CN106960208A - A kind of instrument liquid crystal digital automatic segmentation and the method and system of identification - Google Patents
A kind of instrument liquid crystal digital automatic segmentation and the method and system of identification Download PDFInfo
- Publication number
- CN106960208A CN106960208A CN201710195624.7A CN201710195624A CN106960208A CN 106960208 A CN106960208 A CN 106960208A CN 201710195624 A CN201710195624 A CN 201710195624A CN 106960208 A CN106960208 A CN 106960208A
- Authority
- CN
- China
- Prior art keywords
- image
- liquid crystal
- character
- crystal digital
- automatic segmentation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/153—Segmentation of character regions using recognition of characters or words
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06V10/243—Aligning, centring, orientation detection or correction of the image by compensating for image skew or non-uniform image deformations
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Character Input (AREA)
- Character Discrimination (AREA)
Abstract
The invention discloses the method and system of a kind of instrument liquid crystal digital automatic segmentation and identification, the system of the instrument liquid crystal digital automatic segmentation and identification includes image pre-processing module, decimal point identification module, character cutting module, four modules of character recognition module, and the instrument liquid crystal digital automatic segmentation and knowledge method for distinguishing include image preprocessing, decimal point identification, character cutting, four steps of character recognition.Relative to prior art, the invention has the advantages that：1st, preprocessing part has very high robustness for the image that brightness is changed greatly, and well can make a distinction liquid crystal digital with background；2nd, it can be identified exactly for the decimal point in image；3rd, single liquid crystal digital discrimination is 98%, and whole string recognizes that correct success rate reaches 97% completely.
Description
Technical field
The invention belongs to Digital Image Processing, optical character identification (OCR) technical field, being related to one kind can be to Instrument image
Middle liquid crystal digital carries out cutting and simultaneously knows method for distinguishing, and in particular to the cutting and recognition methods of a kind of gas meter liquid crystal digital and be
System.
Background technology
With developing rapidly for Modernized City Construction, the construction of intelligent plant and intelligent city has showed inevitable
Trend, all kinds of instrument such as ammeter, water meter, gas meter are appeared among the life of the people in large quantities.However, China is many at present
The metering method of water rate, the gas charge in place etc. still relies primarily on the execute-in-place of meter reading personnel, is copied and accepted into cell, then record
Enter in computer, do not only exist the error caused by human factor so, also create the waste of human and material resources.With computer
With the development of electronic communication, continued to bring out for the various remote automatic meter reading systems of different occasions, substitute cumbersome hand
Work is worked, and conveniently tables look-up, user is not disturbed also, with good economic and social benefit.Current remote kilowatt meter reading-out system is a variety of more
Sample, one of which is to be taken pictures by placing a camera before instrument for instrument, then is obtained by OCR identification technologies in instrument
Numerical value.This mode, which is that one kind is relatively inexpensive also in factory or cell using older non intelligent instrument for those, to be had
The reforming mode of effect.By contrast, the long-term human input of this disposable input ratio will substantially be calculated much；And if
These places are replaced by new intelligent meter, then the apparatus and dismounting expense of needs are also substantially much higher, and have to production and living
Certain influence.The present invention is exactly the liquid crystal digital cutting and identification being mainly directed towards in the remote automatic meter reading system of above-mentioned instrument
The problem of.
The technology that the cutting of instrument liquid crystal digital is mainly used with identification is optical character identification (OCR) technology.It is generally existing
Some OCR softwares are general by several portions such as the input and pretreatment of image, printed page analysis, Character segmentation, character recognition and post processing
It is grouped into, wherein：Image input is main to be scanned using optical instrument such as scanner, facsimile machine etc. to target or with number
Camera, which is taken pictures, obtains digital image data, and the factor such as illumination condition, resolution ratio when image is generated influences whether follow-up identification
Effect and accuracy；Pretreatment includes the processes such as image binaryzation, noise remove, Slant Rectify；Printed page analysis is by document
Picture presses paragraph segmentation, because the diversity of document is currently without fixed optimal cutling model；Due to the limitation of photographical condition,
Characters Stuck, disconnected pen are often resulted in, the characteristics of along with Chinese character itself, software for discerning characters needs complicated character cutting work(
Can, a line character is cut into single character；Character recognition technologies belong to statistical learning, and the method for early stage is template matches,
It is later then based on feature extraction；Post processing includes the space of a whole page and recovered, according to the contact of language-specific context correction etc..
OCR will be faced with new challenges applied to instrument.The image of document files typically can be by section
Fall or cut into by row the image of single line text, and the environment residing for Instrument image liquid crystal digital is then more complex, dial plate etc.
The content such as unessential character can all be impacted to printed page analysis on part and display screen, therefore can not be easily by target
String Region is cut out, it is necessary to for different instrument processing method reasonable in design.The problem of another is important is instrument
Table is larger by illumination effect when taking pictures so that the brightness irregularities of image, is frequently encountered the problem of regional area is excessively bright, this
The identification of analysis and character to image has large effect, and is not in then this kind of ask to file scanning or when taking pictures
Topic, therefore the influence of exclusion illumination is an important job when handling Instrument image.In addition, for most liquid crystal digital instrument
For table, the position of decimal point is significant to recognition result, and the accuracy rate for improving decimal point identification should also be weight
The content wanted.Obviously, OCR is applied to need more to close, it is necessary to find the problem of considering many new during instrument
Reason method.
Zeng Zhongjie etc. deliver based on Character segmentation splice field formula liquid crystal digital identification (computer engineering with application,
2013,49 (12)：110-112) as shown in figure 1, main include pretreatment, three steps of Character segmentation and character recognition, first
By being pre-processed to original image, positioning, slant correction, targets improvement and the binaryzation of liquid crystal display are realized；Then root
The segmentation of character is realized according to the upright projection of character string；It is last to carry out character recognition by projection properties.This method exists following
Weak point：
1st, for the larger figure of illumination variation, this method can not obtain preferable binary map.
2nd, lack the identification of decimal point, and occur decimal point in many high-precision meters and scaling position is not solid
Fixed, therefore the identification of decimal point is indispensable.
3rd, character recognition uses the features such as the peak value of relatively simple projection to be judged, due to preprocess method and
Threshold value is set, and the influence for distorting the factor such as crooked of image in itself, still suffers from partial character identification mistake, such as 0 and 8
So close character of form.
The content of the invention
There is provided the side of a kind of instrument liquid crystal digital automatic segmentation and identification for the problem of present invention exists for the above method
Method and system.
The purpose of the present invention is achieved through the following technical solutions：
A kind of instrument liquid crystal digital automatic segmentation and knowledge method for distinguishing, including the identification of image preprocessing, decimal point, character are cut
Point, four steps of character recognition, comprise the following steps that：
First, image preprocessing
(1) processing is filtered to original image using LoG operators；
(2) by filtered image binaryzation, binary map is obtained；
(3) by the edge of hough change detection liquid crystal displays go forward side by side line tilt correct；
(4) on the basis of liquid crystal platen edge is detected, with reference to the height of character and the height of liquid crystal display into fixed proportion
This constraints, intercepts target string approximate region；
(5) binary map to above-mentioned interception area is expanded, corroded, and then calculates connected component, remove it is less and
Excessive component, obtains connected graph, and level, upright projection are finally done to connected graph, obtains the accurate location of character zone.
2nd, decimal point is recognized
(1) with a larger threshold value to the image binaryzation after LoG operator filterings, obtain one only comprising decimal point and
The binary map of a small number of noises；
(2) relative coordinate (x, y) of the geometric center of each connected component boundary rectangle in figure is calculated respectively；
(3) compared with the connected graph that is obtained during image preprocessing, the centre coordinate for calculating each connected component exists
Corresponding coordinate (x ', y ') in the figure, if (x ', y ') is located at numerically, then it is assumed that be a part rather than decimal for numeral
Point, the coordinate is removed；All coordinates (x ', y ') of screening, if there remains one, then it is assumed that the coordinate is exactly that decimal point exists
Relative position in artwork, otherwise artwork is in the absence of decimal point.
3rd, character cutting
(1) cutting is crossed to image using upright projection, divides the image into the image for several individual digits；
(2) for two adjacent narrower images, increase the alternative cutting scheme of two images of a merging, obtain one
Individual identification path profile.
4th, character recognition
(1) storehouse is recognized using existing individual character, to corresponding the be possible to individual character image of every section of camber line in identification path profile
It is identified；
(2) for there is the identification path profile of mulitpath, the average distance recognized under different paths is calculated；
(3) recognition result of the individual character recognition result on optimal path as character string is exported.
A kind of instrument liquid crystal digital automatic segmentation for realizing the above method and the system of identification, including image preprocessing mould
Block, decimal point identification module, character cutting module, character recognition module, wherein：
The pretreatment module is handled original image using LoG operators, then passes through hough change detection liquid crystal
The edge of screen go forward side by side line tilt correction and target positioning.
The decimal point identification module carries out binaryzation to the image after LoG operator filterings in pretreatment module, obtains only
Binary map comprising decimal point and a small number of noises, then by obtaining the position of decimal point to position reasonableness test.
The character cutting module uses upright projection and cutting excessively to divide the image into the image for several individual digits,
For the substantially narrower image of width, the alternative cutting scheme that increase by one merges two neighboring narrower image.
The character recognition module recognizes all possible list that storehouse is obtained to character cutting module using existing individual character
Word image is identified, and the part merged for there is narrower image carries out similarity distance contrast, obtains optimal cutting scheme.
Relative to prior art, the invention has the advantages that：
1st, preprocessing part has very high robustness for the image that brightness is changed greatly, can be well by liquid crystal digital
Made a distinction with background；
2nd, it can be identified exactly for the decimal point in image；
3rd, single liquid crystal digital discrimination is 98%, and whole string recognizes that correct success rate reaches 97% completely.
Brief description of the drawings
Fig. 1 is the field formula liquid crystal digital recognizer framework spliced based on Character segmentation；
Fig. 2 is that hough becomes scaling method schematic diagram, and (a) counts every with straight line where parameter θ and r denotation coordinations (x, y), (b)
To the number of the foreground point on parameter θ and r line correspondences；
Two kinds of structural elements that Fig. 3 is expansion, corrosion is used, the structural element of (a) square, (b) cross structure
Element；
Fig. 4 is the extraction of connected component, and (a) image A, (b) A containing 2 connected components connected component marks image,
(c) 3 × 3 structural element S；
Fig. 5 is identification path schematic diagram；
Fig. 6 is Instrument image (by taking gas meter as an example)；
Fig. 7 is the flow chart of the process of liquid crystal digital in identification Instrument image；
Fig. 8 is that preprocessing part handles the Laplce's Gauss operator used during image；
Fig. 9 is the image obtained after being filtered using LoG operators to original image；
Figure 10 is to remove the connected graph obtained after larger and less connected component to Fig. 9 expansions, burn into；
Figure 11 is the image of the digital region of the Object LC being truncated to after pretreatment；
Figure 12 is the schematic diagram of character cutting；
Figure 13 is to add a variety of words that will be formed after adjacent narrower block combines in character cutting cutting scheme
Accord with the schematic diagram of cutting scheme.
Embodiment
Technical scheme is further described below in conjunction with the accompanying drawings, but is not limited thereto, it is every to this
Inventive technique scheme is modified or equivalent, without departing from the spirit and scope of technical solution of the present invention, all should be covered
In protection scope of the present invention.
Embodiment one：Present embodiments provide for the system of a kind of instrument liquid crystal digital automatic segmentation and identification,
The system is made up of image preprocessing, decimal point identification, character cutting, four modules of character recognition, wherein：
The pretreatment module is mainly handled using Laplce's Gauss operator (LoG), the picture larger to luminance difference
The effect of effective noise reduction process can be reached, preferable binary map is obtained；Then the edge of hough change detection liquid crystal displays is passed through
Go forward side by side line tilt correction and target positioning.
The decimal point identification module uses a larger threshold value to the image after LoG operator filterings in pretreatment module
Binaryzation is carried out, the binary map only comprising decimal point and a small number of noises is obtained, then by being obtained to modes such as position reasonableness tests
To the position of decimal point.
The character cutting module uses upright projection and cutting excessively to divide the image into the image for several individual digits,
For the substantially narrower image of width, the alternative cutting scheme that increase by one merges two neighboring narrower image.
The character recognition module recognizes all possible list that storehouse is obtained to character cutting module using existing individual character
Word image is identified, and the part merged for there is narrower image carries out similarity distance contrast, obtains optimal cutting scheme.
Embodiment two：Present embodiments provide for a kind of instrument liquid crystal digital automatic segmentation and knowledge method for distinguishing,
Methods described includes following four step：Image preprocessing, decimal point identification, character cutting, character recognition.Each step is detailed
Thin process is as follows：
1st, image preprocessing
(1) using Laplce's Gauss operator (LoG) to image filtering.
LoG functions are defined as follows：
Gσ(x, y) is the Gaussian function (also referred to as normal distyribution function) of two dimension, is defined as follows：
Wherein, x, y are the variables of function in two-dimensional Cartesian coordinate system, and parameter σ is the standard deviation of Gaussian function, and σ ＞ 0,
Which determine the sphere of action (or intensity of normal distribution) of Gaussian function.σ is bigger, then Gaussian function and LoG functions
Sphere of action is bigger.
When decimal spot diameter in 2 σ=figure or liquid crystal digital stroke width, convolution kernel size=(6 σ, 6 σ), effect is best.
(2) filtered image binaryzation is obtained into binary map.Threshold value is calculated by Otsu algorithm (OTSU).Da-Jin algorithm is again
It is called maximum variance between clusters, inter-class variance computing formula：
w0(t)*(u0(t)-u)2+w1(t)*(u1(t)-u)2。
Wherein, t is the threshold value of classification, and w0 is background ratio, and u0 is background mean value, and w1 is prospect ratio, and u1 is that prospect is equal
Value, u is the average of entire image.L to gray level image gray level, the image that tonal range is [0, L-1], calculate when with each
The result of above formula when gray level is threshold value, it is exactly optimal threshold value to make the threshold value t that above formula obtains maximum.
(3) hough change detection liquid crystal platen edges are passed through.The minimum length of detection of straight lines and inclination angle model by limiting
Enclose, can reject due to the flase drop that noise jamming is caused.
The algorithm idea of hough change detection straight lines：
The straight line cluster of passing point (x, y) meets following constraint equation in two-dimensional Cartesian coordinate system：
X*cos θ+y*sin θ=r.
Wherein, r is geometry vertical range of the origin to straight line, and angle, θ is the angle between vertical line r and X-axis.Such as Fig. 2 (a)
It is shown.
θ angles are divided into several interval, all foreground pixel points in calculating image in -90 degree to 90 degree of scope
The r of each θ angles line correspondence of (x, y) value, every straight line of statistics (θ, r) on foreground point number, shown in such as Fig. 2 (b).
Be considered as having in image when number is higher than some threshold value t an obvious straight line (θ, r).
(4) line tilt correction is entered according to the tilt angle theta of the straight line detected.When the angle of inclination of liquid crystal display is too big,
Gross distortion can occur after slant correction for character string.Therefore in order to ensure recognition accuracy, the algorithm is made to angle of inclination
Limitation, just refuses identification when angle of inclination, which exceedes, to be limited, such as restriction level direction straight incline angle range for (- 15 ,+
15), vertical direction straight incline angle range for [- 90, -75) ∪ (75,90).
(5) on the basis of liquid crystal platen edge is oriented, with reference to the height of character and the height of liquid crystal display into fixed proportion
This constraints, intercepts target string approximate region.
(6) binary map to above-mentioned interception area is expanded, corroded.The purpose of expansion is by the stroke company of liquid crystal digital
Pick up and, then stroke is reverted to original thickness by corrosion, expansion should use the structural element of square, and corrosion should make
Cross-shaped configuration element is used, as shown in Figure 3.Then connected component is calculated, removes less and excessive component, is connected
Figure.The threshold value of wherein minimum connected component should be greater than the size of decimal point, and the threshold value of largest connected component should be at least above numeral
The size of " 8 ".Level, upright projection are finally done to connected graph, the accurate location of character zone is obtained.
Here a kind of method of the extraction connected component based on morphological dilation is briefly introduced.
The image A containing multiple connected components as shown in Fig. 4 (a), from the figure of some point of only connected component A1 inside
As B starts, constantly expanded using the structural element S shown in Fig. 4 (c), due to A1 and other connected components at least one 1
The wide space of individual pixel (dotted line in Fig. 4 (a)), so can guarantee that expansion will not be produced in other connected regions every time
Point, thus need to only be intersected with artwork A with the result after expansion every time, expansion can be just limited in inside A1, until final B fills
Full whole connected component A1, extracts connected component A1 and finishes.Algorithm summary is as follows：
Initialization：B0Some point in=connected component A1
Circulation：
Until Bi+1=Bi
2nd, decimal point is recognized
Decimal point identification uses the image used in preprocessing process after LoG operator filterings.
(1) one two-value only comprising decimal point and a small number of noises is obtained to image binaryzation with a larger threshold value
Figure.0.8 times that image pixel maximum can directly be chosen makees threshold value, or uses maximum with Da-Jin algorithm threshold value and image pixel
The average value of value makees threshold value.
(2) connected component of binary map is calculated, it is decimal point that wherein some connected component, which is possible to corresponding, is counted respectively
Calculate relative coordinate (x, y) of the geometric center of each connected component boundary rectangle in figure.
(3) compared with the connected graph that is obtained in preprocessing process, calculate the centre coordinate of each connected component in the figure
In corresponding coordinate (x ', y '), if (x ', y ') be located at numerically, then it is assumed that be numeral a part rather than decimal point, will
The coordinate removes.All coordinates (x ', y ') of screening, if there remains one, then it is assumed that the coordinate is exactly decimal point in artwork
In relative position, otherwise artwork be not present decimal point.
3rd, character cutting
(1) cutting is crossed to image using upright projection, the image for several individual digits is divided the image into, such as Figure 12 institutes
Show.The threshold value of cutting should be less than the width of liquid crystal digital stroke.
(2) the substantially narrower image of width may not be a complete character after cutting, thus for adjacent two compared with
Narrow image, should increase the cutting scheme of two images of a merging, obtain an identification path profile, as shown in figure 13.
It is about that (character-spacing refers to adjacent two to 0.5 times of character-spacing due to complete character to be cut into the width of every image after two pieces
The distance between individual character center point), therefore judge that its whether complete threshold value can be set to 0.75 times of character-spacing.And character-spacing is substantially
Be with character height into fixed proportion, therefore in practical operation, the ratio of threshold value=character height × character-spacing and character height
×0.75.It is noted here that also needing to find the cutting block where decimal point, complete character block should be regarded as.
4th, character recognition
(1) storehouse is recognized using existing individual character, to corresponding the be possible to individual character image of every section of camber line in identification path profile
It is identified.Used recognition methods is first to tie up Gradient Features to image zooming-out 100, is then entered using vector quantization method
Row classification.
(2) for there is the identification path profile of mulitpath, the average distance recognized under different paths is calculated.
Path profile one shown in Fig. 5 has 3 identification paths, and d1, d2 ..., d6 correspond to cutting image for every section of camber line
Recognize distance.3 paths are respectively { 0,1,2,3,4 }, { 0,1,3,4 }, { 0,2,3,4 }.Calculate the average identification per paths
Distance (d1+d2+d3+d4)/4, (d1+d6+d4)/3, (d5+d3+d4)/3.Path in these three values corresponding to minimum value
It is exactly optimal cutting scheme.
(3) recognition result of the individual character recognition result on optimal path as character string is exported.
Embodiment three：Present embodiment is automatic to the instrument liquid crystal digital of the present invention by taking gas meter image as an example
Cutting and recognition methods are described in detail.
The gas meter image handled in present embodiment is as shown in Figure 6.The development platform of recognizer is VS2013, identification
Program is write with C++.
The gas meter image of collection needs to have following characteristics：
1st, image resolution ratio is more than 50dpi coloured image；
2nd, image storage format is the jpg forms of 24；
3rd, image should include complete LCDs region, tilt and be less than ± 10 °.
If the gas meter image of input, which is not reaching to above-mentioned standard, may reduce discrimination.
In specific implementation process, gas meter image is handled according to the flow shown in Fig. 7, detailed process is such as
Under：
1st, image preprocessing
The size of Laplce's Gauss operator (LoG) is determined according to the height of liquid crystal digital.Gas meter image liquid crystal in Fig. 6
Digital height is 30, and stroke width is about 3, and the LoG operators size used is 9 × 9, σ=1.5, as shown in Figure 8.
(1) processing is filtered to gas meter image using LoG operators, filtered treatment effect is as shown in Figure 9.
(2) by filtered image binaryzation, obtaining binary map.
(3) by hough change detection liquid crystal platen edges, line tilt of going forward side by side correction, according to the liquid crystal display region detected,
The approximate location of liquid crystal digital is obtained in proportion.
(4) by expansion, corrosion, connected component is calculated, removes less and excessive, obtains connected graph such as Figure 10 institutes
Show.
(5) accurate location of character zone is obtained by level, upright projection in the approximate location that (3) step is obtained,
As shown in figure 11.
2nd, decimal point is recognized
(1) image (Fig. 9) of the pretreatment module after LoG operator filterings is used, with a larger threshold value (0.8* image
Pixel maximum gradation value) binaryzation is carried out, obtain a binary map only comprising decimal point and a small number of noises.
(2) several coordinates that decimal point is likely to occur are obtained by calculating connected domain center.
(3) it is (small when removing less connected component because decimal point is smaller by being compared with the connected graph in Figure 10
Several points can be removed, therefore Figure 10 is not in decimal point), if point is appeared in numerically, then it is assumed that be a part for numeral
Rather than decimal point and remove the point coordinates.All alternative coordinates are screened, if surplus next, then it is assumed that decimal point is appeared in
The coordinate position, otherwise in the absence of decimal point.
3rd, character cutting
(1) cutting is crossed to image using upright projection, the image for several individual digits is divided the image into, such as Figure 12 institutes
Show.
(2) for the substantially narrower image of width, the alternative cutting that increase by one merges two neighboring narrower image
Scheme, obtains an identification path profile, as shown in figure 13, each one cut-off of node on behalf in figure, every camber line represents one
Individual individual character image, the two ends of arc represent the starting and ending position of the cutting respectively, and a kind of cutting scheme is represented per paths.Example
Such as：Figure 10 first " 4 " have been cut into two pieces, then have 2 two kinds of cutting schemes of 0- ＞ 1- ＞ 2 and 0- ＞, although and 5- ＞ 6
Also it is very narrow but identified in (2) step as decimal point, therefore be not included.
4th, character recognition
(1) storehouse is recognized using existing individual character, (i.e. Figure 10 is every to recognizing the individual character image that is possible to specified in path profile
The corresponding image of bar camber line) it is identified；
(2) for there is the identification path profile of mulitpath, the average distance recognized under different paths is calculated, obtaining optimal
Cutting route；
(3) recognition result of the individual character recognition result on optimal path as character string is exported.
The above method is tested on 427 gas meter images, including 129 3 type tables, 298 Z-type tables.Know
Other result is：Z-type table is completely correct, only 3 mistakes of 3 type tables, and one of them is to be identified as 17, and one is mistake when projecting
Last two digits are lost, also one is to be identified as 95.It can be seen that the present invention is for qualified gas meter image
With higher discrimination.
Claims (10)
1. a kind of instrument liquid crystal digital automatic segmentation and know method for distinguishing, it is characterised in that methods described is comprised the following steps that：
First, image preprocessing
(1) processing is filtered to original image using LoG operators；
(2) by filtered image binaryzation, binary map is obtained；
(3) by the edge of hough change detection liquid crystal displays go forward side by side line tilt correct；
(4) on the basis of liquid crystal platen edge is detected, with reference to the height of character and the height of liquid crystal display into fixed proportion this
Constraints, intercepts target string approximate region；
(5) binary map to above-mentioned interception area is expanded, corroded, and then calculates connected component, removes less and excessive
Component, obtain connected graph, level, upright projection finally done to connected graph, obtain the accurate location of character zone；
2nd, decimal point is recognized
(1) with a larger threshold value to the image binaryzation after LoG operator filterings, obtain one and only include decimal point and minority
The binary map of noise；
(2) relative coordinate (x, y) of the geometric center of each connected component boundary rectangle in figure is calculated respectively；
(3) compared with the connected graph that is obtained during image preprocessing, calculate the centre coordinate of each connected component in the figure
In corresponding coordinate (x ', y '), if (x ', y ') be located at numerically, then it is assumed that be numeral a part rather than decimal point, will
The coordinate removes；All coordinates (x ', y ') of screening, if there remains one, then it is assumed that the coordinate is exactly decimal point in artwork
In relative position, otherwise artwork be not present decimal point；
3rd, character cutting
(1) cutting is crossed to image using upright projection, divides the image into the image for several individual digits；
(2) for two adjacent narrower images, increase the alternative cutting scheme of two images of a merging, obtain a knowledge
Other path profile；
4th, character recognition
(1) storehouse is recognized using existing individual character, corresponding the be possible to individual character image of every section of camber line in identification path profile is carried out
Identification；
(2) for there is the identification path profile of mulitpath, the average distance recognized under different paths is calculated；
(3) recognition result of the individual character recognition result on optimal path as character string is exported.
2. instrument liquid crystal digital automatic segmentation according to claim 1 and knowledge method for distinguishing, it is characterised in that the step
In one (1), the σ of the σ of the LoG operators size that filtering process is used=6 × 6, and decimal spot diameter or liquid crystal digital stroke in 2 σ=figure
Width.
3. instrument liquid crystal digital automatic segmentation according to claim 1 and knowledge method for distinguishing, it is characterised in that the step
In one (3), the method at the edge of hough change detection liquid crystal displays is as follows：
The straight line cluster of passing point (x, y) meets following constraint equation in two-dimensional Cartesian coordinate system：
X*cos θ+y*sin θ=r,
Wherein, r is geometry vertical range of the origin to straight line, and angle, θ is the angle between vertical line r and X-axis；
θ angles are divided into several interval, all foreground pixel points (x, y) in calculating image in -90 degree to 90 degree of scope
Each θ angles line correspondence r value, every straight line of statistics (θ, r) on foreground point number, when number is higher than some threshold value
Be considered as having in image during t an obvious straight line (θ, r).
4. instrument liquid crystal digital automatic segmentation according to claim 3 and knowledge method for distinguishing, it is characterised in that the threshold value
Calculated by Otsu algorithm.
5. instrument liquid crystal digital automatic segmentation according to claim 1 and knowledge method for distinguishing, it is characterised in that the step
In one (5), using the structural element of square in expansion procedure, cross-shaped configuration element is used during corrosion treatment；
The threshold value of minimum connected component is more than the size of decimal point, size of the threshold value at least above digital " 8 " of largest connected component.
6. instrument liquid crystal digital automatic segmentation according to claim 1 and knowledge method for distinguishing, it is characterised in that the step
In two (1), choose image pixel maximum 0.8 times makees threshold value, or uses with Da-Jin algorithm threshold value and image pixel maximum
Average value make threshold value.
7. instrument liquid crystal digital automatic segmentation according to claim 1 and knowledge method for distinguishing, it is characterised in that the step
In three (1), the threshold value of cutting is less than the width of liquid crystal digital stroke；In step 3 (2), judge the whether narrow threshold value of image=
Ratio × 0.75 of character height × character-spacing and character height.
8. instrument liquid crystal digital automatic segmentation according to claim 1 and knowledge method for distinguishing, it is characterised in that the step
In four (1), individual character recognition methods is：Gradient Features first are tieed up to image zooming-out 100, then divided using vector quantization method
Class.
9. instrument liquid crystal digital automatic segmentation according to claim 1 and knowledge method for distinguishing, it is characterised in that the step
In four (2), character string, which is integrally recognized, has used identification path profile to obtain optimal cutting method.
10. a kind of realize instrument liquid crystal digital automatic segmentation described in claim 1-9 any claims and know method for distinguishing
System, it is characterised in that the system is by image pre-processing module, decimal point identification module, character cutting module, character recognition
Module composition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710195624.7A CN106960208B (en) | 2017-03-28 | 2017-03-28 | Method and system for automatically segmenting and identifying instrument liquid crystal number |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710195624.7A CN106960208B (en) | 2017-03-28 | 2017-03-28 | Method and system for automatically segmenting and identifying instrument liquid crystal number |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106960208A true CN106960208A (en) | 2017-07-18 |
| CN106960208B CN106960208B (en) | 2020-03-31 |
Family
ID=59470621
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710195624.7A Active CN106960208B (en) | 2017-03-28 | 2017-03-28 | Method and system for automatically segmenting and identifying instrument liquid crystal number |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106960208B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109034160A (en) * | 2018-07-06 | 2018-12-18 | 江苏迪伦智能科技有限公司 | A kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks |
| CN109145907A (en) * | 2018-09-11 | 2019-01-04 | 上海中信信息发展股份有限公司 | Text image based on commonly used word word frequency statistic is inverted detection method and device |
| CN109508714A (en) * | 2018-08-23 | 2019-03-22 | 广州市心鉴智控科技有限公司 | A kind of low-cost multi-channel real-time digital instrument board visual identity method and system |
| CN110188746A (en) * | 2019-04-19 | 2019-08-30 | 南京大学 | A kind of segmentation and recognition methods towards digital lcd |
| CN110443220A (en) * | 2019-08-13 | 2019-11-12 | 树根互联技术有限公司 | Digital table image-recognizing method, device, electronic equipment and storage medium |
| CN110659645A (en) * | 2019-08-05 | 2020-01-07 | 沈阳工业大学 | Character recognition method for digital instrument |
| CN111222507A (en) * | 2020-01-10 | 2020-06-02 | 随锐科技集团股份有限公司 | Automatic identification method of digital meter reading and computer readable storage medium |
| CN111582262A (en) * | 2020-05-07 | 2020-08-25 | 京源中科科技股份有限公司 | Segment type liquid crystal picture content identification method, device, equipment and storage medium |
| 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 |
| CN112561903A (en) * | 2020-12-24 | 2021-03-26 | 中铁建设集团基础设施建设有限公司 | Temperature shrinkage crack resistance method suitable for asphalt pavement in cold region |
| CN112726360A (en) * | 2020-12-24 | 2021-04-30 | 中铁建设集团基础设施建设有限公司 | Airport concrete pavement crack repairing method |
| CN112836726A (en) * | 2021-01-12 | 2021-05-25 | 云南电网有限责任公司电力科学研究院 | Pointer instrument indication reading method and device based on video information |
| CN114308718A (en) * | 2021-11-16 | 2022-04-12 | 江汉大学 | Method and device for sorting clothes according to sizes of clothes |
| CN114519694A (en) * | 2021-12-28 | 2022-05-20 | 河南大学 | Seven-segment digital tube liquid crystal display screen identification method and system based on deep learning |
| CN116246265A (en) * | 2023-05-12 | 2023-06-09 | 威海凯思信息科技有限公司 | Property management method and device based on image processing |
| CN110084083B (en) * | 2018-01-26 | 2024-04-19 | 新天科技股份有限公司 | Character detection method based on stroke characteristics |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080008383A1 (en) * | 2006-07-07 | 2008-01-10 | Lockheed Martin Corporation | Detection and identification of postal metermarks |
| 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 |
| CN105654072A (en) * | 2016-03-24 | 2016-06-08 | 哈尔滨工业大学 | Automatic character extraction and recognition system and method for low-resolution medical bill image |
| 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 |
-
2017
- 2017-03-28 CN CN201710195624.7A patent/CN106960208B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080008383A1 (en) * | 2006-07-07 | 2008-01-10 | Lockheed Martin Corporation | Detection and identification of postal metermarks |
| 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 |
| CN105654072A (en) * | 2016-03-24 | 2016-06-08 | 哈尔滨工业大学 | Automatic character extraction and recognition system and method for low-resolution medical bill image |
| 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 |
Non-Patent Citations (5)
| Title |
|---|
| 于金伦 等: "联机手写字符串识别搜索算法研究", 《2007年全国模式识别学术会议》 * |
| 刘宗林 等: "一种数显仪表字符快速自动识别方法", 《图形图像》 * |
| 曾仲杰 等: "基于字符切割拼接的字段式液晶数字识别", 《计算机工程与应用》 * |
| 童文超 等: "LED显示仪表的字符识别方法", 《湖南工业大学学报》 * |
| 谢兰军 等: "数字字符切分中小数点处理", 《科技资讯》 * |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110084083B (en) * | 2018-01-26 | 2024-04-19 | 新天科技股份有限公司 | Character detection method based on stroke characteristics |
| CN109034160B (en) * | 2018-07-06 | 2019-07-12 | 江苏迪伦智能科技有限公司 | A kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks |
| CN109034160A (en) * | 2018-07-06 | 2018-12-18 | 江苏迪伦智能科技有限公司 | A kind of mixed decimal point digital instrument automatic identifying method based on convolutional neural networks |
| CN109508714A (en) * | 2018-08-23 | 2019-03-22 | 广州市心鉴智控科技有限公司 | A kind of low-cost multi-channel real-time digital instrument board visual identity method and system |
| CN109508714B (en) * | 2018-08-23 | 2021-02-09 | 心鉴智控(深圳)科技有限公司 | Low-cost multi-channel real-time digital instrument panel visual identification method and system |
| CN109145907B (en) * | 2018-09-11 | 2020-12-11 | 上海信联信息发展股份有限公司 | Text image inversion detection method and device based on common word frequency statistics |
| CN109145907A (en) * | 2018-09-11 | 2019-01-04 | 上海中信信息发展股份有限公司 | Text image based on commonly used word word frequency statistic is inverted detection method and device |
| CN110188746A (en) * | 2019-04-19 | 2019-08-30 | 南京大学 | A kind of segmentation and recognition methods towards digital lcd |
| CN110659645B (en) * | 2019-08-05 | 2023-01-31 | 沈阳工业大学 | Character recognition method for digital instrument |
| CN110659645A (en) * | 2019-08-05 | 2020-01-07 | 沈阳工业大学 | Character recognition method for digital instrument |
| CN110443220A (en) * | 2019-08-13 | 2019-11-12 | 树根互联技术有限公司 | Digital table image-recognizing method, device, electronic equipment and storage medium |
| CN111222507A (en) * | 2020-01-10 | 2020-06-02 | 随锐科技集团股份有限公司 | Automatic identification method of digital meter reading and computer readable storage medium |
| CN111222507B (en) * | 2020-01-10 | 2024-03-19 | 随锐科技集团股份有限公司 | Automatic identification method for digital meter reading and computer readable storage medium |
| CN111582262A (en) * | 2020-05-07 | 2020-08-25 | 京源中科科技股份有限公司 | Segment type liquid crystal picture content identification method, device, equipment and storage medium |
| 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 |
| CN112726360A (en) * | 2020-12-24 | 2021-04-30 | 中铁建设集团基础设施建设有限公司 | Airport concrete pavement crack repairing method |
| CN112561903A (en) * | 2020-12-24 | 2021-03-26 | 中铁建设集团基础设施建设有限公司 | Temperature shrinkage crack resistance method suitable for asphalt pavement in cold region |
| CN112836726B (en) * | 2021-01-12 | 2022-06-07 | 云南电网有限责任公司电力科学研究院 | Pointer instrument indication reading method and device based on video information |
| CN112836726A (en) * | 2021-01-12 | 2021-05-25 | 云南电网有限责任公司电力科学研究院 | Pointer instrument indication reading method and device based on video information |
| CN114308718A (en) * | 2021-11-16 | 2022-04-12 | 江汉大学 | Method and device for sorting clothes according to sizes of clothes |
| CN114519694A (en) * | 2021-12-28 | 2022-05-20 | 河南大学 | Seven-segment digital tube liquid crystal display screen identification method and system based on deep learning |
| CN114519694B (en) * | 2021-12-28 | 2024-06-18 | 河南大学 | Seven-segment nixie tube liquid crystal display screen identification method and system based on deep learning |
| CN116246265A (en) * | 2023-05-12 | 2023-06-09 | 威海凯思信息科技有限公司 | Property management method and device based on image processing |
| CN116246265B (en) * | 2023-05-12 | 2023-08-08 | 威海凯思信息科技有限公司 | Property management method and device based on image processing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106960208B (en) | 2020-03-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106960208A (en) | A kind of instrument liquid crystal digital automatic segmentation and the method and system of identification | |
| CN108596066B (en) | Character recognition method based on convolutional neural network | |
| CN107491730A (en) | A kind of laboratory test report recognition methods based on image procossing | |
| CN109426814B (en) | Method, system and equipment for positioning and identifying specific plate of invoice picture | |
| CN105654072A (en) | Automatic character extraction and recognition system and method for low-resolution medical bill image | |
| CN105913093A (en) | Template matching method for character recognizing and processing | |
| KR940003829B1 (en) | Position recognizing method | |
| CN106326887B (en) | A kind of method of calibration and device of optical character identification result | |
| RU2640322C2 (en) | Methods and systems of effective automatic recognition of symbols | |
| RU2648638C2 (en) | Methods and systems of effective automatic recognition of symbols using a multiple clusters of symbol standards | |
| CN114549993B (en) | Method, system and device for grading line segment image in experiment and readable storage medium | |
| CN109993154A (en) | The lithium sulfur type instrument intelligent identification Method of substation's simple pointer formula | |
| CN112364834A (en) | Form identification restoration method based on deep learning and image processing | |
| CN107766854B (en) | Method for realizing rapid page number identification based on template matching | |
| CN113962929B (en) | Photovoltaic cell assembly defect detection method and system and photovoltaic cell assembly production line | |
| CN111626145A (en) | Simple and effective incomplete form identification and page-crossing splicing method | |
| CN114463770A (en) | Intelligent question-cutting method for general test paper questions | |
| CN109271882B (en) | Method for extracting color-distinguished handwritten Chinese characters | |
| CN111008635A (en) | OCR-based multi-bill automatic identification method and system | |
| CN110263784A (en) | The English paper achievement of intelligence identifies input method | |
| CN114693917A (en) | Data enhancement method applied to signboard identification | |
| CN112686872B (en) | Wood counting method based on deep learning | |
| CN113989482A (en) | Image processing-based automatic reading identification method for optical image-combination level meter | |
| CN109145912A (en) | A kind of digital instrument reading automatic identifying method | |
| CN113378847B (en) | Character segmentation method, system, computer device and 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |