CN108764134A - A kind of automatic positioning of polymorphic type instrument and recognition methods suitable for crusing robot - Google Patents
A kind of automatic positioning of polymorphic type instrument and recognition methods suitable for crusing robot Download PDFInfo
- Publication number
- CN108764134A CN108764134A CN201810523996.2A CN201810523996A CN108764134A CN 108764134 A CN108764134 A CN 108764134A CN 201810523996 A CN201810523996 A CN 201810523996A CN 108764134 A CN108764134 A CN 108764134A
- Authority
- CN
- China
- Prior art keywords
- instrument
- polymorphic type
- rectangle frame
- crusing robot
- automatic positioning
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of polymorphic type instrument automatic positioning suitable for crusing robot and recognition methods is labeled making training sample set according to all types of instrument scene pictures gathered in advance, the sample set training convolutional neural networks model that will be made;To arbitrary task scene picture, it is detected identification using trained convolutional neural networks instrument model, and post-process to detection recognition result;Final recognition result is finally obtained according to post-processing result.The present invention has high accuracy, high robust, to all having very high accuracy with identification comprising the polymorphic type instrument positioning including pointer gauge, digital instrument, status panel.
Description
Technical field
The present invention relates to image procossing and machine learning field more particularly to a kind of polymorphic types suitable for crusing robot
Instrument is automatically positioned and recognition methods.
Background technology
Instrument is widely used in the industries such as chemical industry, machinery, electric power as a kind of measuring instrument, such as the electric current in power grid
Table, voltmeter, thermometer etc..Traditional instrument positioning identifying method needs artificial naked eyes to judge, needs to spend a large amount of manpowers, labor
Fatigue resistance is big, while efficiency is extremely low, and process is very complicated, and in certain high-risk environments and is not suitable for manual work.
So human meter's inspection at present just gradually develops to intelligent inspection robot inspection, the task of intelligent inspection robot is logical
Autonomous information collection is crossed with processing to know the operating status of equipment, Informational support is provided for O&M maintenance.Automatic positioning and knowledge
Various types of instrument under other patrol task scene is to accurately identify the basis of various meter readings and state, to intelligence
The business application of crusing robot is of great significance.
Number of patent application is 201611031884.2, entitled《A kind of digital instrument reading image-recognizing method》China specially
Profit application, according to the digital instrument image demarcated in advance, area-of-interest is extracted using template matching method in panoramic picture,
Further according to single character zone and decimal point area to be tested in the relative position relation extraction area-of-interest of calibration character;It is right
Single character zone carries out single character recognition using the good convolutional neural networks character model of precondition;Decimal point is waited for
Detection zone utilizes the good Cascade target detections based on piecemeal LBP coding characteristics and Adaboost graders of precondition
Son carries out decimal point detection, and is post-processed to testing result;Finally obtained according to character, decimal point and sign recognition result
Read number.This method carries out accuracy when template matches dependent on the template demarcated in advance, if in task image irradiation
In the case of variation is apparent, the accuracy rate of positioning can substantially reduce, and then recognition accuracy declines to a great extent, in addition this method needle
Identification to digital instrument can not carry out the identification of polymorphic type instrument.
Number of patent application is 201610697650.5, entitled《A kind of gauge pointer image recognition based on symmetric characteristics
Method》Chinese patent application, symmetric characteristics in image are identified by extraction, and are further processed to obtain on this basis
The series of parameters information of identified pointer;In the area-of-interest of panorama sketch, to edge pixel in pointer travel and ginseng
The quantization for examining distance spatially carries out accumulative matrix ballot, obtains several groups finger candidate symmetry axis.Pass through pointer edge pixel
Point synteny characteristic and the corresponding image pixel value approximation consistency feature of pointer symmetry axis line segment, carry finger candidate
Refining so that final symmetrical pixels point belongs to pointer to being substantially all.Finger candidate merges the finger candidate of removal overlapping, selection
The most finger candidate of edge pixel point set is as final pointer recognition result.This method also depends on prior calibration well
Template carry out template matches when accuracy, can by illumination or the strong influence of shooting angle significantly change, and
Identify that object is single.
Existing instrument positioning is with recognition methods just for the positioning and identification that solve certain specific single kind instrument, example
Pointer class instrument can only be identified if pointer class instrument positioning identifying method, be unable to fixation and recognition numeric class instrument, application field
Limitation, cannot be satisfied the various instrument inspection demands under such as substation inspection scene.
Invention content
The purpose of the present invention is to provide a kind of polymorphic type instrument automatic positioning suitable for crusing robot and identification sides
Method can solve intelligent inspection robot and need the need for being accurately positioned with identifying various different instruments under such as substation's scene
It asks, and there is very high accuracy to the positioning and identification of various types instrument.
To achieve the above object, technical scheme is as follows:
A kind of automatic positioning of polymorphic type instrument and recognition methods suitable for crusing robot includes the following steps:
Step 1:Polymorphic type instrument scene picture is collected according to prior, marks the instrument region in scene picture and correspondence
Markup information file makes training sample set;
Step 2:Training sample set is sent into designed convolutional neural networks progress off-line model to train to obtain convolution god
Through network instrument model;
Step 3:Meter type identification is carried out to task scene picture using convolutional neural networks instrument model and position is examined
It surveys, tentatively obtains series instrument type, confidence level and position rectangle frame;
Step 4:The PRELIMINARY RESULTS obtained in step 3 is post-processed, finally obtains meter type in scene picture
And corresponding position.
In said program, in step 1, according to the polymorphic type instrument scene picture collected in advance, scene image is manually selected
The region of middle instrument is recorded the rectangle upper left corner and bottom right angular coordinate and meter type, is made with the boundary of rectangle frame gauged instrument
Training sample set.
In said program, in step 2 convolutional neural networks off-line model training include depth residual error network characterization extract,
Multi resolution feature extraction and deconvolution processing.
In said program, position rectangle frame is post-processed in step 4, is specifically included:
Detect the merging of rectangle frame:If any two rectangle frame RuAnd RvIf meetingAnd it indicates
To be same type of, then two rectangle frames are merged, combined rectangle frame is Ru∪Rv, τ is to merge threshold value;
The removal of pseudo- rectangle frame:If the arbitrarily confidence level C of detection recognition resultconf≤ δ, δ ∈ [0,1], then by the result
It rejects, δ is confidence threshold value.
In said program, in step 4, τ values 0.7, δ values 0.7.
The automatic positioning of polymorphic type instrument and recognition methods suitable for crusing robot of the present invention, has below beneficial to effect
Fruit:1, there is very high accuracy to various instrument, identify various meter types and meter location while can be accurate;2,
The interference such as it can overcome light variation, dial plate impurity, block, accurately identifying meter type, there is high robust;3, the present invention can
To greatly enhance the adaptability of instrument automation detection and identification device, manual work is liberated, improves efficiency.
Description of the drawings
Fig. 1 is pointer gauge sample mark figure of the present invention;
Fig. 2 is state table sample mark figure of the present invention;
Fig. 3 is digital table sample this mark figure of the invention;
Fig. 4 is the recognition result figure of task image pointer gauge of the present invention;
Fig. 5 is the recognition result figure of task image state table of the present invention;
Fig. 6 is the recognition result figure of task image number table of the present invention;
Fig. 7 is instrument sample training flow chart of the present invention;
Fig. 8 is task image Meter recognition flow chart of the present invention.
Specific implementation mode
Technical scheme of the present invention is described in further detail below in conjunction with the accompanying drawings.
A kind of automatic positioning of polymorphic type instrument and recognition methods suitable for crusing robot includes the following steps, such as Fig. 7
Shown in Fig. 8:
Step 1:According to the polymorphic type instrument scene picture collected in advance, instrument in mark scene image is manually selected
Region is recorded the upper left corner and the bottom right angular coordinate of rectangle, is respectively labeled as with the boundary up and down of rectangle frame gauged instrument
Xmin, ymin, xmax, ymax, the type of instrument are denoted as the title of corresponding instrument, by the picture of mark and corresponding markup information
File is fabricated to training sample set.
Step 2:Training sample set is sent into designed convolutional neural networks progress off-line model to train to obtain convolution god
Through network instrument model, wherein the training of convolutional neural networks off-line model includes the extraction of depth residual error network characterization, multiple dimensioned spy
Sign extraction and deconvolution processing.
Step 3:Task scene picture input convolutional neural networks instrument model is subjected to meter type identification and position inspection
It surveys, tentatively obtains series instrument type, confidence level and position rectangle frame;
Step 4:Preliminary recognition result is post-processed, merging overlapping area is larger and is the identification of same object
As a result, removal confidence level is less than the recognition result of threshold value, meter type and corresponding position in scene picture are finally obtained, wherein after
Processing specifically includes:
Detect the merging of rectangle frame:If any two rectangle frame RuAnd RvIf meetingAnd it indicates
To be same type of, then two rectangle frames are merged, combined rectangle frame is Ru∪Rv, τ is to merge threshold value, rule of thumb τ mono-
As value 0.7;
The removal of pseudo- rectangle frame:If the arbitrarily confidence level C of detection recognition resultconf≤ δ, δ ∈ [0,1], then by the result
It rejects, δ is confidence threshold value, rule of thumb the general values of δ 0.7.
By taking indoor instrument task scene picture in power grid as an example, the present invention provides a kind of suitable for the more of crusing robot
Type of meter is automatically positioned and recognition methods, including training sample set makes, training and task scene picture recognition three phases:
1, the sample production phase:The sample production phase is divided into data mark and training set makes.
Data include multiple classifications such as pointer ammeter, and such as pointer gauge, state table and digital table, it is on the scene to manually select instrument
Region in scape image, as shown in Figure 1, Figure 2 and Figure 3, with the boundary up and down of rectangle frame gauged instrument, the type of instrument is remembered
For the title of corresponding instrument, the upper left corner of rectangle and bottom right angular coordinate are denoted as xmin, ymin, xmax, ymax, these are believed respectively
Breath write-in file is recorded.
Training set makes:By the picture of mark and corresponding markup information file, it is divided into trained file and test text
Part makes the training sample set of LMDB formats.
2, the training stage:Off-line model training is carried out using convolutional neural networks, entire convolutional neural networks are divided into three
Part.
2.1 depth residual error networks (ResNet) carry out feature extraction:
Input layer:By instrument scene samples pictures using RGB triple channels as input layer data.
Convolutional layer 1:Use size for the convolution nuclear parameter of 7*7, the number of characteristic pattern is 64.
Pond layer 1:Use window size for the maximum pond of 3*3.
Followed by the main body of ResNet, it is made of, is respectively designated as 4 parts with residual error module:Conv2_x,
Conv3_x, conv4_x, conv5_x, meanwhile, there are one bn layers, scale layers and relu layers, bn layers after each convolutional layer:It will
Parameter normalization to (0,1];Scale layers:Change of scale;Relu layers, nonlinear transformation is carried out to convolution results.
conv2_x:It is made of 3 residual error block coupled in series, each residual error module is made of 3 convolutional layers, respectively:
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 64.
Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 64.
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 256.
conv3_x:It is made of 4 residual error block coupled in series, each residual error module is made of 3 convolutional layers, respectively:
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 128.
Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 128.
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 512.
conv4_x:It is made of 23 residual error block coupled in series, each residual error module is made of 3 convolutional layers, respectively:
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 256.
Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 256.
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 1024.
conv5_x:It is made of 3 residual error block coupled in series, each residual error module is made of 3 convolutional layers, respectively:
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 512.
Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 512.
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 2048.
2.2 multiple dimensioned characteristic extraction parts:
conv6:Size is the convolution nuclear parameter of 2*2, and characteristic pattern number is 1024.
Followed by 1 residual error module, the present invention is named as conv6_x, is made of three convolutional layers, respectively:
Use size for the convolution nuclear parameter of 2*2, the number of characteristic pattern is 256.
Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 256.
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 1024.
conv7:Size is the convolution nuclear parameter of 2*2, and the number of characteristic pattern is 1024.
Followed by 1 residual error module, the present invention is named as conv7_x, is made of three convolutional layers, respectively:
Use size for the convolution nuclear parameter of 2*2, the number of characteristic pattern is 256.
Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 256.
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 1024.
conv8:Size is the convolution nuclear parameter of 3*3, and characteristic pattern number is 1024.
Followed by 1 residual error module, the present invention is named as conv8_x, is made of three convolutional layers, respectively:
Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 256.
Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 256.
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 1024.
conv9:Size is the convolution nuclear parameter of 3*3, and characteristic pattern number is 1024.
Followed by 1 residual error module, the present invention is named as conv9_x, is made of three convolutional layers, respectively:
Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 256.
Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 256.
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 1024.
2.3 warp laminations:
Deconvolution is carried out for the result of Multi resolution feature extraction part, the size of characteristic pattern is made to return to conv6 inputs
Size includes mainly two modules:Prediction module and warp volume module.
Warp volume module:It is divided into 2 branches, common input is conv9.
Branch 1:
Warp lamination:Use size for the deconvolution nuclear parameter of 2*2, the number of characteristic pattern is 512.
Convolutional layer:Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 512.
Branch 2:
Convolutional layer:Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 512.
Convolutional layer:Use size for the convolution nuclear parameter of 3*3, the number of characteristic pattern is 512.
The result of branch 1 and branch 2 is obtained into characteristic pattern to the end into row element dot product
Warp volume module is called successively so that the size of characteristic pattern returns to the input size of conv6.
Prediction module:Input is deconvolution each time as a result, carrying out residual error to the characteristic pattern that deconvolution each time obtains
It calculates, is made of three convolutional layers, respectively:
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 256.
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 256.
Use size for the convolution nuclear parameter of 1*1, the number of characteristic pattern is 1024.
Concat layers:Input is prediction module as a result, output is characteristic pattern series connection result.
Output layer:Title, confidence level and the position coordinates of the corresponding instrument of output.
3, task image cognitive phase
The maximum difference of task scene picture and sample scene picture is that the change in location of instrument and the type of instrument become
Change, as shown in Figure 4, Figure 5 and Figure 6, it is shown that pointer gauge recognition result, state table recognition result and the number of task scene picture
Table recognition result, has outlined the position of pointer gauge, state table and digital table in figure, and is identified respectively in the upper left corner of rectangle frame
The corresponding classification of class instrument and confidence information.
Instrument is carried out in panoramic picture according to the good instrument model of precondition in the cognitive phase of task scene graph piece
Table detects and identification, if detection recognition failures, illustrates that instrument is not present;If detection identifies successfully, instrument in figure is marked
Position, classification and confidence information.
There may be the case where flase drop during detecting identification, it is therefore desirable to be post-processed, be specifically included:
(1) merging of rectangle frame is detected:In detecting identification process, may exist and " frame occur to same object identification
The phenomenon that center ", so needing to merge this, if any two rectangle frame RuAnd RvIf meeting
And it is denoted as same type of, then merges two rectangle frames, combined rectangle frame is Ru∪Rv, rule of thumb τ generally take
Value 0.7.
(2) removal of pseudo- rectangle frame:If the arbitrarily confidence level C of detection recognition resultconf≤ δ, δ ∈ [0,1], then by the knot
Fruit is rejected, rule of thumb the general values of δ 0.7.
Above-described specific implementation mode has carried out further the purpose of the present invention, technical solution and advantageous effect
It is described in detail, it should be understood that the foregoing is merely the specific implementation mode of the present invention, the guarantor being not intended to limit the present invention
Range is protected, all within the spirits and principles of the present invention, any modification, equivalent substitution, improvement and etc. done should all be contained in this hair
Within bright protection domain.
Claims (5)
1. a kind of automatic positioning of polymorphic type instrument and recognition methods suitable for crusing robot, it is characterised in that:Including following
Step:
Step 1:Polymorphic type instrument scene picture is collected according to prior, marks the instrument region in scene picture and corresponding mark
Message file makes training sample set;
Step 2:Training sample set is sent into designed convolutional neural networks progress off-line model to train to obtain convolutional Neural net
Network instrument model;
Step 3:Meter type identification and position detection are carried out to task scene picture using convolutional neural networks instrument model,
Tentatively obtain series instrument type, confidence level and position rectangle frame;
Step 4:The PRELIMINARY RESULTS obtained in step 3 is post-processed, finally obtains in scene picture meter type and right
Answer position.
2. the automatic positioning of polymorphic type instrument and recognition methods according to claim 1 suitable for crusing robot, special
Sign is:In the step 1, according to the polymorphic type instrument scene picture collected in advance, instrument in scene image is manually selected
Region is recorded the rectangle upper left corner and bottom right angular coordinate and meter type, training sample is made with the boundary of rectangle frame gauged instrument
Collection.
3. the automatic positioning of polymorphic type instrument and recognition methods according to claim 1 suitable for crusing robot, special
Sign is:The training of convolutional neural networks off-line model includes the extraction of depth residual error network characterization, multiple dimensioned spy in the step 2
Sign extraction and deconvolution processing.
4. the automatic positioning of polymorphic type instrument and recognition methods according to claim 1 suitable for crusing robot, special
Sign is:Position rectangle frame is post-processed in the step 4, is specifically included:
Detect the merging of rectangle frame:If any two rectangle frame RuAnd RvIf meetingAnd it is denoted as same
Type, then two rectangle frames are merged, combined rectangle frame is Ru∪Rv, τ is to merge threshold value;
The removal of pseudo- rectangle frame:If the arbitrarily confidence level C of detection recognition resultconf≤ δ, δ ∈ [0,1], then reject the result, δ
For confidence threshold value.
5. the automatic positioning of polymorphic type instrument and recognition methods according to claim 1 suitable for crusing robot, special
Sign is:τ values 0.7, δ values 0.7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810523996.2A CN108764134A (en) | 2018-05-28 | 2018-05-28 | A kind of automatic positioning of polymorphic type instrument and recognition methods suitable for crusing robot |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810523996.2A CN108764134A (en) | 2018-05-28 | 2018-05-28 | A kind of automatic positioning of polymorphic type instrument and recognition methods suitable for crusing robot |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108764134A true CN108764134A (en) | 2018-11-06 |
Family
ID=64003094
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810523996.2A Pending CN108764134A (en) | 2018-05-28 | 2018-05-28 | A kind of automatic positioning of polymorphic type instrument and recognition methods suitable for crusing robot |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108764134A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109635842A (en) * | 2018-11-14 | 2019-04-16 | 平安科技(深圳)有限公司 | A kind of image classification method, device and computer readable storage medium |
CN109784257A (en) * | 2019-01-08 | 2019-05-21 | 北京全路通信信号研究设计院集团有限公司 | Transformer thermometer detection and identification method |
CN109840497A (en) * | 2019-01-30 | 2019-06-04 | 华南理工大学 | A kind of pointer-type water meter reading detection method based on deep learning |
CN109858474A (en) * | 2019-01-08 | 2019-06-07 | 北京全路通信信号研究设计院集团有限公司 | Detection and identification method for transformer oil surface temperature controller |
CN110059617A (en) * | 2019-04-17 | 2019-07-26 | 北京易达图灵科技有限公司 | A kind of recognition methods of target object and device |
CN110097018A (en) * | 2019-05-08 | 2019-08-06 | 深圳供电局有限公司 | Transformer substation instrument detection method and device, computer equipment and storage medium |
CN111193860A (en) * | 2019-09-09 | 2020-05-22 | 浙江国自机器人技术有限公司 | One-frame calibration method for working point positions of inspection robot |
CN111738264A (en) * | 2020-05-08 | 2020-10-02 | 上海允登信息科技有限公司 | Intelligent acquisition method for data of display panel of machine room equipment |
CN113538370A (en) * | 2021-07-14 | 2021-10-22 | 宁波旗芯电子科技有限公司 | Power grid inspection method and device, computer equipment and storage medium |
CN117733881A (en) * | 2023-08-28 | 2024-03-22 | 广州信邦智能装备股份有限公司 | Configuration method of inspection robot, inspection robot and medium program |
Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104657748A (en) * | 2015-02-06 | 2015-05-27 | 中国石油大学(华东) | Vehicle type recognition method based on convolutional neural network |
CN105335710A (en) * | 2015-10-22 | 2016-02-17 | 合肥工业大学 | Fine vehicle model identification method based on multi-stage classifier |
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 |
CN106339707A (en) * | 2016-08-19 | 2017-01-18 | 亿嘉和科技股份有限公司 | Instrument pointer image recognition method based on symmetrical characteristics |
CN106530284A (en) * | 2016-10-21 | 2017-03-22 | 广州视源电子科技股份有限公司 | Welding spot type detection and device based on image recognition |
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 |
CN106599890A (en) * | 2015-10-14 | 2017-04-26 | 山东鲁能智能技术有限公司 | Transformer substation patrol robot digital type instrument identification algorithm |
CN107092863A (en) * | 2017-03-24 | 2017-08-25 | 重庆邮电大学 | A kind of readings of pointer type meters recognition methods of Intelligent Mobile Robot |
CN107491781A (en) * | 2017-07-21 | 2017-12-19 | 国家电网公司 | A kind of crusing robot visible ray and infrared sensor data fusion method |
CN107610097A (en) * | 2017-08-16 | 2018-01-19 | 深圳市天益智网科技有限公司 | Instrument localization method, device and terminal device |
CN107844794A (en) * | 2016-09-21 | 2018-03-27 | 北京旷视科技有限公司 | Image-recognizing method and device |
CN108009543A (en) * | 2017-11-29 | 2018-05-08 | 深圳市华尊科技股份有限公司 | A kind of licence plate recognition method and device |
CN108052946A (en) * | 2017-12-11 | 2018-05-18 | 国网上海市电力公司 | A kind of high pressure cabinet switch automatic identifying method based on convolutional neural networks |
-
2018
- 2018-05-28 CN CN201810523996.2A patent/CN108764134A/en active Pending
Patent Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN104657748A (en) * | 2015-02-06 | 2015-05-27 | 中国石油大学(华东) | Vehicle type recognition method based on convolutional neural network |
CN106599890A (en) * | 2015-10-14 | 2017-04-26 | 山东鲁能智能技术有限公司 | Transformer substation patrol robot digital type instrument identification algorithm |
CN105335710A (en) * | 2015-10-22 | 2016-02-17 | 合肥工业大学 | Fine vehicle model identification method based on multi-stage classifier |
CN106339707A (en) * | 2016-08-19 | 2017-01-18 | 亿嘉和科技股份有限公司 | Instrument pointer image recognition method based on symmetrical characteristics |
CN107844794A (en) * | 2016-09-21 | 2018-03-27 | 北京旷视科技有限公司 | Image-recognizing method and device |
CN106530284A (en) * | 2016-10-21 | 2017-03-22 | 广州视源电子科技股份有限公司 | Welding spot type detection and device based on image recognition |
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 |
CN107092863A (en) * | 2017-03-24 | 2017-08-25 | 重庆邮电大学 | A kind of readings of pointer type meters recognition methods of Intelligent Mobile Robot |
CN107491781A (en) * | 2017-07-21 | 2017-12-19 | 国家电网公司 | A kind of crusing robot visible ray and infrared sensor data fusion method |
CN107610097A (en) * | 2017-08-16 | 2018-01-19 | 深圳市天益智网科技有限公司 | Instrument localization method, device and terminal device |
CN108009543A (en) * | 2017-11-29 | 2018-05-08 | 深圳市华尊科技股份有限公司 | A kind of licence plate recognition method and device |
CN108052946A (en) * | 2017-12-11 | 2018-05-18 | 国网上海市电力公司 | A kind of high pressure cabinet switch automatic identifying method based on convolutional neural networks |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109635842A (en) * | 2018-11-14 | 2019-04-16 | 平安科技(深圳)有限公司 | A kind of image classification method, device and computer readable storage medium |
CN109784257B (en) * | 2019-01-08 | 2021-10-12 | 北京全路通信信号研究设计院集团有限公司 | Transformer thermometer detection and identification method |
CN109858474A (en) * | 2019-01-08 | 2019-06-07 | 北京全路通信信号研究设计院集团有限公司 | Detection and identification method for transformer oil surface temperature controller |
CN109784257A (en) * | 2019-01-08 | 2019-05-21 | 北京全路通信信号研究设计院集团有限公司 | Transformer thermometer detection and identification method |
CN109858474B (en) * | 2019-01-08 | 2021-10-19 | 北京全路通信信号研究设计院集团有限公司 | Detection and identification method for transformer oil surface temperature controller |
CN109840497A (en) * | 2019-01-30 | 2019-06-04 | 华南理工大学 | A kind of pointer-type water meter reading detection method based on deep learning |
CN110059617A (en) * | 2019-04-17 | 2019-07-26 | 北京易达图灵科技有限公司 | A kind of recognition methods of target object and device |
CN110097018A (en) * | 2019-05-08 | 2019-08-06 | 深圳供电局有限公司 | Transformer substation instrument detection method and device, computer equipment and storage medium |
CN111193860A (en) * | 2019-09-09 | 2020-05-22 | 浙江国自机器人技术有限公司 | One-frame calibration method for working point positions of inspection robot |
CN111193860B (en) * | 2019-09-09 | 2021-05-11 | 浙江国自机器人技术有限公司 | One-frame calibration method for working point positions of inspection robot |
CN111738264A (en) * | 2020-05-08 | 2020-10-02 | 上海允登信息科技有限公司 | Intelligent acquisition method for data of display panel of machine room equipment |
CN113538370A (en) * | 2021-07-14 | 2021-10-22 | 宁波旗芯电子科技有限公司 | Power grid inspection method and device, computer equipment and storage medium |
CN117733881A (en) * | 2023-08-28 | 2024-03-22 | 广州信邦智能装备股份有限公司 | Configuration method of inspection robot, inspection robot and medium program |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108764134A (en) | A kind of automatic positioning of polymorphic type instrument and recognition methods suitable for crusing robot | |
CN111612763B (en) | Mobile phone screen defect detection method, device and system, computer equipment and medium | |
CN103528617B (en) | A kind of cockpit instrument identifies and detection method and device automatically | |
CN106529537B (en) | A kind of digital instrument reading image-recognizing method | |
US20240153272A1 (en) | Method for detecting urine formed elements based on deep learning and context relation | |
CN109239102A (en) | A kind of flexible circuit board open defect detection method based on CNN | |
CN109711288A (en) | Remote sensing ship detecting method based on feature pyramid and distance restraint FCN | |
CN106909941A (en) | Multilist character recognition system and method based on machine vision | |
CN106546581A (en) | Detection paper card intelligent checking system and detection paper card intelligent analysis method | |
CN109583324A (en) | A kind of pointer meters reading automatic identifying method based on the more box detectors of single-point | |
CN105913093A (en) | Template matching method for character recognizing and processing | |
CN109856156A (en) | A kind of display panel tiny flaw determination method and device based on AOI | |
CN108830332A (en) | A kind of vision vehicle checking method and system | |
CN105574161B (en) | A kind of brand logo key element recognition methods, device and system | |
CN110796661B (en) | Fungal microscopic image segmentation detection method and system based on convolutional neural network | |
CN109978872B (en) | White matter microstructure characteristic screening system and method based on white matter fiber tracts | |
CN105223208B (en) | A kind of circuit board detecting template and preparation method thereof, circuit board detecting method | |
CN105466921A (en) | Simultaneous detection method of many samples | |
CN110348494A (en) | A kind of human motion recognition method based on binary channels residual error neural network | |
CN115184380A (en) | Printed circuit board welding spot abnormity detection method based on machine vision | |
CN107610119A (en) | The accurate detection method of steel strip surface defect decomposed based on histogram | |
CN115019294A (en) | Pointer instrument reading identification method and system | |
CN114266881A (en) | Pointer type instrument automatic reading method based on improved semantic segmentation network | |
CN114264648B (en) | Water quality detection test paper identification method and system | |
CN117197112A (en) | Industrial screen printing defect detection method and system based on sample comparison |
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: 20181106 |