CN110427943A - A kind of intelligent electric meter technique for partitioning based on R-CNN - Google Patents
A kind of intelligent electric meter technique for partitioning based on R-CNN Download PDFInfo
- Publication number
- CN110427943A CN110427943A CN201910798558.1A CN201910798558A CN110427943A CN 110427943 A CN110427943 A CN 110427943A CN 201910798558 A CN201910798558 A CN 201910798558A CN 110427943 A CN110427943 A CN 110427943A
- Authority
- CN
- China
- Prior art keywords
- ammeter
- dial plate
- network
- image
- cnn
- 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
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000000638 solvent extraction Methods 0.000 title claims abstract description 10
- 238000012549 training Methods 0.000 claims abstract description 19
- 238000000605 extraction Methods 0.000 claims abstract description 15
- 230000008569 process Effects 0.000 claims abstract description 9
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 238000010845 search algorithm Methods 0.000 claims abstract description 4
- 238000013527 convolutional neural network Methods 0.000 claims description 45
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 8
- 238000001514 detection method Methods 0.000 claims description 5
- 230000005611 electricity Effects 0.000 claims description 4
- 239000013598 vector Substances 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 238000009826 distribution Methods 0.000 claims description 2
- 238000011156 evaluation Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000003709 image segmentation Methods 0.000 claims 1
- 230000000694 effects Effects 0.000 description 5
- 238000005286 illumination Methods 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000003706 image smoothing Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 241000196324 Embryophyta Species 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004925 denaturation Methods 0.000 description 1
- 230000036425 denaturation Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000005304 joining Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- 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
Abstract
The invention discloses a kind of intelligent electric meter technique for partitioning based on R-CNN, off-line training, it acquires the ammeter picture largely taken with camera, and these ammeter picture samples are pre-processed with image pre-processing method, again to sample addition label in order to e-learning, the label is set as the reading value of dial plate, obtains the network of dial plate for identification;A series of on-line prediction comprising the ammeter dial plate band of position is extracted, and generates candidate regions by selective search algorithm first;Identification process is that the picture for having scaled each candidate region is input in R-CNN network and carries out feature extraction, it then whether is target area by SVM Network Recognition candidate region picture, that is ammeter dial plate region, dial plate information is identified by multilayer R-CNN network and fully-connected network, exports the ammeter reading finally identified.Recognition methods provided by the invention can rapidly and accurately identify that ammeter is read, and minimum to the limitation of ammeter image, practicability is good.
Description
Technical field
The invention belongs to ammeter numerical value intelligently to read technical field more particularly to a kind of intelligent electric meter number based on R-CNN
It is worth recognition methods.
Background technique
Ammeter is widely used in the various fields in social production life as a kind of measuring instrument, and plays act foot
The effect of weight.Currently, the reading to ammeter is mainly completed by artificial interpretation, however this method is influenced by human factor
Greatly, poor reliability, low efficiency.
In recent years, with the continuous development of digital image processing techniques and perfect, the identification of ammeter also obtains huge
Progress.For the automatic interpretation of ammeter, existing scientific research personnel is studied using machine vision technique, has expanded automatic identification
The applicable range of system can effectively identify common dynamometer instrument etc..They obtain pointer image using outline method,
And with Hough transform detect two pointers between angle, complete Recognition of Reading, have also been proposed to digital instrument from
Dynamic recognition methods.The localization method of pointer and graduation mark mainly has outline method, Hough transform method at this stage, these methods are mostly
Assuming that instrument imaging circumstances are ideal, i.e. dial plate places that regular, camera optical axis is vertical with dial plate plane, uniform illumination.
Outline method has higher accuracy of identification in illumination abundance, but when environment is less desirable, accuracy of identification is very poor;Hough method
When handling thicker pointer, due to the interference of uneven illumination or other factors, in the straight line and practical pointer that may detect that
There are certain deviations for heart line, affect the accuracy of identification of algorithm.
As machine learning and deep learning algorithm are widely used in feature autonomous learning and indicate, it is based on deep learning
Algorithm excellent performance is shown in intelligent recognition task.Typically there is a kind of intelligent electric meter identification based on deep learning
Method, this method identify ammeter picture by multilayer convolutional neural networks, relative to traditional intelligent electric meter identification side
Method, this method can reach good recognition effect by off-line training.However, traditional based on convolutional neural networks
There are still detection speed in ammeter identification for online learning methods such as (Convolutional Neural Network, CNN) slowly,
The problems such as accuracy of identification is low.On the one hand, traditional CNN can not accurately identify target position, i.e. ammeter in ammeter image
Dial plate region, this causes to need manually to pre-process picture before carrying out intelligent electric meter identification.On the other hand, in tradition
The ammeter recognition methods based on CNN in, network model training function is single, and time consumption for training is long, so that recognition efficiency is lower.And
And the environment of actual test work is complicated, and such as: have that various forms of dial plates, illuminating ray be uneven, glass of instrument surface
Reflection, the optical axis of camera and dial plate plane out of plumb and dial plate inclination etc. can be generated, these factors all can be to Instrument image
Identification cause difficulty, most of recognition methods will generate biggish error, or even failure.
Summary of the invention
It is an object of the invention to solve above-mentioned technical problem, a kind of intelligent electric meter numerical identification based on R-CNN is provided
Method is somebody's turn to do the intelligent electric meter technique for partitioning based on R-CNN using R-CNN network and is completed at the same time target area identification and ammeter
Scale identification, improves the efficiency and accuracy rate of ammeter identification.
Technical scheme is as follows: a kind of intelligent electric meter technique for partitioning based on R-CNN, further includes instructing offline
Experienced and on-line prediction;Wherein
Off-line training acquires the ammeter picture largely taken with camera, and with image pre-processing method to this
A little ammeter picture samples are pre-processed, described image preprocess method be image of gauge with pointer is carried out image gray processing,
Image smoothing filtering technique, image enhancement and image binaryzation processing, it is to be pre-treated complete and then to sample addition label in order to net
Network study, the label are set as the reading value of dial plate, obtain the network of dial plate for identification;
On-line prediction comprising the ammeter dial plate band of position is extracted, and is generated first by selective search algorithm a series of
Then candidate region is fixed the scaling of size for each candidate region, then is identified to it;Identification process
It is that the picture for having scaled each candidate region is input in R-CNN network and carries out feature extraction, then passes through SVM Network Recognition
Whether candidate region picture is target area, i.e. ammeter dial plate region;If having identified ammeter dial plate region, pass through multilayer
R-CNN network and fully-connected network identify dial plate information, export the ammeter reading finally identified, otherwise, repeat to be searched with selectivity
Rope algorithm generates a series of candidate regions, scales to candidate region, identifies dial plate by carrying out feature extraction in R-CNN network
Corresponding steps identify dial plate after Network Recognition goes out ammeter dial plate region, then through multilayer R-CNN network and fully-connected network
Information exports the ammeter reading finally identified.
Beneficial effects of the present invention: complex environment of the present invention towards Electric Power Patrol maintenance ammeter proposes a kind of based on area
The intelligent electric meter numerical value of domain convolutional neural networks (Region-based Convolutional Neural Network, R-CNN)
Recognition methods.This method is completed at the same time target area identification using R-CNN network and ammeter scale identifies, especially completes mesh
The automatic identification for marking region substantially increases the efficiency and accuracy rate of ammeter identification, relies on researcher's experience people relative to current
For design requirement image can just be read with automatic identification for, undoubtedly a kind of huge breakthrough, more adaptation actual demand,
Live camera is thoroughly got rid of to the limitation of dial plate shooting visual angle and range.The present invention is based on the countings of the intelligent meter of R-CNN
Being worth identifying system can be by the scale for the dial plate image and dial plate that history artificially collects to intelligent electricity for its off-line training
Table identification network is trained, for on-line prediction, using trained network for meter identification, when detection, first
The image information of ammeter digital dial is acquired by camera on wearable device, then by collect image information into
Row pretreatment, finally by intelligent electric meter identification network, to prediction, treated that image identifies, reaches and obtains correct meter
Purpose.
Detailed description of the invention
Fig. 1 is workflow schematic diagram of the invention.
Fig. 2 is that the present invention is based on the intelligent electric meters of R-CNN network to identify structural schematic diagram.
Specific embodiment
Embodiments of the present invention are illustrated by particular specific embodiment below, those skilled in the art can be by this explanation
Content disclosed by book is understood other advantages and efficacy of the present invention easily.
Present invention will be further explained below with reference to the attached drawings and examples:
The present embodiment provides a kind of intelligent electric meter technique for partitioning based on R-CNN, the intelligent meter based on R-CNN
Numerical identification system mainly includes two steps of off-line training and on-line prediction.For off-line training, operator passes through history people
The scale of dial plate image and dial plate that work is collected into is trained intelligent electric meter identification network.For on-line prediction, operator
It is identified using trained network for meter.
In the specific implementation: the person of being operated first acquires the image of ammeter digital dial by the camera on wearable device
Information, then by pre-processing to collecting image information, finally by intelligent electric meter identification network to prediction processing after
Image identified, achieve the purpose that obtain correct meter.It wherein, mainly include two steps: ammeter for ammeter identification
Regional location identification and the identification of ammeter information information.The position of dial plate is only first detected from complicated image, it just can be with
Ammeter information identification is carried out again, this is also one of the present invention and the maximum difference of the prior art.Ammeter identification groundwork be
The approximate region of dial plate is searched out in piece image and then confirms exact position and the size of dial plate again, and the image of shooting is (i.e.
The live corresponding picture of ammeter) it include a large amount of background information, but for ammeter information identification, these backgrounds are that have no
Meaning, so needing to handle ammeter picture, and then background information dial plate image as few as possible is obtained, then to dial plate
Image carries out ammeter information identification.The main working process of its present embodiment constituted is as shown in Figure 1.
Wherein, positioning feature point is also known as the alignment of ammeter dial plate, and in general, the ammeter identification of narrow sense only includes automatic ammeter
Most latter two link of identification process figure (Fig. 1), that is, the algorithm research of feature extraction and ammeter identification, this is being currently to answer
It can quick and precisely be identified with quite maturation, the standard picture in the case where meeting design condition.And in the present invention, using R-CNN net
Network and full Connection Neural Network identify ammeter, identify the calibration information of ammeter.When using R-CNN network training, need
Calibration information identification is carried out to the dial plate of extraction, i.e., be fully-connected network in the last layer, by fully-connected network by real part
Calibration information be converted into specific numerical value, to reach the identification of accurate scale.
Traditional algorithm of target detection relies primarily on the artificial design feature of researcher's experience, with mentioning for convolutional neural networks
Out, it has been found that can use convolutional neural networks and automatically extract feature, also, the feature extracted not only contains ammeter
Calibration information, the feature for also including the location information of ammeter, and extracting have displacement, scale, translation and deformation etc. no
Denaturation will be divided into following four step specific in ammeter dial plate identification of the invention:
(1) it generates candidate region (region proposal): for each image of input, using selective search
(Selective Search, SS) method generates 1K-2K candidate region.Using the method for over-segmentation, by the image of input point
It is segmented into zonule, about 1K-2K is a;According to the zonule being split to form, according to the grain distribution joining relation of dial plate image, pairing
And the highest two neighboring region of possibility merges, and constantly repeats this process, until whole image is merged into an entirety
Region;Export all regions merged out, that is, candidate region.
(2) feature extraction (feature extraction): for the candidate region of generation, depth convolutional Neural net is used
Network extracts feature, that is, 4096 dimensional feature vectors of the full articulamentum output of neural network.Before carrying out feature extraction, need pair
The candidate frame of input carries out size normalization, and naturalization is 227 × 227 at uniform sizes size.
(3) classification judgement (classification): linear SVM (Support Vector is used
Machine, SVM) classifier identifies the feature extracted, judges whether to belong to ammeter dial plate.It is instructed using CNN network
When practicing, need to carry out classification based training to the target with bounding box (Bounding Box, Bbox).
(4) position refine (rect refine): for each classification, recurrence device (regressor) basis is used
Evaluation criterion-degree of overlapping (IOU) of ammeter dial plate detection algorithm carries out refine to the position of candidate frame, generates prediction window
Coordinate.
Image Acquisition is one of first link for realizing wearable device automatic interpretation and most important link, is somebody's turn to do
Link is mainly the image that pointer instrument is collected in actual production and living environment, is transmitted to the wearable of operator and sets
Standby instrument carries out interpretation.There are many image capture devices that can complete the Image Acquisition in pointer instrument automatic recognition system at present
Demand, such as digital camera, image pick-up card and data image signal capture card etc..Meanwhile in Image Acquisition link, adopt
The picture quality collected is better, and later period recognition effect is more accurate.Good image should be very clear, resolution with higher
Rate, the target area of image and the boundary of background area are clearly demarcated, and background should be simple as far as possible, and disturbing factor is few, including illumination
How much also want moderate.But correspondingly, the image high for acquisition resolution, the requirement to video camera is just very high, price
Also sufficiently expensive.And image capture environment is varied, some environmental backgrounds are complicated, some ambient lightings are very strong or very weak,
These environment are also often to meet in the practice of pointer instrument automatic recognition system.Therefore, it is answered before Image Acquisition
The above-mentioned item of the attention.
Camera may in the image in acquisition instrument image due to being made acquisition by a variety of causes such as illumination, weathers
Containing noise, the accuracy of final recognition result is influenced.Therefore it just needs to obtain using the method for image procossing to from camera
Color image pre-processed, useful information in prominent image removes interference information, knows for the positioning of next step dial plate and pointer
It does not service.When pre-processing to image, image of gauge with pointer will carry out image gray processing, image smoothing filtering technique, image
Enhancing and image binaryzation processing.Fig. 2 shows a kind of, and the dial plate based on R-CNN identifies structure.
As shown in Fig. 2, the intelligent electric meter identification network based on R-CNN mainly includes two steps: ammeter dial plate position area
Domain is extracted and the identification of ammeter dial plate information.Before training network, need to prepare the data set of ammeter mark.The person of being operated first adopts
Collect the ammeter picture largely taken with camera, and they are pre-processed with image pre-processing method, then to sample
This addition label carries out in order to e-learning, and operator's label is set as the reading value of dial plate herein.Ammeter dial plate position area
Domain is extracted, and generates a series of candidate regions by SS method first, then candidate region is fixed for each candidate region
The scaling of size, then it is identified.Identification process is that the picture for having scaled each candidate region is input in CNN network
Feature extraction is carried out, whether is then target area by SVM Network Recognition candidate region picture, is i.e. ammeter dial plate region.Such as
Fruit has identified ammeter dial plate region, then identifies dial plate information by multi-layer C NN network and fully-connected network, exports final identification
Ammeter reading.Otherwise, repeat to generate a series of candidate regions with SS method, candidate region scaled, then in CNN network into
Row feature extraction identifies these steps of dial plate, after Network Recognition goes out ammeter dial plate region, then by multi-layer C NN network and
Fully-connected network identifies dial plate information, exports the ammeter reading finally identified.
In order to train neural network, operator is using all weights in point-to-point mode of learning learning network and partially
It sets.Operator is by fallAnd ΘallBe set to entire neural network recognition function and all parameters.Therefore entire neural network
Expression formula are as follows:
Y=fall(IMG,Θall)
Wherein IMG is the ammeter dial plate image being originally inputted, and y is the output of entire neural network.Operator's neural network
Input be the dial plate image that is saved by manual metering of history, label data is the corresponding scale value of each dial plate image.Net
All parameters in network are updated by ADAM algorithm, with the mean square error (MSE) of the available whole network of this operator
Expression formula is
Wherein label is label vector, and as true dial scale, T is total sample number.Operator passes through minimum
Above formula is trained neural network.
In the present embodiment, operator is first by artificially collecting the dial plate data of history, then to be collected into data into
Row pretreatment, weeds out the poor picture of some display effects, finally by pretreated pictures to convolutional Neural net
Network is trained.The data being collected into are divided into training set, checksum set and test set by operator, wherein training set, checksum set and
Test set separately includes 20000,3000 and 1000 data.Trained convolutional neural networks are used for ammeter and known by operator
Not, be configured to 1060 video card of GeForce GTX, Intel using training equipmentE3-1231V3 processor, it is real
Verifying bright recognition accuracy can achieve 93%.
The above-described embodiments merely illustrate the principles and effects of the present invention, and is not intended to limit the present invention.It is any ripe
The personage for knowing this technology all without departing from the spirit and scope of the present invention, carries out modifications and changes to above-described embodiment.Cause
This, institute is complete without departing from the spirit and technical ideas disclosed in the present invention by those of ordinary skill in the art such as
At all equivalent modifications or change, should be covered by the claims of the present invention.
Claims (4)
1. a kind of intelligent electric meter technique for partitioning based on R-CNN, it is characterised in that: further include off-line training and online pre-
It surveys;Wherein
Off-line training acquires the ammeter picture largely taken with camera, and with image pre-processing method to these electricity
Table picture sample is pre-processed, and described image preprocess method is to carry out image gray processing, image to image of gauge with pointer
Smothing filtering, image enhancement and image binaryzation processing, it is to be pre-treated complete and then to sample addition label in order to network science
It practises, the label is set as the reading value of dial plate, obtains the network of dial plate for identification;
A series of on-line prediction comprising the ammeter dial plate band of position is extracted, and generates candidates by selective search algorithm first
Then region is fixed the scaling of size for each candidate region, then is identified to it;Identification process be by
The picture that each candidate region has scaled, which is input in R-CNN network, carries out feature extraction, then candidate by SVM Network Recognition
Whether region picture is target area, i.e. ammeter dial plate region;If having identified ammeter dial plate region, pass through multilayer R-
CNN network and fully-connected network identify dial plate information, export the ammeter reading finally identified, otherwise, repetition selective search
Algorithm generates a series of candidate regions, scales to candidate region, the phase for identifying dial plate by carrying out feature extraction in R-CNN network
Step is answered, identifies dial plate letter after Network Recognition goes out ammeter dial plate region, then through multilayer R-CNN network and fully-connected network
Breath exports the ammeter reading finally identified.
2. the intelligent electric meter technique for partitioning based on R-CNN according to claim 1, it is characterised in that: carry out the electricity
When the table dial plate band of position is extracted, the R-CNN network is divided into following four step when working:
(1) it generates candidate region: for each image of input, generating 1K-2K candidate regions using selective search method
The image segmentation of input is 1K-2K zonule using the method for over-segmentation by domain;According to the zonule being split to form, according to
The grain distribution of dial plate image merges to the highest two neighboring region of possibility is merged, constantly repeats this merging process,
Until whole image is merged into an overall region, all regions merged out of output are up to candidate region;
(2) feature extraction: for the candidate region of generation, feature is extracted using depth convolutional neural networks, neural network connects entirely
It connects layer and exports 4096 dimensional feature vectors;Before carrying out feature extraction, need to carry out size normalization to the candidate frame of input;
(3) classification judges: being identified using linear SVM classifier to the feature extracted, judges whether to belong to electricity
Table dial plate;In R-CNN network training, need to carry out classification based training to the target with bounding box;
(4) position refine: for each classification, using a recurrence device according to the evaluation criterion-of ammeter dial plate detection algorithm
Degree of overlapping carries out refine to the position of candidate frame, generates the coordinate of the prediction window of on-line prediction.
3. the intelligent electric meter technique for partitioning based on R-CNN according to claim 2, it is characterised in that: carrying out feature
Before extraction, the size for carrying out unified naturalization to the candidate frame of input is 227 × 227.
4. the intelligent electric meter technique for partitioning based on R-CNN according to claim 1, it is characterised in that: carrying out offline
When training, all weights and the biasing in learning network are carried out using point-to-point mode of learning.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910798558.1A CN110427943A (en) | 2019-08-27 | 2019-08-27 | A kind of intelligent electric meter technique for partitioning based on R-CNN |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910798558.1A CN110427943A (en) | 2019-08-27 | 2019-08-27 | A kind of intelligent electric meter technique for partitioning based on R-CNN |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110427943A true CN110427943A (en) | 2019-11-08 |
Family
ID=68417807
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910798558.1A Pending CN110427943A (en) | 2019-08-27 | 2019-08-27 | A kind of intelligent electric meter technique for partitioning based on R-CNN |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110427943A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110909738A (en) * | 2019-11-15 | 2020-03-24 | 杭州远鉴信息科技有限公司 | Automatic reading method of pointer instrument based on key point detection |
CN111126395A (en) * | 2019-12-25 | 2020-05-08 | 福建天晴数码有限公司 | Optimization method of selective search algorithm in R-CNN network and storage medium |
CN113283419A (en) * | 2021-04-29 | 2021-08-20 | 国网浙江省电力有限公司湖州供电公司 | Convolutional neural network pointer instrument image reading identification method based on attention |
CN113556318A (en) * | 2021-06-07 | 2021-10-26 | 广州康辰科技有限公司 | E-commerce verification method based on cloud security |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108597204A (en) * | 2018-05-22 | 2018-09-28 | 广州市暨联牧科信息科技有限公司 | A kind of intelligent meter data recording system and its implementation |
CN109255375A (en) * | 2018-08-29 | 2019-01-22 | 长春博立电子科技有限公司 | Panoramic picture method for checking object based on deep learning |
US20190095739A1 (en) * | 2017-09-27 | 2019-03-28 | Harbin Institute Of Technology | Adaptive Auto Meter Detection Method based on Character Segmentation and Cascade Classifier |
CN109543682A (en) * | 2018-11-23 | 2019-03-29 | 电子科技大学 | A kind of readings of pointer type meters method based on deep learning |
CN110046617A (en) * | 2019-03-15 | 2019-07-23 | 西安交通大学 | A kind of digital electric meter reading self-adaptive identification method based on deep learning |
-
2019
- 2019-08-27 CN CN201910798558.1A patent/CN110427943A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190095739A1 (en) * | 2017-09-27 | 2019-03-28 | Harbin Institute Of Technology | Adaptive Auto Meter Detection Method based on Character Segmentation and Cascade Classifier |
CN108597204A (en) * | 2018-05-22 | 2018-09-28 | 广州市暨联牧科信息科技有限公司 | A kind of intelligent meter data recording system and its implementation |
CN109255375A (en) * | 2018-08-29 | 2019-01-22 | 长春博立电子科技有限公司 | Panoramic picture method for checking object based on deep learning |
CN109543682A (en) * | 2018-11-23 | 2019-03-29 | 电子科技大学 | A kind of readings of pointer type meters method based on deep learning |
CN110046617A (en) * | 2019-03-15 | 2019-07-23 | 西安交通大学 | A kind of digital electric meter reading self-adaptive identification method based on deep learning |
Non-Patent Citations (1)
Title |
---|
KK123K: "R-CNN论文详细解读", 《CSDN》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110909738A (en) * | 2019-11-15 | 2020-03-24 | 杭州远鉴信息科技有限公司 | Automatic reading method of pointer instrument based on key point detection |
CN111126395A (en) * | 2019-12-25 | 2020-05-08 | 福建天晴数码有限公司 | Optimization method of selective search algorithm in R-CNN network and storage medium |
CN113283419A (en) * | 2021-04-29 | 2021-08-20 | 国网浙江省电力有限公司湖州供电公司 | Convolutional neural network pointer instrument image reading identification method based on attention |
CN113283419B (en) * | 2021-04-29 | 2022-07-05 | 国网浙江省电力有限公司湖州供电公司 | Convolutional neural network pointer instrument image reading identification method based on attention |
CN113556318A (en) * | 2021-06-07 | 2021-10-26 | 广州康辰科技有限公司 | E-commerce verification method based on cloud security |
CN113556318B (en) * | 2021-06-07 | 2023-07-07 | 广西叫酒网络科技有限公司 | Electronic commerce verification method based on cloud security |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110427943A (en) | A kind of intelligent electric meter technique for partitioning based on R-CNN | |
CN107229930B (en) | Intelligent identification method for numerical value of pointer instrument | |
CN105373135B (en) | A kind of method and system of aircraft docking guidance and plane type recognition based on machine vision | |
CN111949809B (en) | Intelligent processing method for infrared inspection data of power transmission line | |
CN109583324A (en) | A kind of pointer meters reading automatic identifying method based on the more box detectors of single-point | |
CN109635806B (en) | Ammeter value identification method based on residual error network | |
CN107389701A (en) | A kind of PCB visual defects automatic checkout system and method based on image | |
CN110555475A (en) | few-sample target detection method based on semantic information fusion | |
CN108961330B (en) | Pig body length measuring and calculating method and system based on image | |
CN108918527A (en) | A kind of printed matter defect inspection method based on deep learning | |
CN104166841A (en) | Rapid detection identification method for specified pedestrian or vehicle in video monitoring network | |
CN109872317A (en) | A kind of defect identification method based on power equipments defect identification learning model | |
CN110472575A (en) | A kind of string tomato maturation detection method based on deep learning and computer vision | |
CN108932712A (en) | A kind of rotor windings quality detecting system and method | |
CN113869162A (en) | Violation identification method and system based on artificial intelligence | |
CN109344768A (en) | Pointer breaker recognition methods based on crusing robot | |
CN103745238A (en) | Pantograph identification method based on AdaBoost and active shape model | |
CN110728269B (en) | High-speed rail contact net support pole number plate identification method based on C2 detection data | |
CN114049325A (en) | Construction method and application of lightweight face mask wearing detection model | |
CN112132157A (en) | Raspberry pie-based gait face fusion recognition method | |
CN114066848B (en) | FPCA appearance defect visual detection system | |
CN109344766A (en) | Slide block type breaker recognition methods based on crusing robot | |
CN115019294A (en) | Pointer instrument reading identification method and system | |
CN116229052A (en) | Method for detecting state change of substation equipment based on twin network | |
CN112381175A (en) | Circuit board identification and analysis method based on image processing |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191108 |