CN110490195A - A kind of water meter dial plate Recognition of Reading method - Google Patents
A kind of water meter dial plate Recognition of Reading method Download PDFInfo
- Publication number
- CN110490195A CN110490195A CN201910725909.6A CN201910725909A CN110490195A CN 110490195 A CN110490195 A CN 110490195A CN 201910725909 A CN201910725909 A CN 201910725909A CN 110490195 A CN110490195 A CN 110490195A
- Authority
- CN
- China
- Prior art keywords
- model
- image
- water meter
- dial plate
- trained
- 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
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/63—Scene text, e.g. street names
-
- 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
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/158—Segmentation of character regions using character size, text spacings or pitch estimation
-
- 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
-
- 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
- G06V30/10—Character recognition
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Character Discrimination (AREA)
Abstract
The present invention relates to a kind of water meter dial plate Recognition of Reading methods, what is solved is that long-range meter reading recognition accuracy is low, device resource requirement is high, needs the technical issues of high-speed network, by using step 1, calling mobile phone camera takes pictures to water meter dial reading, edge detection photo obtains reading area profile, it carries out single Character segmentation and is marked after converting character picture to the Pixel Dimensions of standard, obtain sample data set；Step 2 constructs the neural network MoblieNetV2 model of deep learning, is standardized rear input model to sample data set and is trained, in the APP of trained model transplantations to the end Android；Step 3, water meter dial plate photo to be detected is executed into step 1 and obtains the single character picture of meter reading, the technical solution for trained model will be inputted after character picture standardization predicting, preferably resolves the problem, can be actually used in long-range meter reading identification.
Description
Technical field
The present invention relates to water meter dial plate Recognition of Reading field, and in particular to a kind of based on MoblieNetV2 model
The end android water meter dial plate Recognition of Reading method.
Background technique
With the fast development of information technology and artificial intelligence, many fields all start to have turned to intelligent development.Wisdom
City, intelligent electric power, the concepts such as wisdom water utilities are proposed in succession.Although intellectual water meter has obtained certain development in recent years,
Be intellectual water meter popularity rate it is relatively low, and safeguard and management it is costly.Therefore, most of place still uses
Be traditional mechanical water meter.Automatic knowledge in order to solve the problems, such as traditional manual metering low efficiency, to tradition machinery water meter
It is not still a focus on research direction.Currently, having been achieved for many research achievements to the image recognition of tradition machinery water meter.
Such as: the methods of template matching method, depth convolutional neural networks method of identification, Tesseract identification.
For the water meter dial plate Recognition of Reading technique study based on template matching, the process employs traditional template matchings
Method identifies character, although higher to the half-word accuracy of identification being likely to occur in identification, due to the office of template matching method
It is sex-limited, so that the accuracy of meter reading identification is not very high；For Tesseract recognition methods, need using Tesseract
Optical character recognition engine identifies water meter character picture, although can substantially be identified to water meter character picture,
But it is lower for the discrimination of half of character；Although and existing depth convolutional neural networks method of identification recognition accuracy compared with
Height, but existing deep neural network is higher to running equipment resource requirement, is deployed in embedded device or mobile device
It is relatively difficult, water meter image is usually uploaded into cloud or background server identifies, to network transmission speed requirement
It is higher.
For the deficiency of existing mechanical water meter reading image recognition technology, the present invention is based on using one kind
The end the Android water meter image digitization of MoblieNetV2 model knows method for distinguishing, and MoblieNetV2 model transplantations are arrived in realization
The end Android accurately identifies water meter dial reading.
Summary of the invention
The technical problem to be solved by the present invention is to accuracy present in existing long-range meter reading identification technology is low, right
Running equipment resource requirement is higher, needs the problem of high-speed network transmission.A kind of new water meter dial plate Recognition of Reading side is provided
Method, the water meter dial plate Recognition of Reading method have accuracy rate high, but require not running equipment resource and network transmission speed
High feature.
In order to solve the above technical problems, the technical solution adopted is as follows:
A kind of water meter dial plate Recognition of Reading method, the water meter dial plate Recognition of Reading method are based on MoblieNetV2 mould
The end the android water meter dial plate Recognition of Reading of type, method include:
Step 1 first passes through calling mobile phone camera and takes pictures to water meter dial plate, and obtained photo is carried out edge inspection
It surveys, is partitioned into the image in meter reading region, then obtained meter reading area image is subjected to single Character segmentation, and will be single
A character picture is marked after being converted into the Pixel Dimensions of standard, obtains the sample data set for being used for model training；
Step 2 constructs the neural network MoblieNetV2 model of deep learning, and the sample data set of step 1 is carried out
It inputs neural network MoblieNetV2 model after standardization to be trained, by trained neural network MoblieNetV2
In model transplantations to the APP at the end Android；
Step 3, the interface function for calling JavaAPI to support, using inference interface function to the nerve of step 2
Network MoblieNetV2 model is loaded, and is completed model and is called, water meter dial plate photo to be detected is used to the side of step 1
The single character picture for the meter reading that formula obtains, and it is trained by being inputted after character picture standardization
MoblieNetV2 model is predicted, realizes the identification of the meter reading based on the end deep learning Android.
Wherein, Inference interface function be in JavaAPI interface one of interface function, it can be achieved that in mobile phone
End carries out load importing to model, and not needing to network online can realize and load to model, because JavaAPI interface is
Local interface at Android phone end.
In the above problem, for optimization, further, step 1 is specifically included:
Water meter photo is subjected to gray processing processing, obtains the morphological feature of image；
Using gaussian filtering, to gray processing, treated that image is smoothed, and eliminates various interference noises；
Binary conversion treatment is carried out, so that profile is more clear in image, using the method for edge detection to water meter dial plate figure
As carrying out profile lookup, the reading area of original water gauge image is split according to profile coordinate position is searched, after obtaining segmentation
Meter reading area image；
Equal proportion segmentation is carried out to the meter reading area image after segmentation again, obtains the image of single character；Then will
Each character picture is converted into the Pixel Dimensions of standard, and each character picture is marked, defined label interval, model
It encloses, obtains sample data set.
Further, neural network MoblieNetV2 model is defined as depth and separates convolutional network, the separable volume of depth
Product network includes depthwise convolution sum pointwise convolution；
Wherein, the convolution kernel that a 3*3 is used alone to each input channel in depthwise convolution carries out convolution behaviour
Make, pointwise convolution carries out convolution operation using 1*1 convolution kernel, for changing the number in channel.
Further, step 2 includes defining k number to be used as a batch according to collection, and same batch carries out batch operation, directly
To the complete all data sets of training:
Step 2.1, the sample data set of each batch is standardized, makes the mean value 0 of picture, variance is
1, the formula of standardization are as follows:
Wherein, μ indicates the mean value of image, and x indicates that image array, N indicate that image x pixel quantity, σ indicate standard variance, k
For positive integer；
Step 2.2, the image data set of each batch standardization is input to neural network MoblieNetV2 model
It is trained in the middle, it is a number greater than 0 and being less than or equal between 0.1 that initial learning rate, which is arranged, is iterated training simultaneously
The adjust automatically learning rate in the value range of 0-0.1, until accuracy reaches preset accuracy index or more, expression training
It completes, and saves trained model file；
Step 2.3, trained neural network MoblieNetV2 model is switched into the model file that format is .tflite,
Then .tflite model file is transplanted in the APP at the end Android, so that APP is called.
Beneficial effects of the present invention: the present invention is a kind of end Android meter reading knowledge based on MoblieNetV2 model
Not, it compared with existing meter reading identification technology, realizes and uses the MoblieNetV2 mould of deep learning at the end Android
Type carries out meter reading identification, and identification accurate rate is high.In addition, convolution of the used MoblieNetV2 network model than standard
Neural network model calculation amount reduces 8 to 9 times, further reduced requirement of the network model to mobile device operational capability.Together
When overcome and need to upload to water meter image cloud or background server reuses the network model of deep learning and knows
Otherwise, realize that being just able to use deep learning model at the end Android accurately identifies water meter number.
Detailed description of the invention
Present invention will be further explained below with reference to the attached drawings and examples.
Fig. 1, the end the Android water meter dial plate Recognition of Reading method flow based on MoblieNetV2 model in embodiment 1
Figure.
Fig. 2, the inverted_bottleneck layer structure chart in embodiment 1.
Fig. 3, activation primitive relu6 pictorial diagram.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments to the present invention
It is further elaborated.It should be appreciated that described herein, specific examples are only used to explain the present invention, is not used to limit
The fixed present invention.
Embodiment 1
The end the Android water meter dial plate Recognition of Reading method based on MoblieNetV2 model that the present embodiment provides a kind of:
Step 1 carries out the reading area of water meter image using the method for gray processing, gaussian filtering and edge detection
Segmentation, and equal proportion segmentation is carried out to the meter reading area image after segmentation, single character picture is obtained, then by each
Character picture switchs to the Pixel Dimensions of standard；
Step 2 is marked each character sample image that step 1 obtains, marker spacing 0.5, range from 0 to
9.5, totally 20 groups of sample data sets；
Step 3 uses at standardized processing mode the character image data of 20 groups of obtained sample data sets
Reason；
Step 4 builds the neural network MoblieNetV2 model of deep learning, initial learning rate is set greater than 0
And a number being less than or equal between 0.1, treated that 20 groups of sample data sets are input to MoblieNetV2 to standardized
It is trained in model, and learning rate is adjusted automatically according to training result, adjusting range is between 0-0.1；
Step 5 passes through the interface of JavaAPI support in the APP application of trained model transplantations to Android
Function loads trained model using inference interface function；
Step 6 obtains water to be identified in such a way that APP calling mobile phone camera is taken pictures or opens mobile phone photo album
Table dial reading image, and use the method for gray processing, gaussian filtering and edge detection to water water meter dial reading image
Meter reading region is split, and then carries out equal proportion segmentation to meter reading area image, obtains single character picture, then will
Single character picture switchs to the Pixel Dimensions of standard and is input in trained model one by one after being standardized
It is identified, last position in the result of identification is retained into its recognition result, remaining removes fractional part by the way of being rounded
Point, to obtain more accurate recognition result.
(1) it obtains the sample data set for being used for model training: water meter photo being subjected to gray processing processing, obtains the shape of image
State feature；Using gaussian filtering, to gray processing, treated that image is smoothed, and eliminates various interference noises；Again by image
Binary conversion treatment is carried out to take turns water meter dial plate image using the method for edge detection so that profile is more clear in image
Exterior feature is searched, and is split according to profile coordinate position is searched to the reading area of original water gauge image；To the meter reading after segmentation
Area image carries out equal proportion segmentation again, obtains the image of single character；Then all by the Pixel Dimensions of each character picture
It is converted into 224*224, and each character picture is marked.Marker spacing is 0.5, and range is from 0 to 9.5, totally 20 groups of samples
Notebook data collection.
(2) build the neural network MoblieNetV2 model of deep learning: MobileNetV2 network model uses
Depth separates convolutional network, and it includes depthwise convolution sum pointwise convolution that depth, which separates convolutional network,.Wherein,
The convolution kernel that 3*3 is used alone to each input channel in depthwise convolution carries out convolution operation, and pointwise volumes
Product carries out convolution operation using 1*1 convolution kernel, for changing the number in channel.MoblieNetV2 model structure such as table 1:
Table 1
| Input | Operation | Extension ratio | Output channel | Number of repetition | Convolution step-length |
| 224*224*3 | conv2d(3*3) | - | 32 | 1 | 2 |
| 112*112*32 | inverted_bottleneck | 1 | 16 | 1 | 1 |
| 112*112*16 | inverted_bottleneck | 6 | 24 | 2 | 2 |
| 56*56*24 | inverted_bottleneck | 6 | 32 | 3 | 2 |
| 28*28*32 | inverted_bottleneck | 6 | 64 | 4 | 2 |
| 14*14*64 | inverted_bottleneck | 6 | 96 | 3 | 1 |
| 14*14*96 | inverted_bottleneck | 6 | 160 | 3 | 2 |
| 7*7*160 | inverted_bottleneck | 6 | 320 | 1 | 1 |
| 7*7*320 | conv2d(1*1) | - | 1280 | 1 | 1 |
| 7*7*1280 | avg_pool2d | - | - | 1 | - |
| 1*1*1280 | conv2d(1*1) | - | 20 | - |
Wherein, conv2d (3*3), conv2d (1*1) indicate that convolution kernel is respectively the convolution operation of 1*1 and 3*3；avg_
Pool2d indicates average pondization operation；Extension ratio explanation: if input is 56*56*24, then input channel number is 24, corresponding
Extension ratio be 6, therefore, the port number of extension is 24*6=144；For number of repetition, when number of repetition is greater than 1, the
Using the convolution step-length in table when primary repetition, remaining convolution step-length is 1.
Inverted_bottleneck layers of structure such as Fig. 2, in inverted_bottleneck layers:
Convolution operation is carried out using the con2d that convolution kernel is 1*1 to the input of input, port number is extended to (ratio
Example * input channel number), then activated by activation primitive relu6, corresponding output channel can be divided using separable_conv2d
From convolution operation, i.e., the convolution kernel for a 3*3 being used alone to each channel carries out convolution algorithm, and obtained result is again with activation
Function relu6 activation；Then convolution operation is carried out to output result using the con2d that convolution kernel is 1*1, by port number boil down to
Corresponding output channel number.Wherein, activation primitive relu6 figure is as shown in figure 3, value interval uses activation between [0,6]
Function relu6 can retain more features information.
20 groups of sample data sets are input in MoblieNetV2 model and are trained, using every 64 data sets as one
A batch carries out batch training, until having trained all data sets, comprising:
The sample data set of each batch is standardized, the mean value 0 of picture, variance 1 are made.Standardization
The formula of processing is as follows:
Wherein, μ indicates the mean value of image, and x indicates that image array, N indicate that image x pixel quantity, σ indicate standard variance.
The single character image data collection of every a collection of standardization is input in model and is trained, setting is initial
Learning rate is a number greater than 0 and being less than or equal between 0.1, is then iterated training and in the value range of 0-0.1
Interior adjust automatically learning rate, until training accuracy reaches 96% or more, expression training is completed, and saves trained model text
Part.
Trained model is deployed in the APP of the end Android.It needs trained model switching to .tflite mould
Then type file is transplanted in the APP at the end Android, so that APP is called.
The end Android APP stress model file identifies water meter image.
The interface function for calling JavaAPI to support first carries out trained model using inference interface function
Load；
Calling mobile phone camera is taken pictures or is opened mobile phone photo album and imports water meter dial reading image；
After operating to the water meter dial reading image of importing according to the method for obtaining sample set, 5 single words will be obtained
Image is accorded with, then single character picture is standardized, then by the individual digit image data after 5 standardizations
It is input in trained MoblieNetV2 model and is predicted one by one, final realize accurately identifies meter reading.
Although the illustrative specific embodiment of the present invention is described above, in order to the technology of the art
Personnel are it will be appreciated that the present invention, but the present invention is not limited only to the range of specific embodiment, to the common skill of the art
For art personnel, as long as long as various change the attached claims limit and determine spirit and scope of the invention in, one
The innovation and creation using present inventive concept are cut in the column of protection.
Claims (4)
1. a kind of water meter dial plate Recognition of Reading method, it is characterised in that: the water meter dial plate Recognition of Reading method is based on
The end the Android water meter of MoblieNetV2 model, method include:
Step 1 first passes through calling mobile phone camera and takes pictures to water meter dial plate, and photo is carried out edge detection, segmentation water outlet
Meter reading area image, then meter reading area image is subjected to single Character segmentation, and convert mark for single character picture
It is marked after quasi- Pixel Dimensions, obtains the sample data set for being used for model training；
Step 2 constructs the neural network MoblieNetV2 model of deep learning, and the sample data set of step 1 is carried out standard
It inputs in neural network MoblieNetV2 model and is trained after change processing, by trained neural network MoblieNetV2 mould
Type is transplanted in the APP at the end Android；
Step 3, the interface function for calling JavaAPI to support, using inference interface function to the neural network of step 2
MoblieNetV2 model is loaded, and is completed model and is called；Water meter dial plate photo to be detected is obtained by the way of step 1
The single character picture of the meter reading taken, and trained MoblieNetV2 mould will be inputted after character picture standardization
Type is predicted, realizes the identification of the meter reading based on the end deep learning Android.
2. water meter dial plate Recognition of Reading method according to claim 1, it is characterised in that: step 1 specifically includes:
Water meter photo is subjected to gray processing processing, obtains the morphological feature of image；
Using gaussian filtering, to gray processing, treated that image is smoothed, and eliminates various interference noises；
Carry out binary conversion treatment so that profile is more clear in image, using edge detection method to water meter dial plate image into
Row profile is searched, and is split according to profile coordinate position is searched to the reading area of original water gauge image, the water after being divided
Meter reading area image；
Equal proportion segmentation is carried out to the meter reading area image after segmentation again, obtains the image of single character；It then will be each
It opens character picture and is converted into the Pixel Dimensions of standard, and each character picture is marked, defined label interval, range,
Obtain sample data set.
3. water meter dial plate Recognition of Reading method according to claim 2, it is characterised in that: neural network MoblieNetV2
Model is defined as depth and separates convolutional network, and it includes pointwise volumes of depthwise convolution sum that depth, which separates convolutional network,
Product；
Wherein, the convolution kernel that a 3*3 is used alone to each input channel in depthwise convolution carries out convolution operation,
Pointwise convolution carries out convolution operation using 1*1 convolution kernel, for changing the number in channel.
4. water meter dial plate Recognition of Reading method according to claim 3, it is characterised in that: step 2 includes defining k number
It is used as a batch according to collection, same batch carries out batch operation, until having trained all data sets:
Step 2.1, the sample data set of each batch is standardized, makes the mean value 0 of picture, variance 1, mark
The formula of standardization processing are as follows:
Wherein, μ indicates the mean value of image, and x indicates that image array, N indicate that image x pixel quantity, σ indicate that standard variance, k are positive
Integer；
Step 2.2, the image data set of each batch standardization is input in neural network MoblieNetV2 model
It is trained, it is a number greater than 0 and being less than or equal between 0.1 that initial learning rate, which is arranged, is iterated training and in 0-
Adjust automatically learning rate in 0.1 value range, until accuracy reaches preset accuracy index or more, expression has been trained
At, and save trained model file；
Step 2.3, trained neural network MoblieNetV2 model is switched into the model file that format is .tflite, then
.tflite model file is transplanted in the APP at the end Android, so that APP is called.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910725909.6A CN110490195A (en) | 2019-08-07 | 2019-08-07 | A kind of water meter dial plate Recognition of Reading method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910725909.6A CN110490195A (en) | 2019-08-07 | 2019-08-07 | A kind of water meter dial plate Recognition of Reading method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110490195A true CN110490195A (en) | 2019-11-22 |
Family
ID=68550086
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910725909.6A Pending CN110490195A (en) | 2019-08-07 | 2019-08-07 | A kind of water meter dial plate Recognition of Reading method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110490195A (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110956133A (en) * | 2019-11-29 | 2020-04-03 | 上海眼控科技股份有限公司 | Training method of single character text normalization model, text recognition method and device |
| CN111091558A (en) * | 2019-12-12 | 2020-05-01 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon swing bolster spring jumping fault image identification method |
| CN111160129A (en) * | 2019-12-12 | 2020-05-15 | 中国南方电网有限责任公司超高压输电公司广州局 | Intelligent remote meter reading method based on image recognition |
| CN111199194A (en) * | 2019-12-25 | 2020-05-26 | 吉林大学 | Automobile intelligent cabin instrument testing method based on machine vision and deep learning |
| CN111339832A (en) * | 2020-02-03 | 2020-06-26 | 中国人民解放军国防科技大学 | Method and device for detecting face synthetic image |
| CN112113638A (en) * | 2020-09-17 | 2020-12-22 | 北京慧怡物联科技有限责任公司 | Water meter function self-checking device and method |
| CN112215178A (en) * | 2020-10-19 | 2021-01-12 | 南京大学 | Chemical experiment recording system based on pen type interaction |
| CN112434693A (en) * | 2020-11-03 | 2021-03-02 | 辽宁长江智能科技股份有限公司 | Digital water meter identification method and system |
| CN113139541A (en) * | 2021-04-24 | 2021-07-20 | 西安交通大学 | Power distribution cabinet dial nixie tube visual identification method based on deep learning |
| CN113191351A (en) * | 2021-02-01 | 2021-07-30 | 青岛理工大学 | Reading identification method and device of digital electric meter and model training method and device |
| CN113255650A (en) * | 2021-06-24 | 2021-08-13 | 北京市水利自动化研究所 | Rapid and accurate water meter metering identification method based on slimSSD model |
| CN113647920A (en) * | 2021-10-21 | 2021-11-16 | 青岛美迪康数字工程有限公司 | Method and device for reading vital sign data in monitoring equipment |
| CN113902914A (en) * | 2021-09-03 | 2022-01-07 | 华南理工大学 | Water meter reading automatic identification method based on microcontroller and convolutional neural network |
| CN113936280A (en) * | 2021-11-23 | 2022-01-14 | 河海大学 | Embedded instrument code disc character automatic identification system and method |
| CN116645682A (en) * | 2023-07-24 | 2023-08-25 | 济南瑞泉电子有限公司 | Water meter dial number identification method and system |
| CN117274971A (en) * | 2023-11-20 | 2023-12-22 | 深圳拓安信物联股份有限公司 | Image processing method applied to water meter data extraction and electronic equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101789080A (en) * | 2010-01-21 | 2010-07-28 | 上海交通大学 | Detection method for vehicle license plate real-time positioning character segmentation |
| CN108647686A (en) * | 2018-05-11 | 2018-10-12 | 同济大学 | A kind of water meter image Recognition of Reading method based on convolutional neural networks |
| CN108830271A (en) * | 2018-06-13 | 2018-11-16 | 深圳市云识科技有限公司 | A kind of digital displaying meter Recognition of Reading method based on convolutional neural networks |
| CN108875696A (en) * | 2018-07-05 | 2018-11-23 | 五邑大学 | The Off-line Handwritten Chinese Recognition method of convolutional neural networks is separated based on depth |
-
2019
- 2019-08-07 CN CN201910725909.6A patent/CN110490195A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101789080A (en) * | 2010-01-21 | 2010-07-28 | 上海交通大学 | Detection method for vehicle license plate real-time positioning character segmentation |
| CN108647686A (en) * | 2018-05-11 | 2018-10-12 | 同济大学 | A kind of water meter image Recognition of Reading method based on convolutional neural networks |
| CN108830271A (en) * | 2018-06-13 | 2018-11-16 | 深圳市云识科技有限公司 | A kind of digital displaying meter Recognition of Reading method based on convolutional neural networks |
| CN108875696A (en) * | 2018-07-05 | 2018-11-23 | 五邑大学 | The Off-line Handwritten Chinese Recognition method of convolutional neural networks is separated based on depth |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110956133A (en) * | 2019-11-29 | 2020-04-03 | 上海眼控科技股份有限公司 | Training method of single character text normalization model, text recognition method and device |
| CN111091558A (en) * | 2019-12-12 | 2020-05-01 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon swing bolster spring jumping fault image identification method |
| CN111160129A (en) * | 2019-12-12 | 2020-05-15 | 中国南方电网有限责任公司超高压输电公司广州局 | Intelligent remote meter reading method based on image recognition |
| CN111199194A (en) * | 2019-12-25 | 2020-05-26 | 吉林大学 | Automobile intelligent cabin instrument testing method based on machine vision and deep learning |
| CN111339832A (en) * | 2020-02-03 | 2020-06-26 | 中国人民解放军国防科技大学 | Method and device for detecting face synthetic image |
| CN111339832B (en) * | 2020-02-03 | 2023-09-12 | 中国人民解放军国防科技大学 | Face synthetic image detection method and device |
| CN112113638B (en) * | 2020-09-17 | 2024-02-02 | 北京慧怡科技有限责任公司 | Water meter function self-checking device and method |
| CN112113638A (en) * | 2020-09-17 | 2020-12-22 | 北京慧怡物联科技有限责任公司 | Water meter function self-checking device and method |
| CN112215178A (en) * | 2020-10-19 | 2021-01-12 | 南京大学 | Chemical experiment recording system based on pen type interaction |
| CN112215178B (en) * | 2020-10-19 | 2024-05-28 | 南京大学 | Chemical experiment recording system based on pen type interaction |
| CN112434693A (en) * | 2020-11-03 | 2021-03-02 | 辽宁长江智能科技股份有限公司 | Digital water meter identification method and system |
| CN113191351A (en) * | 2021-02-01 | 2021-07-30 | 青岛理工大学 | Reading identification method and device of digital electric meter and model training method and device |
| CN113191351B (en) * | 2021-02-01 | 2022-04-01 | 青岛理工大学 | Reading identification method and device of digital electric meter and model training method and device |
| CN113139541B (en) * | 2021-04-24 | 2023-10-24 | 西安交通大学 | Power distribution cabinet dial nixie tube visual identification method based on deep learning |
| CN113139541A (en) * | 2021-04-24 | 2021-07-20 | 西安交通大学 | Power distribution cabinet dial nixie tube visual identification method based on deep learning |
| CN113255650A (en) * | 2021-06-24 | 2021-08-13 | 北京市水利自动化研究所 | Rapid and accurate water meter metering identification method based on slimSSD model |
| CN113902914A (en) * | 2021-09-03 | 2022-01-07 | 华南理工大学 | Water meter reading automatic identification method based on microcontroller and convolutional neural network |
| CN113902914B (en) * | 2021-09-03 | 2024-04-02 | 华南理工大学 | Automatic water meter reading identification method based on microcontroller and convolutional neural network |
| CN113647920A (en) * | 2021-10-21 | 2021-11-16 | 青岛美迪康数字工程有限公司 | Method and device for reading vital sign data in monitoring equipment |
| CN113936280A (en) * | 2021-11-23 | 2022-01-14 | 河海大学 | Embedded instrument code disc character automatic identification system and method |
| CN113936280B (en) * | 2021-11-23 | 2024-04-05 | 河海大学 | Automatic character recognition system and method for code disc of embedded instrument |
| CN116645682A (en) * | 2023-07-24 | 2023-08-25 | 济南瑞泉电子有限公司 | Water meter dial number identification method and system |
| CN116645682B (en) * | 2023-07-24 | 2023-10-20 | 济南瑞泉电子有限公司 | Water meter dial number identification method and system |
| CN117274971A (en) * | 2023-11-20 | 2023-12-22 | 深圳拓安信物联股份有限公司 | Image processing method applied to water meter data extraction and electronic equipment |
| CN117274971B (en) * | 2023-11-20 | 2024-04-12 | 深圳拓安信物联股份有限公司 | Image processing method applied to water meter data extraction and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110490195A (en) | A kind of water meter dial plate Recognition of Reading method | |
| CN110570396B (en) | Industrial product defect detection method based on deep learning | |
| Hossain et al. | Improving consumer satisfaction in smart cities using edge computing and caching: A case study of date fruits classification | |
| CN104346440B (en) | A kind of across media hash indexing methods based on neutral net | |
| Kang et al. | Nuclei segmentation in histopathological images using two-stage learning | |
| CN109635744A (en) | A kind of method for detecting lane lines based on depth segmentation network | |
| Li et al. | Light-weight spliced convolution network-based automatic water meter reading in smart city | |
| CN111127360B (en) | Gray image transfer learning method based on automatic encoder | |
| CN115457004B (en) | Intelligent detection method of zinc paste based on computer vision | |
| CN110457982A (en) | A kind of crop disease image-recognizing method based on feature transfer learning | |
| CN108985342A (en) | A kind of uneven classification method based on depth enhancing study | |
| CN110175615A (en) | The adaptive visual position recognition methods in model training method, domain and device | |
| CN112784749A (en) | Target model training method, target object identification method, target model training device, target object identification device and medium | |
| CN113191385A (en) | Unknown image classification automatic labeling method based on pre-training labeling data | |
| CN114283162A (en) | Real scene image segmentation method based on contrast self-supervision learning | |
| CN112132827A (en) | Pathological image processing method and device, electronic equipment and readable storage medium | |
| Huang et al. | Fine-art painting classification via two-channel deep residual network | |
| CN110084327A (en) | Bill Handwritten Digit Recognition method and system based on the adaptive depth network in visual angle | |
| CN113902914B (en) | Automatic water meter reading identification method based on microcontroller and convolutional neural network | |
| CN110136113B (en) | Vagina pathology image classification method based on convolutional neural network | |
| WO2020119624A1 (en) | Class-sensitive edge detection method based on deep learning | |
| CN109858501A (en) | A kind of two phase flow pattern feature extracting method | |
| CN110647897A (en) | Zero sample image classification and identification method based on multi-part attention mechanism | |
| CN108090504A (en) | Object identification method based on multichannel dictionary | |
| Yuan et al. | Image segmentation based on modified superpixel segmentation and spectral clustering |
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: 20191122 |
|
| RJ01 | Rejection of invention patent application after publication |