric energy meter image mainly comprises the following steps:
(1) precise positioning of the digital area. The part adopts the same detection method as the bar code detection, and the YOLO V3 tiny detection method is used for positioning the number indicating area; efficient concurrency performance can be guaranteed. Similarly, other target detection methods can be used in this part of the invention to determine the digital region, such as Fastser-RCNN, Corner Net, etc.
(2) And after the position of the digital area is determined, cutting off the digital area according to the positions and carrying out binarization.
(3) And (5) dividing the number. Different segmentation methods are adopted according to phenotypes: aiming at the character wheel type electric meter, because the digit display digit is fixed, an accurate single digit region can be directly obtained by adopting a mean value segmentation mode. For the LCD screen type electric meter, the display digit is not fixed. And horizontally projecting the binarized digital area image, counting peak-valley information, and cutting off the parts among valley intervals meeting the width condition to obtain a single digital area.
(4) And (4) data set production, wherein different data sets are marked according to different phenotypes for subsequent recognition network training. In this section, word wheel meters have problems with whole words, occlusion, and half words. Therefore, there is a problem in obtaining more appropriate training data. Therefore, the invention provides a way of generating character wheel electric meter data:
s1, firstly, intercepting and obtaining a complete number display image of a plurality of groups of 0-9 real character wheel electric meters;
s2, counting the number intervals displayed by the same character wheel in the real character wheel ammeter;
and S3, manufacturing a plurality of groups of sequential 0-9 strip digital images according to intervals.
And S4, randomly intercepting and adding Gaussian blur (Gaussian blur), Gaussian noise (Gaussian noise and the change of illumination conditions of the HSV color space of the image) to obtain a large amount of half words and whole word wheel ammeter data.
(5) And (5) constructing a network. A handwriting font recognition network is used as a basic network (backbone), and the convolution layer of the handwriting font recognition network is replaced by a residual error module used in ResNet to enhance the network performance so as to achieve a better recognition effect.
(6) And (5) training the constructed network by adopting the data set in the step (4). When in use, the cut single digital image is input into the trained network after the operations of size normalization and the like are carried out, and the final recognition result is obtained.
And finally, recording the meter data to be corrected, wherein the correction coefficient lambda meets the following conditions:
wherein S is the area of the bar code region in cm2N is the number of whole words, N is the number of half words, k is the filtering threshold, fxIs a one-dimensional Gaussian function and is a Gaussian processing function when Gaussian blur is added.
While embodiments of the invention have been described above, it is not limited to the applications set forth in the description and the embodiments, which are fully applicable in various fields of endeavor to which the invention pertains, and further modifications may readily be made by those skilled in the art, it being understood that the invention is not limited to the details shown and described herein without departing from the general concept defined by the appended claims and their equivalents.
Claims (10)
1. A revocation list management method based on image recognition is characterized by comprising the following steps:
step one, collecting images of withdrawn electric energy meters;
secondly, performing phenotype identification, bar code identification and dial plate number indication identification on the image in sequence by adopting an image identification technology;
the phenotype identification is to input the collected withdrawn electric energy meter image, identify and return the model information of the withdrawn electric energy meter by building an electric energy meter type classification network;
the bar code identification is to detect the area of the bar code in the image according to the characteristics of the bar code and then identify the bar code according to the coding rule and the normalization theory;
the dial plate number indication identification is to segment the numbers after positioning the digital area and then to make a data set for identification;
and step three, generating and recording electric meter data.
2. The method for managing revocation meters based on image recognition according to claim 1, wherein in the second step, the phenotype recognition specifically comprises:
building a neural network for classifying the types of the electric meters;
marking an ammeter type data set, and training a neural network for classifying the ammeter types;
preprocessing the image to obtain a preprocessed image, wherein the preprocessing comprises image size normalization processing;
inputting the preprocessed image into a trained neural network for classifying the types of the electric meters, extracting image features through a convolution layer and an activation function layer, continuously simplifying the image features through a pooling layer, inputting the simplified image features into a full-link layer for classification, and finally judging and outputting a specific phenotype.
3. The withdrawn meter management method based on image recognition as claimed in claim 2, wherein the neural network of the electric meter type classification has a specific structure as follows:
a neural network module taking a convolutional layer and a Relu activation layer as bases;
the basic neural network modules are 8 groups in total, and 1-8 groups are numbered in sequence; the convolution kernel size in each group was 3 x 3;
adding a maxporoling pooling layer behind the 2 nd, 4 th, 6 th and 8 th groups of basic modules; the sizes of the pooling layers are 2, 2 and 4 respectively;
and introducing three full-connection layers after the 8 th group of pooling layers for meter type classification.
4. The method for managing the meter revocation list based on image identification as claimed in claim 2, wherein the method for detecting the area where the barcode is located in the second step comprises:
a single-stage convolutional neural network method, a Fastser-RCNN method, or a Corner Net method.
5. The image recognition-based revocation list management method of claim 4, wherein the single-stage convolutional neural network method specifically comprises:
step a, inputting the image into a network, and extracting an image characteristic diagram through a convolution layer, a batch normalization layer, a Leaky Relu layer combined module and a residual error module used by a ResNet network to obtain a characteristic diagram of an input electric energy meter image;
b, carrying out target detection on the multi-scale characteristic images obtained at different stages by adopting an FPN method, and distinguishing a bar code region and a background region;
c, filtering the detection result obtained by multiple scales by adopting a non-maximum inhibition method to obtain the region where the bar code is located;
wherein the filtering threshold is 0.45.
6. The method of claim 5, wherein the barcode recognition comprises:
according to the area of the detected bar code, performing binarization processing on the area of the bar code by using a binarization method adaptive to a local area to obtain a strip-shaped binarization image;
scanning the binary image line by line, and judging the line characteristics conforming to the bar code;
and calculating the average value of the space widths of the bar codes according to all the arrays meeting the identification processing conditions, and identifying the corresponding bar code characters according to the bar code coding rule and normalization.
7. The image recognition-based meter revocation management method of claim 1, wherein said dial-reading recognition comprises:
s1, positioning the digital area of the image by adopting a single-stage convolution neural network method;
step S2, segmenting the digital area, and carrying out binarization processing to obtain a single digital area;
and step S3, performing size normalization operation on the single digital image, and inputting the single digital image into the trained basic network to obtain a final recognition result.
8. The revocation list management method based on image recognition of claim 7,
the method for constructing the trained basic network comprises the following steps:
different data sets are manufactured according to different phenotypes, a handwritten font recognition network is used as a basic network, a convolution layer of the handwritten font recognition network is replaced by a residual error module used in ResNet to enhance the network performance, and then the basic network is trained by the data sets to obtain a trained basic network.
9. The method of claim 8, wherein the data set is created by:
intercepting complete number display images of a plurality of groups of 0-9 real character wheel electric meters;
counting the display number intervals of the same character wheel in the real character wheel ammeter;
making a plurality of groups of sequenced 0-9 long digital images according to digital intervals;
and randomly intercepting and adding Gaussian blur, and obtaining half-word and whole-word character wheel ammeter data according to Gaussian noise and the change of illumination conditions of the HSV color space of the image.
10. The method for managing withdrawn meters based on image recognition according to claim 1, wherein in the third step, the recorded meter data needs to be corrected, and the correction coefficient λ satisfies:
wherein S is the area of the bar code region, N is the number of whole words, N is the number of half words, kappa is the filtering threshold, fxIs a one-dimensional gaussian function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910793065.9A CN110688900A (en) | 2019-08-27 | 2019-08-27 | Withdrawal meter management method based on image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910793065.9A CN110688900A (en) | 2019-08-27 | 2019-08-27 | Withdrawal meter management method based on image recognition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110688900A true CN110688900A (en) | 2020-01-14 |
Family
ID=69108434
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910793065.9A Pending CN110688900A (en) | 2019-08-27 | 2019-08-27 | Withdrawal meter management method based on image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110688900A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111695576A (en) * | 2020-06-09 | 2020-09-22 | 国网河北省电力有限公司营销服务中心 | Electric energy meter appearance rapid identification method based on neural network |
| CN112884002A (en) * | 2021-01-18 | 2021-06-01 | 四川大学 | Intelligent electric meter version classification method |
| CN114047471A (en) * | 2021-11-02 | 2022-02-15 | 中国南方电网有限责任公司超高压输电公司贵阳局 | Electric energy meter calibration method and device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016212603A (en) * | 2015-05-07 | 2016-12-15 | 株式会社ケイオーエス | Bar code recognition device and bar code recognition method |
| CN108416412A (en) * | 2018-01-23 | 2018-08-17 | 浙江瀚镪自动化设备股份有限公司 | A kind of logistics compound key recognition methods based on multitask deep learning |
| CN109271987A (en) * | 2018-08-28 | 2019-01-25 | 上海鸢安智能科技有限公司 | A kind of digital electric meter number reading method, device, system, computer equipment and storage medium |
| CN109635806A (en) * | 2018-12-12 | 2019-04-16 | 国网重庆市电力公司信息通信分公司 | Ammeter technique for partitioning based on residual error network |
| CN109657682A (en) * | 2018-11-29 | 2019-04-19 | 国网河北省电力有限公司电力科学研究院 | A kind of electric energy meter registration recognition methods based on deep neural network and the soft cutting of multi-threshold |
-
2019
- 2019-08-27 CN CN201910793065.9A patent/CN110688900A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016212603A (en) * | 2015-05-07 | 2016-12-15 | 株式会社ケイオーエス | Bar code recognition device and bar code recognition method |
| CN108416412A (en) * | 2018-01-23 | 2018-08-17 | 浙江瀚镪自动化设备股份有限公司 | A kind of logistics compound key recognition methods based on multitask deep learning |
| CN109271987A (en) * | 2018-08-28 | 2019-01-25 | 上海鸢安智能科技有限公司 | A kind of digital electric meter number reading method, device, system, computer equipment and storage medium |
| CN109657682A (en) * | 2018-11-29 | 2019-04-19 | 国网河北省电力有限公司电力科学研究院 | A kind of electric energy meter registration recognition methods based on deep neural network and the soft cutting of multi-threshold |
| CN109635806A (en) * | 2018-12-12 | 2019-04-16 | 国网重庆市电力公司信息通信分公司 | Ammeter technique for partitioning based on residual error network |
Non-Patent Citations (2)
| Title |
|---|
| 奔跑的小仙女: "一种使用新的特征金字塔方式的单阶段目标检测器", 《HTTPS://BLOG.CSDN.NET/QQ_43211132/ARTICLE/DETAILS/93040023》 * |
| 张晓颖: "电能表数字图像识别技术研究与实现", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111695576A (en) * | 2020-06-09 | 2020-09-22 | 国网河北省电力有限公司营销服务中心 | Electric energy meter appearance rapid identification method based on neural network |
| CN112884002A (en) * | 2021-01-18 | 2021-06-01 | 四川大学 | Intelligent electric meter version classification method |
| CN114047471A (en) * | 2021-11-02 | 2022-02-15 | 中国南方电网有限责任公司超高压输电公司贵阳局 | Electric energy meter calibration method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108399405B (en) | Business license identification method and device | |
| CN111476109A (en) | Bill processing method, bill processing apparatus, and computer-readable storage medium | |
| CN110688900A (en) | Withdrawal meter management method based on image recognition | |
| CN111428599B (en) | Bill identification method, device and equipment | |
| CN108717543B (en) | Invoice identification method and device and computer storage medium | |
| CN110874618B (en) | OCR template learning method and device based on small sample, electronic equipment and medium | |
| CN112446351A (en) | Medical bill intelligent recognition system solution | |
| CN108596137A (en) | A kind of commodity scanning record method based on image recognition algorithm | |
| CN109840520A (en) | A kind of invoice key message recognition methods and system | |
| CN111444793A (en) | Bill recognition method, equipment, storage medium and device based on OCR | |
| CN112395996A (en) | Financial bill OCR recognition and image processing method, system and readable storage medium | |
| CN110032654B (en) | Supermarket commodity entry method and system based on artificial intelligence | |
| CN110781890A (en) | Identification card identification method and device, electronic equipment and readable storage medium | |
| CN116343228A (en) | Intelligent reading method and system for water meter | |
| CN111368828A (en) | Multi-bill identification method and device | |
| CN110647824B (en) | Value-added tax invoice layout extraction method based on computer vision technology | |
| CN115471858A (en) | Data processing method and device for bill management | |
| CN114882605A (en) | Intelligent attendance management system and method based on Internet | |
| CN111292068B (en) | Contract information auditing method and device, electronic equipment and storage medium | |
| CN111460293B (en) | Information pushing method and device and computer readable storage medium | |
| Vorozhtsova et al. | System approach to development of intellectual information mobile system for electric power metering | |
| CN116311299A (en) | Method, device and system for identifying structured data of table | |
| CN115760808A (en) | Method, system and device for measuring size of plate glass and readable storage medium | |
| CN112699824B (en) | Method and device for detecting constant of electric energy meter and storage medium | |
| CN114240924A (en) | Power grid equipment quality evaluation method based on digitization technology |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200114 |
|
| RJ01 | Rejection of invention patent application after publication |