CN109034160A - A kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks - Google Patents
A kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks Download PDFInfo
- Publication number
- CN109034160A CN109034160A CN201810734321.2A CN201810734321A CN109034160A CN 109034160 A CN109034160 A CN 109034160A CN 201810734321 A CN201810734321 A CN 201810734321A CN 109034160 A CN109034160 A CN 109034160A
- Authority
- CN
- China
- Prior art keywords
- decimal point
- neural networks
- convolutional neural
- led
- picture
- 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.)
- Granted
Links
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
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/153—Segmentation of character regions using recognition of characters or words
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
Abstract
The invention discloses a kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks, include the following steps: for the digital instrument LED picture sample of acquisition to be divided into independent LED character picture, LED character picture is sent into network model after pretreatment and is trained;Picture to be identified is inputted trained network model to identify.Wherein, network model is made of LED character convolutional neural networks model and decimal point convolutional neural networks model, and the preprocessing process of LED character picture includes LED numeral sample image preprocessing step and decimal point samples pictures pre-treatment step.The present invention is sent into training in network model after carrying out region cutting after the LED character picture scaling comprising decimal point, i.e., recurrence orientation problem is converted to classification problem.Because decimal point and LED character recognition are two different networks, model recognition result will not be interfered with each other, more flexible in terms of Networked E-Journals.
Description
Technical field
This technology fields are computer field, and in particular to the identification of the digital electric meter into picture is applied to number
The automatic identification of word ammeter.
Background technique
LED digital electric meter is common in novel meter, compared to tradition machinery formula ammeter, has accuracy height, low in energy consumption,
It is small in size, the advantages that being easily recognized, it is widely used in chemical industry, electronics, the industrial circles such as electric power.However under many occasions, these
The Recognition of Reading of LED ammeter needs manual work, not only large labor intensity, inefficiency, and certain scenes have risk
Factor, such as the LED of high voltage substation read operation.
Number of patent application is 201710195995.5, entitled " a kind of substation inspection robot autonomous classification method and one
Kind of inspecting robot " Chinese patent application, first binary conversion treatment carried out to LED number, it is rear by vertical direction
Projection carries out Character segmentation, obtains the cutting picture of single character;After picture is zoomed to unified size, pass through template matching
Method carries out score judgement to digital picture to be predicted, using the highest template serial number of matching score as recognition result.This method
It is very high to LED digital picture quality requirement, it obscures, inclination, illumination variation will lead to binaryzation cannot distinguish LED character well
With the edge of background.During subsequent template matching, matching is caused to malfunction.Lack versatility and robustness.
Number of patent application, which is 201710195624.7, entitled " a kind of instrument liquid crystal digital automatic segmentation and to know method for distinguishing
And system " Chinese patent application, disclose the method and system of a kind of automatic segmentation of liquid crystal instrument number and identification, including figure
As pretreatment, decimal point identification, character cutting, four modules of character recognition.Picture is filtered using LoG operator, Otsu
Dilation erosion operation is carried out after binaryzation, obtains connected graph, it is both horizontally and vertically projected for single character
Cutting.It extracts feature vector and uses the single character zone of individual character identification library identification cutting.It is larger by being arranged for decimal point
Threshold value binaryzation is carried out to image so that only retaining the binary picture comprising decimal point and noise, pass through and determine binaryzation
Region boundary rectangle carries out judgement comparison with binaryzation character zone before, confirms the relative position of decimal point.This method process
There are biggish limitations.In terms of character two-value cutting because light environment variation cause to cannot be distinguished after binaryzation character with
Background, decimal point location determination are set by threshold value, and process does not have versatility and stability in this way for institute.And character is cut
Divide and decimal point location depends on artificial priori decision logic.Algorithm can fail in true environment complicated and changeable.
LED digital electric meter is because being widely used in multiple fields, so it comes in every shape, leads to LED font typeface, font face
Color, the differences such as LED backplane texture can not be to various shapes in using conventional exercises method flow (such as characteristics extraction+SVM)
Reasonable threshold value is arranged in the LED font picture that state varies in color.Model after leading to training is on the picture of certain poor qualities
Recognition effect is poor.
Summary of the invention
Goal of the invention: it is directed to the above-mentioned prior art, proposes a kind of mixed decimal point digital instrument based on convolutional neural networks
Automatic identifying method can have good generalization to a variety of different LED fonts to the CNN network of LED character building.
Technical solution: a kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks, including it is as follows
Step: the digital instrument LED picture sample of acquisition is divided into independent LED character picture, the LED character picture is by pre-
Network model is sent into after processing to be trained;Picture to be identified is inputted trained network model to identify;Wherein, described
Network model is made of LED character convolutional neural networks model and decimal point convolutional neural networks model, the LED character picture
Preprocessing process include LED numeral sample image preprocessing step and decimal point samples pictures pre-treatment step, it is specific:
The LED numeral sample image preprocessing step includes:
A1: label calibration is carried out to LED character picture;
A2: data augmentation is carried out to the LED character picture of tape label;
A3: all LED character picture sizes are unified;
A4: gray processing is carried out to all LED character pictures, only retains luminance information;
A5: by after all gray processings sample LED carry out data be packaged encapsulation, be divided into training and test two data packets with
It is tested for the LED character convolutional neural networks model training;
The decimal point samples pictures pre-treatment step includes:
B1: label calibration is carried out to LED character picture;
B2: data augmentation is carried out to the LED character picture of tape label;
B3: all LED character picture sizes are unified;
B4: character picture is subjected to 3*3 cutting, randomly selects two sub-regions in preceding 8 regions, label is set as 0, such as
The fruit LED picture contains decimal point, then the 9th area label in the lower right corner is 1;
B5: the decimal point samples pictures of all cuttings are subjected to gray processing, only retain luminance information;
B6: being packaged packing for all decimal point samples pictures, is divided into described in two data packets confessions of training and test
The test of decimal point convolutional neural networks model training;
When carrying out picture using trained network model: first inputting the LED after pre-processing to images to be recognized
Then character convolutional neural networks model carries out 3x3 cutting to pretreated images to be recognized, take described in the feeding of the 9th region
There is detection according to decimal point and identifies successful order, determines the position of decimal point, finally in decimal point convolutional neural networks model
The LED character convolutional neural networks model and decimal point convolutional neural networks model recognition result are spliced.
Further, the LED character convolutional neural networks model and decimal point convolutional neural networks model are all made of and repair
Main track unit is as nonlinear activation function, and accordingly network weight uses xavier mode;Wherein, modified line unit isProportion range isY in formulaiIndicate nonlinear activation function value, xiIndicate letter
Number variable, m, n respectively represent network layer and output and input port number.
Further, the mobile mean value m of the index for gradient being updated using Adam network weight more new algorithm in model trainingtWith
Squared gradient vt, update rule are as follows:
Wherein,It is the single order moments estimation to gradient,For the second order moments estimation to gradient, η is learning rate, β1、β2, ε be
Hyper parameter, t indicate renewal time step.
Further, the LED character convolutional neural networks model includes 4 convolutional layers, 4 maximum pond layers and 2
A full articulamentum is constituted, and the decimal point convolutional neural networks model includes that 3 convolutional layers, 1 maximum pond layer and 2 are complete
Articulamentum is constituted.
Further, the LED character convolutional neural networks model and decimal point convolutional neural networks model are all made of
Softmax loses the foundation updated as network weight, Softmax loss function are as follows:Wherein,
T is network categorical measure, yiFor the label value of sample, siFor the predicted value of network query function,A is yiVector it is single
Element value, the serial number that j is.
Further, the training effect of network is judged in model training according to network current predictive accuracy rate, specifically:
Wherein,Representative takes label yiThe maximum serial number of vector;Representative takes net
Network calculates predicted value siMaximum possible label value;Accuracy indicates accuracy rate, and BatchSize indicates batch size.
The utility model has the advantages that this specification displaying is applied to LED character and decimal point recognition methods using convolutional neural networks,
Template matching or characteristics extraction carried out after comparison binary conversion treatment be sent into SVM or BP network the methods of identify have
Recognition accuracy is high, and pretreatment is simple, wide usage, the strong advantage of stability.In practical various illumination, the different LED word of pattern
Body and decimal point, can be practical.This method is applied to instrument monitoring such as and identifies, under the scenes such as substation inspection, it will big
It is big to reduce cost of labor, improve detection efficiency.
Detailed description of the invention
Fig. 1 is the method for the present invention flow chart;
Fig. 2 is decimal point 3x3 cutting schematic diagram;
Fig. 3 is LED character convolutional neural networks model schematic;
Fig. 4 is decimal point convolutional neural networks model schematic.
Specific embodiment
Further explanation is done to the present invention with reference to the accompanying drawing.
As shown in Figure 1, a kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks, including it is as follows
Step: the digital instrument LED picture sample of acquisition is divided into independent LED character picture, LED character picture is by pretreatment
It is sent into network model afterwards to be trained, then picture to be identified is inputted into trained network model and is identified.Wherein, network mould
Type is made of LED character convolutional neural networks model and decimal point convolutional neural networks model.LED character picture it is pretreated
Journey includes LED numeral sample image preprocessing step and decimal point samples pictures pre-treatment step, specific:
LED numeral sample image preprocessing step includes:
A1: label calibration is carried out to LED character picture, i.e., type calibration is carried out to the single character that interception comes out, needed
It identifies 0-9, A, B, C and does not work and unusual character, distribute 14 labels altogether, do not work and abnormal merge into same class;
A2: data augmentation is carried out to the LED character picture of tape label;
A3: all LED character picture sizes are unified to 48*48;
A4: gray processing is carried out to all LED character pictures, only retains luminance information;
A5: by after all gray processings sample LED carry out data be packaged encapsulation, be divided into training and test two data packets with
It is tested for LED character convolutional neural networks model training.
Decimal point samples pictures pre-treatment step includes:
B1: label calibration is carried out to LED character picture, the label containing decimal point is 1, and the label without decimal point is 0;
B2: data augmentation is carried out to the LED character picture of tape label;
B3: all LED character picture sizes are unified to 48*48;
B4: 48*48 character picture is subjected to 3*3 cutting, the first eight cutting region must be free of decimal point, in preceding 8 areas
Two sub-regions are randomly selected in domain, label is set as 0, if the LED picture contains decimal point, the 9th region mark in the lower right corner
Label are 1;Wherein, taking 2 at random in preceding 8 regions is to guarantee that positive and negative sample proportion will not differ too big, goes bail for demonstrate,prove negative sample at random
Originally rich.
B5: by the decimal point samples pictures of all cuttings, i.e., the positive sample of the 9th region formation after all samples pictures cuttings
This progress gray processing only retains luminance information;
B6: being packaged packing for all decimal point samples pictures, is divided into two data packets of training and test for decimal
Point convolutional neural networks model training test.
When carrying out picture using trained network model: being zoomed in and out to images to be recognized and gray processing pre-processes
(Resize+gray) the LED character convolutional neural networks model is first inputted after, then to pretreated images to be recognized into
Row 3x3 cutting takes the 9th region feeding decimal point convolutional neural networks model, and it is successfully secondary detection identification occur according to decimal point
Sequence determines the position of decimal point, finally identifies LED character convolutional neural networks model and decimal point convolutional neural networks model
As a result spliced.Such as 12.34, decimal point identifies successfully in second character zone, then decimal point is located at second, finally
Splicing result is (LED:1 234, Dot:2) -> 12.34.
LED character convolutional neural networks model and decimal point convolutional neural networks model be all made of modified line unit (Relu,
Rectified Linear Unit) it is used as nonlinear activation function, accordingly network weight uses xavier mode, to guarantee
Network can be trained normally.Wherein, modified line unit isProportion range isY in formulaiIndicate nonlinear activation function value, xiRepresentative function variable, m, n respectively represent network layer
Output and input port number.
Using Adam (the Adaptive Moment for the learning rate that can adaptively modify each parameter in model training
Estimation) network weight more new algorithm updates the mobile mean value m of index of gradienttWith squared gradient vt, update rule are as follows:
Wherein,It is the single order moments estimation to gradient,For the second order moments estimation to gradient, η is learning rate, and t is indicated more
New time step, β1、β2, ε be hyper parameter, β1It is defaulted as 0.9, β2Being defaulted as 0.999, ε is to be defaulted as 10e-8.
LED character convolutional neural networks model includes 4 convolutional layers, 4 maximum pond layers and 2 full articulamentum structures
At specific structure are as follows:
Type | Configutations | Size |
Input | 48x48gray-scale image | 48x48x1 |
Convolution | #maps:20,k:3x3,s:1,p:1 | 48x48x20 |
MaxPooling | Window:2x2,s:2 | 24x24x20 |
Convolution | #maps:50,k:3x3,s:1,p:1 | 24x24x50 |
MaxPooling | Window:2x2,s:2 | 12x12x50 |
Convolution | #maps:128,k:3x3,s:1,p:1 | 12x12x128 |
MaxPooling | Window:2x2,s:2 | 6x6x128 |
Convolution | #maps:256,k:3x3,s:1 | 6x6x256 |
MaxPooling | Window:2x2,s:2 | 3x3x256 |
Fully-Connection | #maps:1024 | 1024x1 |
Fully-Connection | #maps:14 | 14x1 |
Decimal point convolutional neural networks model includes that 3 convolutional layers, 1 maximum pond layer and 2 full articulamentums are constituted,
Specific structure are as follows:
LED character convolutional neural networks model and decimal point convolutional neural networks model are all made of Softmax loss conduct
The foundation that network weight updates, Softmax loss function are as follows:Wherein, T is network classification number
The length of amount, i.e. network output;yiIt is the vector that length is T for the label value of sample;siIt is long for the predicted value of network query function
Degree is the vector of T;A is yiThe individual element value of vector, j are the serial number of a, and j=0 represents yi0th member in vector
Element value.Here Softmax is the calculating loss of network, i.e., using Softmax loss as criterion, using Adam algorithm to net
Network weight is updated.
It not only needs according to network losses, while also to judge according to network current predictive accuracy rate in the training process
The training effect of network, specifically:
Wherein,Representative takes label yiThe maximum serial number of vector;Label compared with predicted value before can be into
Row vector, it is assumed that classification number is 14, and label 3=[0,0,0,1,0,0,0,0,0,0,0,0,0,0] can be taken using argmax
Outgoing label 3, serial number is since 0.Representative takes label siThe maximum serial number of vector;Accuracy indicates accurate
Rate, BatchSize indicate batch size.
In the prior art, for decimal point fixation and recognition, there is classification and return two kinds of training methods.Classification: will contain small
The picture sample of several points is set as 1, and the picture sample without decimal point is set as 0.It returns: size in the picture sample of decimal point is contracted
It puts to 48*48, using coordinate value of the decimal point in picture as label data.If classification method is directly used, training identification
Effect all can be excessively poor, because decimal point region accounting in entire LED character picture is too small.And homing method is difficult to train,
It is difficult to apply.
LED character picture comprising decimal point is zoomed to the size of 48*48 by the present invention, then carries out the cutting of the region 3*3
At the sub-pictures of 9 16*16, decimal point must be located at the 9th cutting region.Subregional label setting is cut containing decimal point
It is 1, other cutting area labels without decimal point are set as 0.Will these sub-pictures be sent into CNN network in training, i.e., return
Orientation problem is returned to be converted to classification problem.Because decimal point and LED character recognition are two different networks, model identification
As a result it will not interfere with each other, it is more flexible in terms of Networked E-Journals.
The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also answered
It is considered as protection scope of the present invention.
Claims (6)
1. a kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks, which is characterized in that including as follows
Step: the digital instrument LED picture sample of acquisition is divided into independent LED character picture, the LED character picture is by pre-
Network model is sent into after processing to be trained;Picture to be identified is inputted trained network model to identify;Wherein, described
Network model is made of LED character convolutional neural networks model and decimal point convolutional neural networks model, the LED character picture
Preprocessing process include LED numeral sample image preprocessing step and decimal point samples pictures pre-treatment step, it is specific:
The LED numeral sample image preprocessing step includes:
A1: label calibration is carried out to LED character picture;
A2: data augmentation is carried out to the LED character picture of tape label;
A3: all LED character picture sizes are unified;
A4: gray processing is carried out to all LED character pictures, only retains luminance information;
A5: carrying out data for the sample LED after all gray processings and be packaged encapsulation, is divided into two data packets of training and test for institute
State the test of LED character convolutional neural networks model training;
The decimal point samples pictures pre-treatment step includes:
B1: label calibration is carried out to LED character picture;
B2: data augmentation is carried out to the LED character picture of tape label;
B3: all LED character picture sizes are unified;
B4: carrying out 3*3 cutting for character picture, randomly select two sub-regions in preceding 8 regions, and label is set as 0, if should
LED picture contains decimal point, then the 9th area label in the lower right corner is 1;
B5: the decimal point samples pictures of all cuttings are subjected to gray processing, only retain luminance information;
B6: being packaged packing for all decimal point samples pictures, is divided into two data packets of training and test for the decimal
Point convolutional neural networks model training test;
When carrying out picture using trained network model: first inputting the LED character after pre-processing to images to be recognized
Then convolutional neural networks model carries out 3x3 cutting to pretreated images to be recognized, the 9th region is taken to be sent into the decimal
There is detection according to decimal point and identifies successful order, the position of decimal point determined, finally by institute in point convolutional neural networks model
It states LED character convolutional neural networks model and decimal point convolutional neural networks model recognition result is spliced.
2. the mixed decimal point digital instrument automatic identifying method according to claim 1 based on convolutional neural networks, special
Sign is that the LED character convolutional neural networks model and decimal point convolutional neural networks model are all made of modified line unit work
For nonlinear activation function, accordingly network weight uses xavier mode;Wherein, modified line unit isProportion range isY in formulaiIndicate nonlinear activation function value, xiIt indicates
Function variable, m, n respectively represent network layer and output and input port number.
3. the mixed decimal point digital instrument automatic identifying method according to claim 2 based on convolutional neural networks, special
Sign is, updates the index movement mean value m of gradient in model training using Adam network weight more new algorithmtWith squared gradient vt,
Update rule are as follows:
Wherein,It is the single order moments estimation to gradient,For the second order moments estimation to gradient, η is learning rate, β1、β2, ε be super ginseng
Number, t indicate renewal time step.
4. the mixed decimal point digital instrument automatic identifying method according to claim 2 based on convolutional neural networks, special
Sign is that the LED character convolutional neural networks model includes 4 convolutional layers, 4 maximum pond layers and 2 full articulamentums
It constitutes, the decimal point convolutional neural networks model includes 3 convolutional layers, 1 maximum pond layer and 2 full articulamentum structures
At.
5. the mixed decimal point digital instrument automatic identifying method according to claim 3 based on convolutional neural networks, special
Sign is that the LED character convolutional neural networks model and decimal point convolutional neural networks model are all made of Softmax loss and make
For the foundation that network weight updates, Softmax loss function are as follows:Wherein, T is network classification number
Amount, yiFor the label value of sample, siFor the predicted value of network query function,A is yiThe individual element value of vector, j
Serial number.
6. the mixed decimal point digital instrument automatic identifying method according to claim 5 based on convolutional neural networks, special
Sign is, judges the training effect of network in model training according to network current predictive accuracy rate, specifically:
Wherein,Representative takes label yiThe maximum serial number of vector;Representative takes network meter
Calculate predicted value siMaximum possible label value;Accuracy indicates accuracy rate, and BatchSize indicates batch size.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810734321.2A CN109034160B (en) | 2018-07-06 | 2018-07-06 | A kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810734321.2A CN109034160B (en) | 2018-07-06 | 2018-07-06 | A kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109034160A true CN109034160A (en) | 2018-12-18 |
CN109034160B CN109034160B (en) | 2019-07-12 |
Family
ID=64641172
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810734321.2A Active CN109034160B (en) | 2018-07-06 | 2018-07-06 | A kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109034160B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109902751A (en) * | 2019-03-04 | 2019-06-18 | 福州大学 | A kind of dial digital character identifying method merging convolutional neural networks and half-word template matching |
CN110033037A (en) * | 2019-04-08 | 2019-07-19 | 重庆邮电大学 | A kind of recognition methods of digital instrument reading |
CN110197227A (en) * | 2019-05-30 | 2019-09-03 | 成都中科艾瑞科技有限公司 | A kind of meter reading intelligent identification Method of multi-model fusion |
CN110298347A (en) * | 2019-05-30 | 2019-10-01 | 长安大学 | A kind of recognition methods of the automobile exhaust analyzer screen based on GrayWorld and PCA-CNN |
CN110346516A (en) * | 2019-07-19 | 2019-10-18 | 精英数智科技股份有限公司 | Fault detection method and device, storage medium |
CN111368824A (en) * | 2020-02-24 | 2020-07-03 | 河海大学常州校区 | Instrument identification method, mobile device and storage medium |
CN112200160A (en) * | 2020-12-02 | 2021-01-08 | 成都信息工程大学 | Deep learning-based direct-reading water meter reading identification method |
CN112348018A (en) * | 2020-11-16 | 2021-02-09 | 杭州安森智能信息技术有限公司 | Digital display type instrument reading identification method based on inspection robot |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101030258A (en) * | 2006-02-28 | 2007-09-05 | 浙江工业大学 | Dynamic character discriminating method of digital instrument based on BP nerve network |
CN101079108A (en) * | 2007-06-29 | 2007-11-28 | 浙江工业大学 | DSP based multiple channel mechanical digital display digital gas meter automatic detection device |
US20090060396A1 (en) * | 2007-08-30 | 2009-03-05 | Xerox Corporation | Features generation and spotting methods and systems using same |
CN105184265A (en) * | 2015-09-14 | 2015-12-23 | 哈尔滨工业大学 | Self-learning-based handwritten form numeric character string rapid recognition method |
CN105654130A (en) * | 2015-12-30 | 2016-06-08 | 成都数联铭品科技有限公司 | Recurrent neural network-based complex image character sequence recognition system |
CN105809179A (en) * | 2014-12-31 | 2016-07-27 | 中国科学院深圳先进技术研究院 | Pointer type instrument reading recognition method and device |
CN106529537A (en) * | 2016-11-22 | 2017-03-22 | 亿嘉和科技股份有限公司 | Digital meter reading image recognition method |
CN106960208A (en) * | 2017-03-28 | 2017-07-18 | 哈尔滨工业大学 | A kind of instrument liquid crystal digital automatic segmentation and the method and system of identification |
CN108133216A (en) * | 2017-11-21 | 2018-06-08 | 武汉中元华电科技股份有限公司 | The charactron Recognition of Reading method that achievable decimal point based on machine vision is read |
-
2018
- 2018-07-06 CN CN201810734321.2A patent/CN109034160B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101030258A (en) * | 2006-02-28 | 2007-09-05 | 浙江工业大学 | Dynamic character discriminating method of digital instrument based on BP nerve network |
CN101079108A (en) * | 2007-06-29 | 2007-11-28 | 浙江工业大学 | DSP based multiple channel mechanical digital display digital gas meter automatic detection device |
US20090060396A1 (en) * | 2007-08-30 | 2009-03-05 | Xerox Corporation | Features generation and spotting methods and systems using same |
CN105809179A (en) * | 2014-12-31 | 2016-07-27 | 中国科学院深圳先进技术研究院 | Pointer type instrument reading recognition method and device |
CN105184265A (en) * | 2015-09-14 | 2015-12-23 | 哈尔滨工业大学 | Self-learning-based handwritten form numeric character string rapid recognition method |
CN105654130A (en) * | 2015-12-30 | 2016-06-08 | 成都数联铭品科技有限公司 | Recurrent neural network-based complex image character sequence recognition system |
CN106529537A (en) * | 2016-11-22 | 2017-03-22 | 亿嘉和科技股份有限公司 | Digital meter reading image recognition method |
CN106960208A (en) * | 2017-03-28 | 2017-07-18 | 哈尔滨工业大学 | A kind of instrument liquid crystal digital automatic segmentation and the method and system of identification |
CN108133216A (en) * | 2017-11-21 | 2018-06-08 | 武汉中元华电科技股份有限公司 | The charactron Recognition of Reading method that achievable decimal point based on machine vision is read |
Non-Patent Citations (2)
Title |
---|
崔行臣 等: "数显仪表数字实时识别系统的设计与实现", 《计算机工程与设计》 * |
程敏: "一种基于显著性检测的LED仪表字符自动识别方法", 《信息与电脑(理论版)》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109902751A (en) * | 2019-03-04 | 2019-06-18 | 福州大学 | A kind of dial digital character identifying method merging convolutional neural networks and half-word template matching |
CN109902751B (en) * | 2019-03-04 | 2022-07-08 | 福州大学 | Dial digital character recognition method integrating convolution neural network and half-word template matching |
CN110033037A (en) * | 2019-04-08 | 2019-07-19 | 重庆邮电大学 | A kind of recognition methods of digital instrument reading |
CN110197227A (en) * | 2019-05-30 | 2019-09-03 | 成都中科艾瑞科技有限公司 | A kind of meter reading intelligent identification Method of multi-model fusion |
CN110298347A (en) * | 2019-05-30 | 2019-10-01 | 长安大学 | A kind of recognition methods of the automobile exhaust analyzer screen based on GrayWorld and PCA-CNN |
CN110298347B (en) * | 2019-05-30 | 2022-11-01 | 长安大学 | Method for identifying automobile exhaust analyzer screen based on GrayWorld and PCA-CNN |
CN110197227B (en) * | 2019-05-30 | 2023-10-27 | 成都中科艾瑞科技有限公司 | Multi-model fusion intelligent instrument reading identification method |
CN110346516A (en) * | 2019-07-19 | 2019-10-18 | 精英数智科技股份有限公司 | Fault detection method and device, storage medium |
CN111368824A (en) * | 2020-02-24 | 2020-07-03 | 河海大学常州校区 | Instrument identification method, mobile device and storage medium |
CN111368824B (en) * | 2020-02-24 | 2022-09-23 | 河海大学常州校区 | Instrument identification method, mobile device and storage medium |
CN112348018A (en) * | 2020-11-16 | 2021-02-09 | 杭州安森智能信息技术有限公司 | Digital display type instrument reading identification method based on inspection robot |
CN112200160A (en) * | 2020-12-02 | 2021-01-08 | 成都信息工程大学 | Deep learning-based direct-reading water meter reading identification method |
Also Published As
Publication number | Publication date |
---|---|
CN109034160B (en) | 2019-07-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109034160B (en) | A kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks | |
CN106529537B (en) | A kind of digital instrument reading image-recognizing method | |
CN110059694A (en) | The intelligent identification Method of lteral data under power industry complex scene | |
CN109446925A (en) | A kind of electric device maintenance algorithm based on convolutional neural networks | |
CN103049763B (en) | Context-constraint-based target identification method | |
CN109712118A (en) | A kind of substation isolating-switch detection recognition method based on Mask RCNN | |
CN109977780A (en) | A kind of detection and recognition methods of the diatom based on deep learning algorithm | |
CN109919934B (en) | Liquid crystal panel defect detection method based on multi-source domain deep transfer learning | |
CN108830188A (en) | Vehicle checking method based on deep learning | |
Hongwei et al. | Solder joint inspection method for chip component using improved AdaBoost and decision tree | |
CN110070536A (en) | A kind of pcb board component detection method based on deep learning | |
CN110310259A (en) | It is a kind of that flaw detection method is tied based on the wood for improving YOLOv3 algorithm | |
CN106557173B (en) | Dynamic gesture identification method and device | |
CN110991435A (en) | Express waybill key information positioning method and device based on deep learning | |
CN109583324A (en) | A kind of pointer meters reading automatic identifying method based on the more box detectors of single-point | |
CN109063706A (en) | Verbal model training method, character recognition method, device, equipment and medium | |
CN104268538A (en) | Online visual inspection method for dot matrix sprayed code characters of beverage cans | |
CN104680144A (en) | Lip language recognition method and device based on projection extreme learning machine | |
CN110598693A (en) | Ship plate identification method based on fast-RCNN | |
CN110569843B (en) | Intelligent detection and identification method for mine target | |
CN108537222A (en) | A kind of image-recognizing method and system for electric instrument | |
CN108764134A (en) | A kind of automatic positioning of polymorphic type instrument and recognition methods suitable for crusing robot | |
CN108921152A (en) | English character cutting method and device based on object detection network | |
CN107038416A (en) | A kind of pedestrian detection method based on bianry image modified HOG features | |
CN110223310A (en) | A kind of line-structured light center line and cabinet edge detection method based on deep learning |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
PP01 | Preservation of patent right | ||
PP01 | Preservation of patent right |
Effective date of registration: 20220517 Granted publication date: 20190712 |
|
PD01 | Discharge of preservation of patent | ||
PD01 | Discharge of preservation of patent |
Date of cancellation: 20230314 Granted publication date: 20190712 |