CN106909941A - Multilist character recognition system and method based on machine vision - Google Patents
Multilist character recognition system and method based on machine vision Download PDFInfo
- Publication number
- CN106909941A CN106909941A CN201710107177.5A CN201710107177A CN106909941A CN 106909941 A CN106909941 A CN 106909941A CN 201710107177 A CN201710107177 A CN 201710107177A CN 106909941 A CN106909941 A CN 106909941A
- Authority
- CN
- China
- Prior art keywords
- image
- character
- unit
- ammeter
- template
- 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 37
- 230000011218 segmentation Effects 0.000 claims abstract description 24
- 238000012545 processing Methods 0.000 claims abstract description 22
- 238000003860 storage Methods 0.000 claims abstract description 17
- 238000000605 extraction Methods 0.000 claims abstract description 14
- 238000013528 artificial neural network Methods 0.000 claims abstract description 13
- 238000007781 pre-processing Methods 0.000 claims abstract description 12
- 238000010606 normalization Methods 0.000 claims abstract description 11
- 230000002708 enhancing effect Effects 0.000 claims abstract description 10
- 230000000877 morphologic effect Effects 0.000 claims abstract description 10
- 238000006243 chemical reaction Methods 0.000 claims abstract description 9
- 238000005520 cutting process Methods 0.000 claims abstract description 8
- 238000012937 correction Methods 0.000 claims description 8
- 238000001514 detection method Methods 0.000 claims description 5
- 238000002271 resection Methods 0.000 claims description 5
- 238000012549 training Methods 0.000 claims description 5
- 230000002457 bidirectional effect Effects 0.000 claims description 4
- 241001212149 Cathetus Species 0.000 claims description 3
- 239000006002 Pepper Substances 0.000 claims description 3
- 230000009286 beneficial effect Effects 0.000 claims description 3
- 230000008030 elimination Effects 0.000 claims description 3
- 238000003379 elimination reaction Methods 0.000 claims description 3
- 238000000926 separation method Methods 0.000 claims description 3
- 230000007797 corrosion Effects 0.000 claims description 2
- 238000005260 corrosion Methods 0.000 claims description 2
- 238000003702 image correction Methods 0.000 claims 1
- 150000003839 salts Chemical class 0.000 claims 1
- 238000004519 manufacturing process Methods 0.000 abstract description 5
- 241001269238 Data Species 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 5
- 210000004205 output neuron Anatomy 0.000 description 4
- 210000004027 cell Anatomy 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 244000144992 flock Species 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
-
- 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
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- 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
- 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
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/247—Aligning, centring, orientation detection or correction of the image by affine transforms, e.g. correction due to perspective effects; Quadrilaterals, e.g. trapezoids
-
- 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
Abstract
Multilist character recognition system and method the present invention relates to be based on machine vision, system include image capture module, network storage module and software processing module；Wherein, software processing module includes meter group cutting unit, image pre-processing unit, image tilt corrector unit, image enhancing unit, character zone positioning unit, Character segmentation unit, image normalization unit, character feature extraction unit and BP neural network recognition unit；Image pre-processing unit includes gray processing processing unit, binary conversion treatment unit and Morphological scale-space unit；The present invention can simultaneously recognize multiple ammeter images, greatly save human resources and working time, it is used to store ammeter image and ammeter identification data by using FTP high in the clouds, realize resource-sharing, and solve the great amount of images of production scene camera Real-time Collection and the storage problem of a large amount of ammeter datas, release PC memory, improves ammeter recognition efficiency.
Description
Technical field
The present invention relates to Digital Image Processing and instrument registration identification field, more particularly to the multilist based on machine vision
Character recognition system and method.
Background technology
The features such as digital displaying meter is due to its high precision, good stability, easy to install, small volume, at present by widely
It is applied in the industry-by-industries such as machine-building, oil, chemical industry, is especially widely used in industrial control field.General industry field
The data of monitoring can be needed during using digital displaying meter using multiple digital displaying meter centralized displayings, space resources had so not only been saved but also sharp
Creation data is monitored in real time in staff.
Although finding that the kiln temperature of Pearl River Delta area majority pot bank is monitored and controlled through visiting investigation installs a large amount of
Live indicating meter, but these instrument do not have transporting function, it is necessary to live reading, for monitoring industry spot operation shape
Condition.Although there is digital temperature control instrument, need for manually carrying out work of checking meter, periodically by data delivery in the form of reporting
It is analyzed to Production Supervisor.Control Room instrument is numerous, and manual metering is wasted time and energy poorly efficient, and easily causes erroneous judgement；And,
Common character recognition is all single table identification of the table of a figure one, and a video camera only shoots an ammeter, both took up room, and unrestrained
Take resource, be unfavorable for the long term growth of enterprise.
The content of the invention
It is an object of the invention to overcome the deficiencies in the prior art, there is provided one kind save manpower, high working efficiency, can be simultaneously
The multiple ammeters of identification, ammeter discrimination high, the great amount of images of effective solution production scene camera Real-time Collection and big electrometer
The storage problem of data, release PC memory, the multilist character recognition system based on machine vision for economizing on resources.
To achieve the above object, technical scheme provided by the present invention is：Multilist character recognition system based on machine vision
System, including image capture module, network storage module and software processing module；
Image capture module is the camera with mixed-media network modules mixed-media, many ammeter images of camera real-time capture；
Network storage module is ftp server, and many ammeter images that camera mixed-media network modules mixed-media will be captured are uploaded to FTP clothes
Business device；
Software processing module, many ammeter images to being downloaded to from network storage module are processed and recognized；
Wherein, software processing module include meter group cutting unit, image pre-processing unit, image tilt corrector unit,
Image enhancing unit, character zone positioning unit, Character segmentation unit, image normalization unit, character feature extraction unit with
And BP neural network recognition unit；
Meter group cutting unit, gathers the image of each ammeter working condition in meter group and preserves, as subsequent match
Template, global search is carried out in meter group target image using based on NCC template matching methods, find and template matches
Ammeter simultaneously splits；
Image tilt corrector unit, Canny rim detections are carried out to digital display instrument, in obtaining figure using Hough transform method
Straight line most long and its slope, calculate image inclination angle, and skew correction is carried out to it；
Image enhancing unit, is strengthened 20% pixel before ammeter and other pixels is suppressed；
Character zone positioning unit, the greatest gradient difference MGD from region determines character area more than the number of times statistics of threshold value
Domain；
Character segmentation unit, bidirectional projection is carried out to ammeter, and the starting and ending position of each character is oriented from left to right
Put, and cut out, the exact position of character is obtained by perpendicular cuts and horizontal resection；
Character feature extraction unit, describing character using the coarse grid of partial statistics characteristic carries out feature extraction；
BP neural network recognition unit, is recognized using the recognition methods based on BP neural network with on-line training function
Character.
Further, described image pretreatment unit includes gray processing processing unit, binary conversion treatment unit and form
Learn processing unit；
Wherein, gray processing processing unit, averagely more rational gray level image is obtained using being weighted to RGB three-components；
Morphological scale-space unit, the edge for the treatment of objective area in image, being adhered and salt-pepper noise between elimination character,
So that foreground image areas diminish, the background area inside foreground image is exaggerated；
Binary conversion treatment unit, binaryzation is realized using maximum variance between clusters Otsu, the target and image that will be recognized
Background make a distinction.
To achieve the above object, the present invention also provides a kind of side for the multilist character recognition system based on machine vision
Method, the method is comprised the following steps：
(1) IMAQ；
Will carry mixed-media network modules mixed-media held before meter group to be measured proper range, carry out the real-time of many ammeter images
Catch；
(2) image is uploaded；
Many ammeter images that camera mixed-media network modules mixed-media will be captured are uploaded to ftp server；
(3) software processing；
Many ammeter images to being downloaded to from network storage module are processed and recognized, finally by ammeter recognition result
Local PC given discs are saved in .txt text formattings, and are derived with Excel forms.
Further, video procession is comprised the following steps in step (3)：
1) meter group segmentation；
Live ammeter installs irregular, and the ammeter quantity that video camera once shoots is more and position disunity, thus need by
Every piece of ammeter Fast Segmentation is positioned and carries out Real time identification；Meter group is split by template matches；First in collection meter group
The image of each ammeter working condition is simultaneously preserved, as the template of subsequent match, using based on NCC template matching methods in ammeter
Global search is carried out in group target image, the ammeter with template matches is found and is split；
And NCC template matching methods are based on, its step is as follows：
(1 acquisition template pixel simultaneously calculates average and standard variance, pixel and average diff data samples；
(2 according to template size, and, moving window, calculates and often move a picture from top to bottom on target image from left to right
The NCC values of pixel and template pixel, compare with threshold value in the rear hatch of element, more than threshold value then record position；
(3, according to positional information is obtained, template matches recognition result are marked using red rectangle；
(4 systems show segmentation result.
NCC computing formula are：Wherein,
The gray value of pixel p is represented, μ represents window all pixels average value, and σ represents standard variance, and g represents template pixel value, m tables
Show the sum of all pixels of template, m-1 represents the free degree.
2) image preprocessing；
Preprocessing process includes gray processing, binaryzation and Morphological scale-space；Gray processing uses and RGB three-components is weighted
Averagely obtain more rational gray level image；Using corrosion etc. Morphological scale-space objective area in image edge, eliminate character it
Between be adhered and salt-pepper noise so that foreground image areas diminish, the background area inside foreground image be exaggerated, beneficial to character
Identification；Binaryzation is realized using maximum variance between clusters Otsu, the target and the background of image that will be recognized make a distinction；
3) image slant correction；
There is obvious straight line information after carrying out Canny rim detections to digital display instrument, in image, using Hough transform method
Straight line most long and its slope in figure are obtained, if straight line formula is:Y=kx+b, in formula：K is straight slope, b be straight line on the y axis
Intercept, the slope of its cathetus is defined as：The tangent value of straight line and rectangular coordinate system positive axis angular separation θ, its expression formula
For：K=arctan θ, calculate image inclination angle, and skew correction is carried out to it；
4) image enhaucament；
20% pixel before ammeter is strengthened and other pixels are suppressed, reached enhancing target character and suppress the back of the body
The purpose of scape, is easy to subsequent operation.
5) character zone positioning；
Because character zone and background contrasts are strong, its Grad is larger, and the greatest gradient difference MGD from region is more than
The number of times statistics of threshold value determines character zone；
6) Character segmentation；
Character zone is generally made up of multiple characters, it is therefore desirable to string segmentation for single character could be recognized, this
Scheme carries out bidirectional projection using sciagraphy to ammeter, and the starting and ending position of each character is oriented from left to right, and cuts
Cut out, the exact position of character is obtained by perpendicular cuts and horizontal resection；
7) image normalization；
Feature extraction for convenience, the image that the image normalization split to size is 15 × 21.
8) character feature is extracted；
Describing character using the coarse grid of partial statistics characteristic carries out feature extraction；
9) BP neural network identification；
Character is recognized using the recognition methods based on BP neural network with on-line training function.
This programme principle is as follows：
Camera real-time capture meter group image, and image is uploaded to by FTP by the mixed-media network modules mixed-media on camera
Server；When identification is needed, many ammeter images are downloaded to local PC from FTP high in the clouds, then application program is carried out to image
Meter group segmentation, image preprocessing, image slant correction, image enhaucament, character zone positioning, Character segmentation, image normalization,
Character feature is extracted, BP neural network recognizes a series of identifying processing, finally by ammeter recognition result with .txt text formattings
Local PC given discs are saved in, and are derived with Excel forms.
Compared with prior art, this programme has advantages below and beneficial effect：
Multiple ammeter images can be simultaneously recognized, human resources and working time is greatlyd save, is used to by using FTP high in the clouds
Storage ammeter image and ammeter identification data, realize resource-sharing, and solve the big of production scene camera Real-time Collection
The storage problem of spirogram picture and a large amount of ammeter datas, releases PC memory, improves ammeter recognition efficiency.
Brief description of the drawings
Fig. 1 is the structural representation of present system；
Fig. 2 is many ammeter character recognizing process figures of the invention；
Fig. 3 is Character segmentation course projection schematic diagram in the present invention；
Fig. 4 is grid search-engine extraction schematic diagram in the present invention；
Fig. 5 is three layers of BP network structures in the present invention.
Specific embodiment
With reference to specific embodiment, the invention will be further described：
Referring to shown in accompanying drawing 1, the multilist character recognition system based on machine vision described in the present embodiment, including image is adopted
Collection module 1, network storage module 2 and software processing module 3；Wherein, software processing module 3 includes meter group cutting unit 3-
1st, image pre-processing unit 3-2, image tilt corrector unit 3-3, image enhancing unit 3-4, character zone positioning unit 3-5,
Character segmentation unit 3-6, image normalization unit 3-7, character feature extraction unit 3-8 and BP neural network recognition unit；
Image pre-processing unit 3-2 includes gray processing processing unit 3-2-1, binary conversion treatment unit 3-2-2 and Morphological scale-space list
First 3-2-3.
During work, flow is as shown in Fig. 2 image capture module 1 is the camera with 4G mixed-media network modules mixed-medias, and real-time capture is more
Ammeter image, and many ammeter images that will be captured are uploaded to network storage module 2FTP servers, it is necessary to when recognizing, from FTP
Server downloads many ammeter images to software processing module 3, and specific treatment identification step is as follows：
Meter group cutting unit 3-1 treatment：
Meter group is split by template matches；First gather the image of each ammeter working condition in meter group and protect
Deposit, as the template of subsequent match, global search carried out in meter group target image using based on NCC template matching methods,
Find the ammeter with template matches and split；
NCC spans are [- 1,1], and target image is that each pixel of meter group image can be regarded as RGB numbers
Value, equivalent to a set for sample data, subset of the template image equivalent to it.When in template and target image another
It is 1 that sample data is mutually matched then its NCC values, represents that correlation is very high, if -1 represents completely uncorrelated, based on this
Individual principle, realizes that image is based on algorithms of template matching recognition.First to data normalization, formula is：Wherein f tables
Show the gray value of pixel p, μ represents window all pixels average value, and σ represents standard variance, it is assumed that g represents template pixel value,
ThenWherein m represents the sum of all pixels of template, and m-1 is the free degree.
Based on NCC template matching methods, its step is as follows：
(1 acquisition template pixel simultaneously calculates average and standard variance, pixel and average diff data samples；
(2 according to template size, and, moving window, calculates and often move a picture from top to bottom on target image from left to right
The NCC values of pixel and template pixel, compare with threshold value in the rear hatch of element, more than threshold value then record position；
(3, according to positional information is obtained, template matches recognition result are marked using red rectangle；
(4 systems show segmentation result.
Gray processing processing unit 3-2-1 treatment：
Image gray processing treatment, using weighted mean method：Image is converted into HSI color spaces from RGB color,
I is exactly the gray value of image in HSI spaces, and H and S show respectively the tone and saturation degree of image, and RGB three-components are mapped to H
The formula of component：
The formula of saturation degree component is：
Strength component is：
Three components are weighted with different weights, the conversion from coloured image to gray-scale map is obtained, formula is：I
=030R+0.59G+0.11B.
Binary conversion treatment unit 3-2-2 treatment：
Using maximum variance between clusters, maximum between-cluster variance can select suitable threshold value to image according to the gray value of image
Binary conversion treatment is carried out, image pixel is divided into the class of target and background two, defines the difference that inter-class variance g describes target and background,
For given image, there is suitable threshold value to maximize the inter-class variance g of the image, the threshold value is just the optimal threshold of binaryzation
Value.Wherein ω0And ω1The respectively ratio of foreground and background, μ0And μ1The respectively average of foreground and background pixel, μ is whole
The pixel average of image, g is the maximum between-cluster variance of requirement.
μ=ω0×μ0+ω1×μ1
G=ω0(μ0-μ)2+ω1(μ1-μ)2
G=ω0ω1(μ0-μ1)2
Morphological scale-space unit 3-2-3 treatment：
The edge of objective area in image is processed using etch state so that foreground image areas diminish, foreground image
Internal background area is exaggerated.
Image tilt corrector unit 3-3 treatment：
There is obvious straight line information after carrying out Canny rim detections to digital display instrument, in image, using Hough transform method
Straight line most long and its slope in figure are obtained, if straight line formula is:Y=kx+b, in formula：K is straight slope, b be straight line on the y axis
Intercept, the slope of its cathetus is defined as：The tangent value of straight line and rectangular coordinate system positive axis angular separation θ, its expression formula
For：K=arctan θ, calculate image inclination angle, and skew correction is carried out to it；
Image enhancing unit 3-4 treatment：
20% pixel before ammeter is strengthened and other pixels are suppressed, reached enhancing target character and suppress the back of the body
The purpose of scape, step is as follows：
The maximum of pixel, then carries out descending sort, 20% after sequence to pixel value used first in statistics picture
The value level of that point, finally strengthens every bit in image.
Character zone positioning unit 3-5 treatment：
Enter the Grad Gin of character and from character into the Grad Gout symbols of background just conversely, ladder from background
The absolute value of degree is close to equal.In a certain region of neighbouring selection of character, greatest gradient value Gmax and the minimum ladder in region are tried to achieve
Angle value Gmin, then the greatest gradient difference MGD in the region, as：
MGD=Gmax-Gmin
One threshold value of setting, MGD is probably character row or column more than threshold value.Due to the presence of noise jamming, non-character
Image there is also situations of indivedual MGD more than threshold value sometimes.Therefore, character area is determined more than the number of times statistics of threshold value from MGD
Domain, the method can be accurately positioned to character zone.
Character segmentation unit 3-6 treatment：
Because character zone color pixel cell is more, flock together more；There is a fixed gap between character and character, these
Space is the background pixel that pixel value is 0 mostly.According to this feature, upright projection is done to ammeter field area；Numeric area pair
The projection answered has obvious peak value, and vertical segmentation is realized as Rule of judgment；Level throwing is done to each numeral after vertical segmentation
Shadow, according to floor projection to realizing the horizontal segmentation of individual digit.By just having obtained character after perpendicular cuts and horizontal resection
Exact position.
Principle is as follows：
As shown in figure 3, image function is f (x, y), a is projecting direction, and b is perpendicular direction, and b is pressed from both sides with x-axis for projection
Angle is θ, then f (x, y) is along the definition of a：
As fixed θ, p (b, θ) is the function of b, is an one-dimensional waveform.Change θ, can obtain on different directions f (x,
Y) projection.Projection in x-axis and y-axis is defined as Px, Py：
With x-axis as axis of projection, the gray value summation of all picture points in statistics and x-axis orthogonal direction.Projection histogram can
Intuitively to reflect very much space distribution situation of the picture point relative to axial direction.It is bianry image f (x, y) of M*N for size,
Projective distribution function HX (x) of x-axis is：
Due to there is black picture element at where each character, and have certain without black picture element area between character.As long as taking x
Axle as projection transverse axis, upright projection, using the black picture element number that projects as ordinate；Again using y-axis as projection transverse axis, water
Flat projection, the black picture element number of projection is used as abscissa.Perspective view can be obtained by by the method, be further according to perspective view
Character segmentation can be come.
Image normalization unit 3-7 treatment：
Segmentation figure picture is normalized to 15 × 21；
Calculate the ratio for needing scaling：
Calculate the size after image scaling：
W=W × scale
H=H × scale
Image after scaling is placed on centre, lower or so zero padding is to 15 × 21 on this image；
Character feature extraction unit 3-8 treatment：
Feature extraction is carried out using the coarse grid of local feature statistical nature.Method is：After being normalized with OTSU methods
Digital picture binaryzation, then divides the image into several local cell domains, and the character pixels density on each zonule is made
It is Expressive Features, that is, counts the percentage shared by image pixel in each zonule.It is 15 × 21 image for size, divides
It is multiple 3 × 5 zonule, totally 15.The number for trying to achieve black picture element in each lattice respectively accounts for total in this lattice
The percentage of pixel, forms the matrix of 3 × 5, the matrix that every row element is converted into 1 × 15 is taken successively, as neutral net
Input data.
Numeral 0 to 9 is processed successively, as the feature of character recognition, as shown in Figure 4.
BP neural network recognition unit 3-9 treatment：
Set up three-layer network, because input is each digital feature for extracting, have 15 data, output it is corresponding
It is 0 to 9 totally ten numerals, so input layer and output layer neuron node number are respectively 15 and 10.It is public using general experience
Formula(hide represents hidden layer number, and m and n is respectively input layer and output layer nodes, and a takes 1-10's
Constant) determine the hidden node number of plies, determine that the final hidden node number of plies is 15 by test of many times：By verification experimental verification,
When hidden layer reaches 15 nodes, the training of system and error all reach preferable situation, so the input layer of network, hidden
15,15 and 10 are respectively containing layer and output layer neuron node number.Taken O to 9 totally ten numerals as number to be identified, per number
Word takes 10 samples and is trained, and has 100 training samples, separately takes 10 samples as identification sample.When input picture 0~
It is 1 on the corresponding position of output neuron after 9, other positions are O；Input numeral 0, first output neuron is l,
Other are 0；Input numeral 1, second output neuron is 1, and other are O：The rest may be inferred.During identification, if certain output god
Weights through unit are more than or equal to 0.8, then it is assumed that the digital picture is corresponding numeral.If all of the weights of output neuron
Both less than 0.8, then system think None- identified, as shown in Figure 5.
Ammeter recognition result is finally saved in local PC given discs with .txt text formattings, and with Excel forms
Derive.
The present embodiment is manipulated to meter group treatment identification by a series of, once can simultaneously recognize 20-30 ammeter.Use
Ftp server stores ammeter image and ammeter identification data, realizes solving production scene camera reality while resource-sharing
When the big electrometer image that gathers and a large amount of ammeter datas storage problem, release PC memory, improve ammeter recognition efficiency.
By on-the-spot test, up to 98%, the recognition time of average each image is within 10ms for the discrimination of character.The system has
Recognition speed is fast, strong antijamming capability, accuracy advantage high, on the premise of recognition correct rate is ensured, substantially increases effect
Rate, the time is only the 1/10 of equivalent amount list table identification.Also, the multilist identification that the system is irregularly placed for different model
It is equally applicable.
The examples of implementation of the above are only the preferred embodiments of the invention, not limit implementation model of the invention with this
Enclose, therefore the change that all shapes according to the present invention, principle are made, all should cover within the scope of the present invention.
Claims (7)
1. based on machine vision multilist character recognition system, including image capture module (1), network storage module (2) and
Software processing module (3)；
Image capture module (1), many ammeter images of real-time capture；
Network storage module (2), many ammeter images that storage is captured；
Software processing module (3), many ammeter images to being downloaded to from network storage module (2) are processed and recognized；
Wherein, software processing module (3) is inclined including meter group cutting unit (3-1), image pre-processing unit (3-2), image
Correction unit (3-3), image enhancing unit (3-4), character zone positioning unit (3-5), Character segmentation unit (3-6), image
Normalization unit (3-7), character feature extraction unit (3-8) and BP neural network recognition unit (3-9)；
Meter group cutting unit (3-1), gathers the image of each ammeter working condition in meter group and preserves, as subsequent match
Template, global search is carried out in meter group target image using based on NCC template matching methods, find and template matches
Ammeter simultaneously splits；
Image tilt corrector unit (3-3), Canny rim detections are carried out to digital display instrument, and figure is obtained using Hough transform method
In straight line most long and its slope, calculate image inclination angle, skew correction is carried out to it；
Image enhancing unit (3-4), is strengthened 20% pixel before ammeter and other pixels is suppressed；
Character zone positioning unit (3-5), the greatest gradient difference MGD from region determines character area more than the number of times statistics of threshold value
Domain；
Character segmentation unit (3-6), bidirectional projection is carried out to ammeter, and the starting and ending position of each character is oriented from left to right
Put, and cut out, the exact position of character is obtained by perpendicular cuts and horizontal resection；
Character feature extraction unit (3-8), describing character using the coarse grid of partial statistics characteristic carries out feature extraction；
BP neural network recognition unit (3-9), is known using the recognition methods based on BP neural network with on-line training function
Malapropism is accorded with.
2. the multilist character recognition system based on machine vision according to claim 1, it is characterised in that：Described image is pre-
Processing unit (3-2) includes gray processing processing unit (3-2-1), binary conversion treatment unit (3-2-2) and Morphological scale-space list
First (3-2-3)；
Wherein, gray processing processing unit (3-2-1), averagely more rational gray-scale map is obtained using being weighted to RGB three-components
Picture；
Morphological scale-space unit (3-2-3), processes the edge of objective area in image, and being adhered between elimination character is made an uproar with the spiced salt
Sound so that foreground image areas diminish, the background area inside foreground image is exaggerated；
Binary conversion treatment unit (3-2-2), binaryzation is realized using maximum variance between clusters Otsu, the target and figure that will be recognized
The background of picture makes a distinction.
3. it is a kind of for described in claim 1 based on machine vision multilist character recognition system method, it is characterised in that：Bag
Include following steps：
(1) IMAQ；
Will carry mixed-media network modules mixed-media held before meter group to be measured proper range, carry out many ammeter images in real time catch
Catch；
(2) image is uploaded；
Many ammeter images that camera mixed-media network modules mixed-media will be captured are uploaded to ftp server；
(3) software processing；
Many ammeter images to being downloaded to from network storage module are processed and recognized, finally by ammeter recognition result with .txt
Text formatting is saved in local PC given discs, and is derived with Excel forms.
4. the method for the multilist character recognition system based on machine vision according to claim 3, it is characterised in that：
Video procession is comprised the following steps in the step (3)：
1) meter group segmentation；
Meter group is split by template matches；First gather the image of each ammeter working condition in meter group and preserve, make
Be the template of subsequent match, global search carried out in meter group target image using based on NCC template matching methods, find with
The ammeter of template matches simultaneously splits；
2) image preprocessing；
Preprocessing process includes gray processing, binaryzation and Morphological scale-space；Gray processing uses and RGB three-components is weighted averagely
Obtain more rational gray level image；Using the edge of the Morphological scale-space objective area in image such as corrosion, between elimination character
It is adhered and salt-pepper noise so that foreground image areas diminish, the background area inside foreground image is exaggerated, beneficial to the knowledge of character
Not；Binaryzation is realized using maximum variance between clusters Otsu, the target and the background of image that will be recognized make a distinction；
3) image slant correction；
There is obvious straight line information after carrying out Canny rim detections to digital display instrument, in image, obtained using Hough transform method
Straight line most long and its slope in figure, if straight line formula is:Y=kx+b, in formula：K is straight slope, and b is straight line cutting on the y axis
Away from the slope of its cathetus is defined as：The tangent value of straight line and rectangular coordinate system positive axis angular separation θ, its expression formula is：k
=arctan θ, calculate image inclination angle, and skew correction is carried out to it；
4) image enhaucament；
20% pixel before ammeter is strengthened and other pixels are suppressed, reached enhancing target character and suppress background
Purpose；
5) character zone positioning；
Greatest gradient difference MGD from region determines character zone more than the number of times statistics of threshold value；
6) Character segmentation；
Bidirectional projection is carried out to ammeter, the starting and ending position of each character is oriented from left to right, and cut out, passed through
Perpendicular cuts and horizontal resection obtain the exact position of character；
7) image normalization；
8) character feature is extracted；
Describing character using the coarse grid of partial statistics characteristic carries out feature extraction；
9) BP neural network identification.
5. the method for the multilist character recognition system based on machine vision according to claim 4, it is characterised in that：
It is described based on NCC template matching methods, its step is as follows：
(1 acquisition template pixel simultaneously calculates average and standard variance, pixel and average diff data samples；
(2 according to template size, on target image from left to right, moving window from top to bottom, calculate often move a pixel it
The NCC values of pixel and template pixel, compare with threshold value in rear hatch, more than threshold value then record position；
(3, according to positional information is obtained, template matches recognition result are marked using red rectangle；
(4 systems show segmentation result.
6. the method for the multilist character recognition system based on machine vision according to claim 5, it is characterised in that：
The NCC computing formula are：
Wherein, data normalization formula isf
The gray value of pixel p is represented, μ represents window all pixels average value, and σ represents standard variance, and g represents template pixel value, m tables
Show the sum of all pixels of template, m-1 represents the free degree.
7. the method for the multilist character recognition system based on machine vision according to claim 4, it is characterised in that：
The character feature is extracted, digital picture binaryzation after being normalized with OTSU methods, then divides the image into several parts small
Region, counts the percentage shared by image pixel in each zonule, using the character pixels density on each zonule as retouching
State feature.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710107177.5A CN106909941A (en) | 2017-02-27 | 2017-02-27 | Multilist character recognition system and method based on machine vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710107177.5A CN106909941A (en) | 2017-02-27 | 2017-02-27 | Multilist character recognition system and method based on machine vision |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106909941A true CN106909941A (en) | 2017-06-30 |
Family
ID=59208786
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710107177.5A Pending CN106909941A (en) | 2017-02-27 | 2017-02-27 | Multilist character recognition system and method based on machine vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106909941A (en) |
Cited By (41)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107507130A (en) * | 2017-08-01 | 2017-12-22 | 江苏理工学院 | A kind of quickly QFN chip pins image obtains and amplification method |
| CN107563374A (en) * | 2017-08-14 | 2018-01-09 | 华中科技大学 | A kind of digital display instrument charactron localization method |
| CN107832763A (en) * | 2017-11-07 | 2018-03-23 | 青岛理工大学 | Graphics and Image Processing and equipment |
| CN107832767A (en) * | 2017-11-14 | 2018-03-23 | 深圳码隆科技有限公司 | Container number identification method, device and electronic equipment |
| CN107862303A (en) * | 2017-11-30 | 2018-03-30 | 平安科技（深圳）有限公司 | Information identifying method, electronic installation and the readable storage medium storing program for executing of form class diagram picture |
| CN108009538A (en) * | 2017-12-22 | 2018-05-08 | 大连运明自动化技术有限公司 | A kind of automobile engine cylinder-body sequence number intelligent identification Method |
| CN108038482A (en) * | 2017-12-22 | 2018-05-15 | 大连运明自动化技术有限公司 | A kind of automobile engine cylinder-body sequence number Visual intelligent identifying system |
| CN108416355A (en) * | 2018-03-09 | 2018-08-17 | 浙江大学 | A kind of acquisition method of the industry spot creation data based on machine vision |
| CN108427959A (en) * | 2018-02-07 | 2018-08-21 | 北京工业大数据创新中心有限公司 | Board state collection method based on image recognition and system |
| CN108573261A (en) * | 2018-04-17 | 2018-09-25 | 国家电网公司 | A kind of read out instrument recognition methods suitable for Intelligent Mobile Robot |
| CN108875739A (en) * | 2018-06-13 | 2018-11-23 | 深圳市云识科技有限公司 | A kind of accurate detecting method of digital displaying meter reading |
| CN109145912A (en) * | 2018-07-09 | 2019-01-04 | 华南理工大学 | A kind of digital instrument reading automatic identifying method |
| CN109360396A (en) * | 2018-09-27 | 2019-02-19 | 长江大学 | Remote meter reading method and system based on image recognition technology and NB-IoT technology |
| CN109357694A (en) * | 2018-08-22 | 2019-02-19 | 安徽慧视金瞳科技有限公司 | A kind of instrument digital detection method |
| CN109373896A (en) * | 2018-10-19 | 2019-02-22 | 东莞复熵智能科技有限公司 | A kind of NI Vision Builder for Automated Inspection and its implementation of one direction measurement displacement |
| CN109382968A (en) * | 2018-10-23 | 2019-02-26 | 宁波汇智恒动自动化科技有限公司 | A kind of vision-based detection and correction system for injection molding machine |
| CN109858480A (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 | A kind of digital instrument recognition methods |
| CN110097048A (en) * | 2019-04-02 | 2019-08-06 | 江苏理工学院 | A kind of SOT chip image quickly corrects and character identifying method |
| CN110119648A (en) * | 2018-02-05 | 2019-08-13 | 国家计算机网络与信息安全管理中心 | A kind of facsimile signal classification method based on optical character identification |
| CN110210477A (en) * | 2019-05-24 | 2019-09-06 | 四川阿泰因机器人智能装备有限公司 | A kind of digital instrument Recognition of Reading method |
| CN110232376A (en) * | 2019-06-11 | 2019-09-13 | 重庆邮电大学 | A kind of gear type digital instrument recognition methods returned using projection |
| CN110309831A (en) * | 2019-06-10 | 2019-10-08 | 北京工业大学 | Non intelligent water meter reading method based on machine vision |
| CN110716513A (en) * | 2019-09-17 | 2020-01-21 | 国网河北能源技术服务有限公司 | Indirect analog quantity original data acquisition method based on EXCEL |
| CN110725711A (en) * | 2019-10-29 | 2020-01-24 | 南京北路自动化系统有限责任公司 | Drilling system based on video and auxiliary drilling inspection method |
| CN110781884A (en) * | 2019-10-17 | 2020-02-11 | 广西电网有限责任公司电力科学研究院 | Method for realizing intelligent reading of electric meter data |
| CN111031222A (en) * | 2019-12-27 | 2020-04-17 | 山东厚德测控技术股份有限公司 | Real-time recognition device and method for character wheel of camera type gas meter |
| CN111027546A (en) * | 2019-12-05 | 2020-04-17 | 北京嘉楠捷思信息技术有限公司 | Character segmentation method and device and computer readable storage medium |
| CN111017728A (en) * | 2019-12-23 | 2020-04-17 | 安徽工业大学 | Crane brake health state assessment method based on machine vision |
| CN111242197A (en) * | 2020-01-07 | 2020-06-05 | 中国石油大学(华东) | Image and text matching method based on double-view-domain semantic reasoning network |
| CN111695576A (en) * | 2020-06-09 | 2020-09-22 | 国网河北省电力有限公司营销服务中心 | Electric energy meter appearance rapid identification method based on neural network |
| CN111860317A (en) * | 2020-07-20 | 2020-10-30 | 青岛特利尔环保集团股份有限公司 | Boiler operation data acquisition method, system, equipment and computer medium |
| CN112101336A (en) * | 2020-09-09 | 2020-12-18 | 杭州测质成科技有限公司 | Intelligent data acquisition mode based on computer vision |
| CN112163546A (en) * | 2020-10-13 | 2021-01-01 | 珠海格力电器股份有限公司 | Water meter value determination method and device based on visual identification system |
| CN112418226A (en) * | 2020-10-23 | 2021-02-26 | 济南信通达电气科技有限公司 | Method and device for identifying opening and closing states of fisheyes |
| CN112464928A (en) * | 2020-11-27 | 2021-03-09 | 广东电网有限责任公司 | Digital meter reading identification method, device, equipment and storage medium |
| CN112966746A (en) * | 2020-11-20 | 2021-06-15 | 扬州大学 | Stable variable gray template generation method suitable for tire defect detection |
| CN113591705A (en) * | 2021-07-30 | 2021-11-02 | 华东理工大学 | Inspection robot instrument recognition system and method and storage medium |
| CN113609906A (en) * | 2021-06-30 | 2021-11-05 | 南京信息工程大学 | Document-oriented table information extraction method |
| CN114445784A (en) * | 2022-03-01 | 2022-05-06 | 武汉聚智惠仁信息技术有限公司 | Method and system for acquiring CRRT screen parameters in real time |
| CN116721042A (en) * | 2023-08-10 | 2023-09-08 | 广东石油化工学院 | Multi-threshold binarization-based image tilt correction method |
| CN116721094A (en) * | 2023-08-03 | 2023-09-08 | 济南邦德激光股份有限公司 | Visual pattern recognition method and system under complex background and laser cutting machine |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103984930A (en) * | 2014-05-21 | 2014-08-13 | 南京航空航天大学 | Digital meter recognition system and method based on vision |
-
2017
- 2017-02-27 CN CN201710107177.5A patent/CN106909941A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103984930A (en) * | 2014-05-21 | 2014-08-13 | 南京航空航天大学 | Digital meter recognition system and method based on vision |
Non-Patent Citations (3)
| Title |
|---|
| 张文杰 等: "基于模板匹配的多目标精确粘连分割算法", 《计算机与现代化》 * |
| 李貌: "仪器仪表数字图像识别及其应用研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
| 林文娅: "基于神经网络的水表字符识别", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
Cited By (50)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107507130A (en) * | 2017-08-01 | 2017-12-22 | 江苏理工学院 | A kind of quickly QFN chip pins image obtains and amplification method |
| CN107563374A (en) * | 2017-08-14 | 2018-01-09 | 华中科技大学 | A kind of digital display instrument charactron localization method |
| CN107832763A (en) * | 2017-11-07 | 2018-03-23 | 青岛理工大学 | Graphics and Image Processing and equipment |
| CN107832767A (en) * | 2017-11-14 | 2018-03-23 | 深圳码隆科技有限公司 | Container number identification method, device and electronic equipment |
| CN107862303B (en) * | 2017-11-30 | 2019-04-26 | 平安科技（深圳）有限公司 | Information identifying method, electronic device and the readable storage medium storing program for executing of form class diagram picture |
| CN107862303A (en) * | 2017-11-30 | 2018-03-30 | 平安科技（深圳）有限公司 | Information identifying method, electronic installation and the readable storage medium storing program for executing of form class diagram picture |
| CN108038482A (en) * | 2017-12-22 | 2018-05-15 | 大连运明自动化技术有限公司 | A kind of automobile engine cylinder-body sequence number Visual intelligent identifying system |
| CN108009538A (en) * | 2017-12-22 | 2018-05-08 | 大连运明自动化技术有限公司 | A kind of automobile engine cylinder-body sequence number intelligent identification Method |
| CN110119648A (en) * | 2018-02-05 | 2019-08-13 | 国家计算机网络与信息安全管理中心 | A kind of facsimile signal classification method based on optical character identification |
| CN108427959A (en) * | 2018-02-07 | 2018-08-21 | 北京工业大数据创新中心有限公司 | Board state collection method based on image recognition and system |
| CN108416355A (en) * | 2018-03-09 | 2018-08-17 | 浙江大学 | A kind of acquisition method of the industry spot creation data based on machine vision |
| CN108573261A (en) * | 2018-04-17 | 2018-09-25 | 国家电网公司 | A kind of read out instrument recognition methods suitable for Intelligent Mobile Robot |
| CN108875739A (en) * | 2018-06-13 | 2018-11-23 | 深圳市云识科技有限公司 | A kind of accurate detecting method of digital displaying meter reading |
| CN109145912A (en) * | 2018-07-09 | 2019-01-04 | 华南理工大学 | A kind of digital instrument reading automatic identifying method |
| CN109357694A (en) * | 2018-08-22 | 2019-02-19 | 安徽慧视金瞳科技有限公司 | A kind of instrument digital detection method |
| CN109360396A (en) * | 2018-09-27 | 2019-02-19 | 长江大学 | Remote meter reading method and system based on image recognition technology and NB-IoT technology |
| CN109373896A (en) * | 2018-10-19 | 2019-02-22 | 东莞复熵智能科技有限公司 | A kind of NI Vision Builder for Automated Inspection and its implementation of one direction measurement displacement |
| CN109382968A (en) * | 2018-10-23 | 2019-02-26 | 宁波汇智恒动自动化科技有限公司 | A kind of vision-based detection and correction system for injection molding machine |
| CN109858480A (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 | A kind of digital instrument recognition methods |
| CN110097048A (en) * | 2019-04-02 | 2019-08-06 | 江苏理工学院 | A kind of SOT chip image quickly corrects and character identifying method |
| CN110210477A (en) * | 2019-05-24 | 2019-09-06 | 四川阿泰因机器人智能装备有限公司 | A kind of digital instrument Recognition of Reading method |
| CN110210477B (en) * | 2019-05-24 | 2023-03-24 | 四川阿泰因机器人智能装备有限公司 | Digital instrument reading identification method |
| CN110309831A (en) * | 2019-06-10 | 2019-10-08 | 北京工业大学 | Non intelligent water meter reading method based on machine vision |
| CN110309831B (en) * | 2019-06-10 | 2022-01-07 | 北京工业大学 | Non-intelligent water meter reading method based on machine vision |
| CN110232376A (en) * | 2019-06-11 | 2019-09-13 | 重庆邮电大学 | A kind of gear type digital instrument recognition methods returned using projection |
| CN110716513A (en) * | 2019-09-17 | 2020-01-21 | 国网河北能源技术服务有限公司 | Indirect analog quantity original data acquisition method based on EXCEL |
| CN110781884A (en) * | 2019-10-17 | 2020-02-11 | 广西电网有限责任公司电力科学研究院 | Method for realizing intelligent reading of electric meter data |
| CN110725711B (en) * | 2019-10-29 | 2023-08-29 | 南京北路智控科技股份有限公司 | Video-based drilling system and auxiliary drilling test method |
| CN110725711A (en) * | 2019-10-29 | 2020-01-24 | 南京北路自动化系统有限责任公司 | Drilling system based on video and auxiliary drilling inspection method |
| CN111027546A (en) * | 2019-12-05 | 2020-04-17 | 北京嘉楠捷思信息技术有限公司 | Character segmentation method and device and computer readable storage medium |
| CN111027546B (en) * | 2019-12-05 | 2024-03-26 | 嘉楠明芯(北京)科技有限公司 | Character segmentation method, device and computer readable storage medium |
| CN111017728A (en) * | 2019-12-23 | 2020-04-17 | 安徽工业大学 | Crane brake health state assessment method based on machine vision |
| CN111031222A (en) * | 2019-12-27 | 2020-04-17 | 山东厚德测控技术股份有限公司 | Real-time recognition device and method for character wheel of camera type gas meter |
| CN111242197A (en) * | 2020-01-07 | 2020-06-05 | 中国石油大学(华东) | Image and text matching method based on double-view-domain semantic reasoning network |
| CN111242197B (en) * | 2020-01-07 | 2023-11-07 | 中国石油大学(华东) | Image text matching method based on double-view semantic reasoning network |
| CN111695576A (en) * | 2020-06-09 | 2020-09-22 | 国网河北省电力有限公司营销服务中心 | Electric energy meter appearance rapid identification method based on neural network |
| CN111860317A (en) * | 2020-07-20 | 2020-10-30 | 青岛特利尔环保集团股份有限公司 | Boiler operation data acquisition method, system, equipment and computer medium |
| CN112101336A (en) * | 2020-09-09 | 2020-12-18 | 杭州测质成科技有限公司 | Intelligent data acquisition mode based on computer vision |
| CN112163546A (en) * | 2020-10-13 | 2021-01-01 | 珠海格力电器股份有限公司 | Water meter value determination method and device based on visual identification system |
| CN112418226A (en) * | 2020-10-23 | 2021-02-26 | 济南信通达电气科技有限公司 | Method and device for identifying opening and closing states of fisheyes |
| CN112966746A (en) * | 2020-11-20 | 2021-06-15 | 扬州大学 | Stable variable gray template generation method suitable for tire defect detection |
| CN112464928A (en) * | 2020-11-27 | 2021-03-09 | 广东电网有限责任公司 | Digital meter reading identification method, device, equipment and storage medium |
| CN112464928B (en) * | 2020-11-27 | 2024-03-15 | 广东电网有限责任公司 | Digital meter reading identification method, device, equipment and storage medium |
| CN113609906A (en) * | 2021-06-30 | 2021-11-05 | 南京信息工程大学 | Document-oriented table information extraction method |
| CN113591705A (en) * | 2021-07-30 | 2021-11-02 | 华东理工大学 | Inspection robot instrument recognition system and method and storage medium |
| CN113591705B (en) * | 2021-07-30 | 2024-03-19 | 华东理工大学 | Inspection robot instrument identification system and method and storage medium |
| CN114445784A (en) * | 2022-03-01 | 2022-05-06 | 武汉聚智惠仁信息技术有限公司 | Method and system for acquiring CRRT screen parameters in real time |
| CN116721094A (en) * | 2023-08-03 | 2023-09-08 | 济南邦德激光股份有限公司 | Visual pattern recognition method and system under complex background and laser cutting machine |
| CN116721094B (en) * | 2023-08-03 | 2023-12-19 | 济南邦德激光股份有限公司 | Visual pattern recognition method and system under complex background and laser cutting machine |
| CN116721042A (en) * | 2023-08-10 | 2023-09-08 | 广东石油化工学院 | Multi-threshold binarization-based image tilt correction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106909941A (en) | Multilist character recognition system and method based on machine vision | |
| CN108121991B (en) | Deep learning ship target detection method based on edge candidate region extraction | |
| CN105046252B (en) | A kind of RMB prefix code recognition methods | |
| CN108596166A (en) | A kind of container number identification method based on convolutional neural networks classification | |
| CN108694393A (en) | A kind of certificate image text area extraction method based on depth convolution | |
| CN105023008A (en) | Visual saliency and multiple characteristics-based pedestrian re-recognition method | |
| CN106022232A (en) | License plate detection method based on deep learning | |
| CN109325935A (en) | A kind of transmission line faultlocating method based on unmanned plane image | |
| CN105404868B (en) | The rapid detection method of text in a kind of complex background based on interaction platform | |
| Lu et al. | Saliency modeling from image histograms | |
| CN108491788A (en) | A kind of intelligent extract method and device for financial statement cell | |
| CN107038416B (en) | Pedestrian detection method based on binary image improved HOG characteristics | |
| CN109918971B (en) | Method and device for detecting number of people in monitoring video | |
| CN106530281A (en) | Edge feature-based unmanned aerial vehicle image blur judgment method and system | |
| Yuan et al. | Learning to count buildings in diverse aerial scenes | |
| CN111160169B (en) | Face detection method, device, equipment and computer readable storage medium | |
| CN106203237A (en) | The recognition methods of container-trailer numbering and device | |
| CN106570475B (en) | A kind of dark-red enameled pottery seal search method | |
| CN106709500A (en) | Image feature matching method | |
| Li et al. | SDBD: A hierarchical region-of-interest detection approach in large-scale remote sensing image | |
| CN111695373B (en) | Zebra stripes positioning method, system, medium and equipment | |
| CN112417931B (en) | Method for detecting and classifying water surface objects based on visual saliency | |
| CN106485639A (en) | The method and apparatus differentiating forged certificate picture | |
| CN109829511B (en) | Texture classification-based method for detecting cloud layer area in downward-looking infrared image | |
| CN112396638A (en) | Image processing method, terminal and computer readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170630 |
|
| RJ01 | Rejection of invention patent application after publication |