CN101916364B - Adaptive dial digital identification method - Google Patents
Adaptive dial digital identification method Download PDFInfo
- Publication number
- CN101916364B CN101916364B CN2010102203746A CN201010220374A CN101916364B CN 101916364 B CN101916364 B CN 101916364B CN 2010102203746 A CN2010102203746 A CN 2010102203746A CN 201010220374 A CN201010220374 A CN 201010220374A CN 101916364 B CN101916364 B CN 101916364B
- Authority
- CN
- China
- Prior art keywords
- character
- zone
- dial
- image
- identification
- 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.)
- Expired - Fee Related
Links
Images
Abstract
The invention discloses an adaptive dial digital identification method. In the device and the method, according to a plurality of successive frames of dial images, the accurate positions of character zones and characters can be determined automatically, and a character template of a current dial is established in real time to improve the adaptability of dial digital identification; a robust skeletal feature matching method is provided to improve the accuracy of the dial digital identification; digital characters are identified by combining a template matching method and the skeletal feature matching method to reduce the complexity of the calculation by only the characteristic matching method; and template matching identification is performed by reestablishing a character template for the half-word characters generated by the rotating shooting of the dial to improve the accuracy of the dial digital identification. The device and the method have certain robustness with changed ambient light, are not limited to the dial character digital identification, and can be used for other digital identification, such as license plate identification.
Description
Technical field
The present invention relates to a kind of dial digital identification method, relate in particular to a kind of adaptive dial digital identification method.
Background technology
Traditional digit recognition method normally; The printing digital character is discerned processing through the method for template matches; The handwriting digital character is discerned processing through the method for characteristic matching, and in addition, carrying out digit recognition through neural metwork training also is one of common method.Because it is inconsistent that the character position of the bianry image after the image threshold processing distributes, lower based on the digit recognition method matching rate of template matches; Because it is noise such as light, lower based on the accuracy rate of the digit recognition method of characteristic matching; Accuracy rate based on the digit recognition method of neural network is relevant with Feature Selection and training time, though recognition accuracy is higher, does not also reach the requirement of current digit recognition application to accuracy.In addition, lower in the image resolution ratio of camera acquisition, under the poor situation of image subjective quality, it is inaccurate that the Character segmentation of bianry image can become, more than the accuracy rate of several kinds of traditional digit recognition methods also can reduce.
Research to digit recognition at present mainly concentrates on both direction, and the one, the research of new recognizer, the 2nd, the research of integrated approach.The former digit recognition accuracy rate is different because of method, can adjust correlation parameter to specific digit recognition application and obtain higher accuracy rate, can not realize the self-adaptation demand of digit recognition well; The latter comprises integrated and sorting technique integrated of characteristic, and is integrated through characteristic of different nature or sorting technique are carried out complementation, to improve whole reliability and discrimination.But existing such integrated digital recognition methods robustness on charcter topology is not strong.
Summary of the invention
The objective of the invention is to deficiency, a kind of dial digital identification method is provided to prior art.
The objective of the invention is to realize through following technical scheme:
A kind of adaptive dial digital identification method is characterized in that this method may further comprise the steps:
(1) location character zone, it specifically may further comprise the steps:
(1.1) read the dial plate image, continuous multiple frames dial plate image is done the frame difference add up, remove because the zone that noise effect produces; According to the position relation of camera and dial plate, confirm the character zone border by the character zone symmetry, to obtain the coarse positioning of character zone;
(1.2) character zone of coarse positioning is judged, got back to the dial plate image read step in the step (1.1) if character zone does not meet preset range, and carry out the character zone coarse positioning once more; The exact position of classifying to confirm character if character zone meets preset range according to character boundary is if big character gets into step (1.3), if small characters gets into step (1.4);
(1.3) isolate blackaandawhite characterumay zone and red wrongly written or mispronounced character symbol zone according to the difference of character zone RGB color; Respectively the image binaryzation processing is carried out in blackaandawhite characterumay zone and red wrongly written or mispronounced character symbol zone, and removed in the binary image because the zone that noise effect produces; Character zone is carried out normalization to be handled; Each character in the character zone of handling through normalization is carried out skeletal extraction, framework characteristic extraction and characteristic matching rate to be calculated; If the framework characteristic matching rate of alphabet all reaches 100% in the character zone; Be judged as the exact position of the character of correctly cutting apart; And relevant character exact position and the character number of record, otherwise the aforesaid operations that carries out step (1.3) again is till definite character exact position;
(1.4) gray-scale value of extraction character zone is with the edge of Canny operator extraction character zone; The exact position that is partitioned into character according to the character independence and the distributing homogeneity at edge; And judge the rationality of this split position, if be judged as the aforesaid operations that do not meet the rationality condition then carry out step (1.4) again till definite character exact position;
(2) set up Character mother plate, it specifically may further comprise the steps:
(2.1) character self-adapting binaryzation: each character zone is carried out adaptive image binaryzation handle, and remove in the binary image because the zone that noise effect produces;
(2.2) character zone normalization;
(2.3) each character in the normalization character zone is carried out skeletal extraction, framework characteristic extraction and characteristic matching rate and calculate, thus identification character;
The size of the characteristic matching rate that (2.4) calculates according to step (2.3) determines whether to be correct character identification result, if then the pixel of this character zone is write the corresponding characters template;
(3) identification dial digital, it specifically may further comprise the steps:
(3.1) judge whether each character in the character zone is the half-word character;
(3.2) if character zone does not have the half-word character; Then carry out character match with Character mother plate; Statistical pixel correct match rate, if the correct match rate greater than certain threshold value, and the matching rate of other characters is well below this correct match degree; Then export recognition result, otherwise can not discern the character in this character zone;
(3.3) character that can not discern for template matches in the step (3.2) carries out the framework characteristic coupling, if the framework characteristic matching rate surpasses certain threshold value then exports recognition result, otherwise the character in this character zone is a unrecognizable character;
(3.4) if character zone has the half-word character, rebuild matching template, carry out character match with the reconstruction matching template, statistical pixel correct match rate is confirmed the recognition result of half-word character.。
The invention has the beneficial effects as follows: the present invention can set up the Character mother plate of current dial plate in real time according to the exact position of automatic location character regional location of the dial plate image of continuous multiple frames and character, has improved the adaptivity of dial digital identification; Propose a kind of framework characteristic matching process of robust, improved the accuracy rate of dial digital identification; Carry out Number character recognition in conjunction with template matching method and framework characteristic matching process, reduced simple computation complexity with feature matching method; Rotate the half-word character reconstruction Character mother plate of taking generation to dial plate and carry out template matches identification, improved the accuracy rate of dial digital identification; The present invention has certain robustness under the situation that ambient light changes; The invention is not restricted to the digit recognition of dial plate character, can also be generalized to other digit recognition and use, such as car plate identification etc.
Description of drawings
Accompanying drawing 1 is the concrete structure figure of apparatus of the present invention;
Accompanying drawing 2 is to extract the related 3*3 neighborhood synoptic diagram of skeleton in the image identification unit of the present invention;
Accompanying drawing 3 is the pixel distribution plans that extract two related Rule of judgment of skeleton in the image identification unit of the present invention;
Accompanying drawing 4 is to extract eight related neighborhood search direction synoptic diagram of skeleton in the image identification unit of the present invention;
Accompanying drawing 5 is algorithm flow charts of image pretreatment unit of the present invention;
Accompanying drawing 6 is wired circuit figure of image identification unit of the present invention and other devices;
Accompanying drawing 7 is frame diagrams of image identification unit of the present invention;
Accompanying drawing 8 is algorithm flow charts of image identification unit of the present invention.
Embodiment
The definition of noun:
1, half-word character definition: because dial plate rotate to produce is not complete character, but former and later two characters is half the, claims that such character is the half-word character.
2, adaptive image binaryzation is handled:
Distribute to confirm threshold value according to the area pixel value, image is carried out binaryzation according to threshold value.Threshold value confirms that step is following:
1., the pixel brightness value to whole zone carries out statistics with histogram;
2., to the non-vanishing histogram of brightness value distribution probability, calculate the accumulated value of each brightness value distribution probability and the accumulated value of brightness value and distribution probability product respectively.Formula is following:
Wherein, Hist_Pro [n] is the distribution probability accumulated value of brightness value from (left+1) to (left+n), and left is the left margin in the non-vanishing interval of brightness histogram probability.N=1 ..., length, length are the non-vanishing histogram length of brightness value distribution probability.Hist_Reg [i+left] is the distribution probability of (i+left) for brightness value.
Hist_Vec [n] is distribution probability and the corresponding brightness on duty long-pending accumulated value of brightness value from (left+1) to (left+n).N=1 ..., length, the histogram length that length brightness value distribution probability is non-vanishing.Hist_Reg [i+left] is the distribution probability of (i+left) for brightness value.
3., ask for the threshold value of the corresponding brightness value of maximal value according to following formula as binaryzation:
Hist_square[n]＝(Hist_Vec[length]*Hist_Pro[n])2/Hist_Pro[n]*(1-Hist_Pro[n])
Wherein, Hist_Pro [n] is the accumulated value of brightness value distribution probability, and Hist_Vec [n] is the accumulated value of brightness value and distribution probability product.N=1 ..., length, length are the non-vanishing histogram length of brightness value distribution probability.
3, skeletal extraction:
3*3 neighborhood in the accompanying drawing 2 expression skeletal extraction is labeled as P2 to each neighborhood point, P2, and .., P9 to the point of the white pixel after each binary conversion treatment in the image, if 1. 2. 3. 4. four conditions are satisfied simultaneously below, then deletes current point P1.Repeat this step, all point is all till the unsuppressible-suppression in image.
①2＜＝NZ(P1)＜＝6
Wherein NZ (P1) remarked pixel point P1 8 neighborhood territory pixel points are the number of white pixel point.
②Z0(P1)＝1
Wherein, Z0 (P1) expression begins to judge that from pixel P2 current pixel point and next pixel increase by 1 if 0 and 1 combination is then counted shown in the ring of accompanying drawing 3 left sides, and up to getting back to point of origin P 2, Z0 (P1) result is this count value.
3. P2*P4*P8=0 or a Z0 (P1)!=1
Wherein, P2, P4, P8 represent the pixel value of three neighborhood points of current pixel point shown in Figure 2; Z0 (P1) expression begins to judge that from pixel P2 current pixel point and next pixel increase by 1 if 0 and 1 combination is then counted shown in the ring of accompanying drawing 5 left sides; Up to getting back to point of origin P 2, Z0 (P1) result is this count value.
4. P2*P4*P6=0 or a Z0 (P4)!=1
Wherein, P2, P4, P6 represent the pixel value of three neighborhood points of current pixel point shown in Figure 2; Z0 (P4) expression begins to judge that from pixel P3 current pixel point and next pixel increase by 1 if 0 and 1 combination is then counted shown in the ring of accompanying drawing 3 the right; Up to getting back to point of origin P 3, Z0 (P4) result is this count value;
4, framework characteristic extracts:
Use the directional diagram of 8 neighborhood search in the accompanying drawing 4 expression skeletal extraction processes.
Seek initial skeleton pixel: each character zone from top to bottom, is from left to right searched for first character skeleton pixel, find the back to travel through skeleton clockwise according to 8 neighborhood directions shown in the accompanying drawing 4 from this some beginning.
Traversal skeleton process: begin from first skeleton pixel; Travel through whole character skeleton; If run into a plurality of branches, then carry out the recurrence traversal in the direction of the clock, till each skeleton pixel of character travels through fully; Remove on the character skeleton some because pseudo-framework informations that noise causes, and framework characteristics such as line number crossed in real-time statistics character flex point, crunode, closed loop, skeleton direction, end points and endpoint location information, character.
1., remove pseudo-framework information: in the skeleton ergodic process, the directivity branch of removing subbranch (length of subbranch is below threshold value) and removing non-character skeleton.
2., extract the skeleton direction character: 8 directional diagrams shown in accompanying drawing 4, in the skeleton ergodic process, add up the skeleton traversal direction of each character respectively.
3., extract framework characteristic: in the skeleton ergodic process,, confirm end points, flex point or the crunode of character skeleton according to judging that current skeleton pixel is take-off point or end points; Set out whether get back to the closed loop that initial point is judged character skeleton according to take-off point; Make up the positional information that the up and down position of each end points in character picture obtains end points.
4., the line number crossed in the statistics character: respectively to each character skeleton laterally with row to center, 1/4 position and 3/4 position travel through and added up the line number.
5, the characteristic matching degree calculates:
To the framework characteristic that extracts, carry out matching degree respectively from closed loop number 0,1,2 and calculate.
1., the closed loop number is 2, then respectively end points number, crunode number, horizontal 1/4 are crossed line number and horizontal 3/4 and cross the line number and judge that meeting then increases matching degree respectively.
2., the closed loop number is 1, then respectively end points number, crunode number, horizontal 1/4 are crossed line number and horizontal 3/4 and cross the line number and judge that meeting then increases matching degree respectively.
3., the closed loop number is 0, then respectively end points number, endpoint location information, horizontal 1/4 are crossed line number and horizontal 3/4 and cross the line number and judge that meeting then increases matching degree respectively.
6, character recognition:
Framework characteristic to extracting carries out character recognition respectively from closed loop number 0,1,2.
1., the closed loop number is 2, then cross line number and horizontal 3/4 and cross the line number and be identified as character " 8 " from end points number, crunode number, horizontal 1/4.
2., the closed loop number is 1, then cross the difference that line number and horizontal line 3/4 cross the line number and be identified as " 0,6,9 " character respectively from end points number, horizontal 1/4.
3., the closed loop number is 0, then from endpoint location, x, the y axle is crossed line number, horizontal 1/4 and 3/4 and is crossed line number and chain code information and differentiate and be " 1,2,3,4,5,7 " character.
A kind of adaptive dial digital identification device:
Accompanying drawing 1 has been showed the concrete structure of adaptive dial digital identification device of the present invention.Native system comprises cmos imaging unit, graphics processing unit, light filling control module, image identification unit and five parts of communications interface unit.
Wherein, The cmos imaging unit links to each other with graphics processing unit, image identification unit and communications interface unit successively; Graphics processing unit links to each other with the light filling control module, and the light filling control module links to each other with the cmos imaging unit, and image identification unit links to each other with the cmos imaging unit.
Graphics processing unit outputs results to the light filling control module, and the light filling control module is controlled the picture quality that the cmos imaging unit is gathered according to the result of graphics processing unit, and image identification unit output result controls the order of cmos imaging unit images acquired.Wherein, cmos imaging unit, graphics processing unit and light filling control module have constituted image-pickup device.
Cmos imaging unit: comprise the cmos imaging module and the transmission part that adopt VGA resolution; Reference object is reached photo-sensitive cell and converted it into the pixel by lens unit is the electronic signal of unit, and exports the dsp processor of this signal to image identification unit.This device can adopt the CMOS camera product of the 8014KA of Ningbo Sunny Optical Technology Company Limited model to realize.
Graphics processing unit: input picture is carried out Filtering Processing, to brightness of image and Contrast Detection, confirm whether picture quality satisfies the requirement of digit recognition, output is the result give the light filling control module.Algorithm flow chart is shown in accompanying drawing 5.
1., image filtering is handled: input picture is carried out medium filtering handle;
2., brightness value statistics with histogram: the pixel brightness value to entire image carries out statistics with histogram;
3., picture contrast detects: analyze the characteristics of image of brightness histogram, judge through the peak value distance of brightness histogram whether characteristics of image is that crest is outstanding, the figure kine bias is bright or the figure kine bias is dark.This contrast results is outputed to the light filling control module.
Light filling control module: adopt LED light filling parts, whether control the needs light filling according to the output result of graphics processing unit.LED light filling device can adopt the LED light emitting diode to realize.
Image identification unit comprises dsp processor, is responsible for the realization of dial digital automatic identification algorithm.At first, this unit controls image-pickup device images acquired is carried out character zone location and location, character exact position, and the character zone rationality is judged; Then, control image-pickup device continuous acquisition image is handled each character respectively at the character zone of having good positioning, and correctly sets up the Character mother plate of current dial plate type; At last, the dial plate image of gathering to be identified is discerned, numeric results is outputed to peripheral hardware through communications interface unit.
Accompanying drawing 6 is the wired circuit figure of image identification unit and other devices, and wherein, U1 is for can adopt the BLACKFIN of ADI company series DSP chip, and PPICLK/PPID0-D7 is a video port, accepts video data and to DSP, handles.SCL, SDA are the I2C bus, carry out image through this bus configuration CMOS camera and read.UART_TX/UART_RX is the serial type interface, and this interface signal converts the RS485 signal and the host computer that are fit to long Distance Transmission into through the MAX485 chip and communicates.U2 can adopt the MAX485 chip of MAXIM company, is used for converting the rs 232 serial interface signal that dsp chip produces into satisfy the RS485 agreement differential level signal, to satisfy the needs of long Distance Transmission.The 8014ka chip that U3 can adopt Ningbo Sunny Optical Technology Company Limited to produce, main effect are to gather vision signal, and the digital format that the digital video signal of gathering converts bt656 into transferred among the DSP further handle.
Communications interface unit: the serial communication, communication protocol and the data layout that comprise standard.Serial communication, like RS485, RS232, modes such as SPI can directly link to each other with miscellaneous equipment or module.
Dial digital identification device with above equipment disposition can identify the dial digital result adaptively, and the character picture of result and jpeg format is outputed to peripheral hardware.
A kind of dial digital identification method of using above-mentioned dial digital identification device shown in the algorithm flow chart of the algorithm frame figure of accompanying drawing 7 and accompanying drawing 8, may further comprise the steps:
1, location character is regional automatically:
Because the character distributing position of different dial plate images is inconsistent; Automatically orient the character zone position of dial plate through certain algorithm according to different dial plate types; This positional information is used for character modeling and character recognition, can reduces because environmental baseline such as light change the influence that causes that character position disturbs.And; Because the color of kinds of characters is different with the distribution of light position; Through the exact position that certain algorithm is oriented each character automatically, can carry out more accurate threshold process respectively to each character zone, reduced because the influence that factors such as character light variation cause.Because cmos imaging module focal length is unadjustable, handles respectively for big character and small characters.
The invention provides a kind ofly, confirm the method for character zone position and character exact position by following mode according to continuous multiple frames dial plate image and preset reasonable regional extent.
The character coarse positioning 1.1 the frame difference adds up: continuous multiple frames dial plate image is done the frame difference add up, remove because the zonule that influences such as noise produce; According to the position relation of camera and dial plate, confirm the zone boundary by regional symmetry; Then regional extent is carried out rationality and judges, and if do not meet preset range would forward to 1.1 once more reading image carry out zone location; If meet preset range then classify to confirm the exact position of character, if big character jumps to 1..2, if small characters jumps to 1.3 according to character types.
If 1.2 be big character, like gas meter, flow meter:
1.2.1 region adaptivity binaryzation: read the dial plate image, at first isolate blackaandawhite characterumay zone and red wrongly written or mispronounced character symbol zone to gas meter, flow meter dial plate feature RGB color difference; Respectively two kinds of character zones are carried out adaptive image binaryzation and handle, and remove in the binary image because the zonule that influences such as noise produce;
1.2.2 character normalization;
1.2.3 skeleton process: normalized image is carried out skeleton process: promptly each character is carried out skeletal extraction, framework characteristic extracts, and the characteristic matching rate is calculated, identification character.
1.2.4 character is accurately located: if the skeleton discrimination of alphabet all reaches 100% in the character zone, be judged as the character position of correctly cutting apart, and relevant character exact position and the character number of record.If do not meet then jump to 1.2.1 and carry out aforesaid operations till the character exact position is confirmed.
If 1.3 be small characters, like water meter: read the dial plate image, at first extract the gray-scale value of character zone, with the edge of Canny operator extraction character zone; The exact position that is partitioned into character according to the character independence and the distributing homogeneity at edge, and judge the rationality of this split position.If do not meet then jump to 1.3 and carry out aforesaid operations till the character exact position is confirmed.
2, set up Character mother plate:
The accuracy of character modeling has determined the accuracy rate of Character mother plate coupling identification, and traditional template matching method has been ignored the different of font architecture type and character distributing position, has reduced the Character mother plate matching rate.Because the charcter topology type of different dial plates is different; And the distributing position Different Effects of template and current character the accuracy rate of template matches; Through certain algorithm current dial plate is carried out real-time character modeling; And the character distributed intelligence of logging template, can improve the accuracy rate of template matches recognition methods.
The invention provides a kind ofly, confirm the circulate method of the Character mother plate of setting up 0-9 of character skeleton coupling recognition result by following mode according to the character zone positional information.
2.1 character self-adapting binaryzation: each character zone is carried out adaptive image binaryzation handle, and remove in the binary image because the zonule that influences such as noise produce;
2.2 character normalization;
2.3 normalized image is carried out skeleton process: promptly each character is carried out skeletal extraction, framework characteristic extracts, and the characteristic matching rate is calculated, identification character.
2.4 set up Character mother plate: determine whether the character identification result that belongs to correct according to the size of characteristic matching rate, if then the pixel of this character zone is write the corresponding characters template.Do not finish if Character mother plate is set up, jump to step 2.1, finish up to Character mother plate foundation.
3, identification dial digital:
Because the matching accuracy rate of the template matching method of character is lower; And the feature matching method computation complexity disturbs susceptibility than higher than higher and algorithm to external world; Advantage in conjunction with two kinds of methods; At first carry out character match through template matching method, the character that does not reach certain matching rate carries out the framework characteristic coupling again, has improved the accuracy of identification to a certain extent and has reduced computation complexity.
The invention provides the Character mother plate that a kind of character zone positional information that obtains according to step 1 and step 2 obtain, confirm the method for the character identification result of dial plate image by following mode.
3.1 character self-adapting binaryzation: each character zone is carried out adaptive image binaryzation handle, and remove in the binary image because the zonule that influences such as noise produce;
3.2 character normalization;
3.3 whether be half-word character: binary image is carried out zone marker if judging character; Add up each regional boundary information; Judge interregional in the otherness on border, the left and right sides and the otherness of up-and-down boundary; If the otherness on border, the left and right sides judges that less than certain threshold value and from up-and-down boundary two zones occupy the certain proportion of region height, judge that then this position belongs to half-word character state.Otherwise, judge that then this position belongs to the complete character state;
3.4 template matches: if character zone does not have the half-word character, then carry out character match, statistical pixel correct match rate with the original match template.For complete character, if the template matches degree greater than certain threshold value, and other character match rates then export recognition result well below this matching degree, otherwise template matches can not be discerned this character.
3.5 skeleton coupling: for the unsuccessful character of template matches in the complete character zone, carry out the framework characteristic coupling according to 2.3 steps, if the skeleton matching degree surpasses certain threshold value then writes the target area, otherwise this complete character is a unrecognizable character.
3.6 half-word Character mother plate coupling: if character zone has the half-word character; Rebuild matching template; Carry out character match with the reconstruction template; Confirm the reference position of character zone respectively, mate that statistical pixel correct match rate is confirmed the recognition result of half-word character with adjacent two characters in the template.
Claims (1)
1. adaptive dial digital identification method is characterized in that this method may further comprise the steps:
(1) location character zone, it specifically may further comprise the steps:
(1.1) read the dial plate image, continuous multiple frames dial plate image is done the frame difference add up, remove because the zone that noise effect produces; According to the position relation of camera and dial plate, confirm the character zone border by the character zone symmetry, to obtain the coarse positioning of character zone;
(1.2) character zone of coarse positioning is judged, got back to the dial plate image read step in the step (1.1) if character zone does not meet preset range, and carry out the character zone coarse positioning once more; The exact position of classifying to confirm character if character zone meets preset range according to character boundary is if big character gets into step (1.3), if small characters gets into step (1.4);
(1.3) isolate blackaandawhite characterumay zone and red wrongly written or mispronounced character symbol zone according to the difference of character zone RGB color; Respectively the image binaryzation processing is carried out in blackaandawhite characterumay zone and red wrongly written or mispronounced character symbol zone, and removed in the binary image because the zone that noise effect produces; Character zone is carried out normalization to be handled; Each character in the character zone of handling through normalization is carried out skeletal extraction, framework characteristic extraction and characteristic matching rate to be calculated; If the framework characteristic matching rate of alphabet all reaches 100% in the character zone; Be judged as the exact position of the character of correctly cutting apart; And relevant character exact position and the character number of record, otherwise the aforesaid operations that carries out step (1.3) again is till definite character exact position;
(1.4) gray-scale value of extraction character zone is with the edge of Canny operator extraction character zone; The exact position that is partitioned into character according to the character independence and the distributing homogeneity at edge; And judge the rationality of this split position, if be judged as the aforesaid operations that do not meet the rationality condition then carry out step (1.4) again till definite character exact position;
(2) set up Character mother plate, it specifically may further comprise the steps:
(2.1) character self-adapting binaryzation: each character zone is carried out adaptive image binaryzation handle, and remove in the binary image because the zone that noise effect produces;
(2.2) character zone normalization;
(2.3) each character in the normalization character zone is carried out skeletal extraction, framework characteristic extraction and characteristic matching rate and calculate, thus identification character;
The size of the characteristic matching rate that (2.4) calculates according to step (2.3) determines whether to be correct character identification result, if then the pixel of this character zone is write the corresponding characters template;
(3) identification dial digital, it specifically may further comprise the steps:
(3.1) judge whether each character in the character zone is the half-word character;
(3.2) if character zone does not have the half-word character; Then carry out character match with Character mother plate; Statistical pixel correct match rate, if the correct match rate greater than certain threshold value, and the matching rate of other characters is well below this correct match degree; Then export recognition result, otherwise can not discern the character in this character zone;
(3.3) character that can not discern for template matches in the step (3.2) carries out the framework characteristic coupling, if the framework characteristic matching rate surpasses certain threshold value then exports recognition result, otherwise the character in this character zone is a unrecognizable character;
(3.4) if character zone has the half-word character, rebuild matching template, carry out character match with the reconstruction matching template, statistical pixel correct match rate is confirmed the recognition result of half-word character.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010102203746A CN101916364B (en) | 2010-07-06 | 2010-07-06 | Adaptive dial digital identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010102203746A CN101916364B (en) | 2010-07-06 | 2010-07-06 | Adaptive dial digital identification method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101916364A CN101916364A (en) | 2010-12-15 |
| CN101916364B true CN101916364B (en) | 2012-06-06 |
Family
ID=43323873
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2010102203746A Expired - Fee Related CN101916364B (en) | 2010-07-06 | 2010-07-06 | Adaptive dial digital identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101916364B (en) |
Families Citing this family (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103268490B (en) * | 2013-05-30 | 2016-01-13 | 电子科技大学 | A kind of digit recognition method adopting both sides three quant's sign |
| CN103618878A (en) * | 2013-12-04 | 2014-03-05 | 洛阳愿景科技有限公司 | Camera shooting direct reading remote water meter device |
| CN104700092B (en) * | 2015-03-26 | 2018-01-23 | 南京理工大学 | A kind of small characters digit recognition method being combined based on template and characteristic matching |
| CN105118262A (en) * | 2015-08-25 | 2015-12-02 | 武汉理工大学 | Community-oriented high-efficiency remote meter reading method and system |
| CN106709484B (en) * | 2015-11-13 | 2022-02-22 | 国网吉林省电力有限公司检修公司 | Digital identification method of digital instrument |
| CN105931468B (en) * | 2016-07-06 | 2018-09-11 | 成都市红亿科技有限公司 | A kind of method for automatically regulating of the external light compensating lamp of Car license recognition |
| CN106441519B (en) * | 2016-09-20 | 2019-04-02 | 杭州电子科技大学 | A kind of detection method of turn-taking for gas meter, flow meter machinery dial plate |
| CN107301778A (en) * | 2017-07-11 | 2017-10-27 | 深圳市丰巨泰科电子有限公司 | A kind of parking occupancy management system based on Car license recognition |
| CN108052943A (en) * | 2017-12-29 | 2018-05-18 | 杭州占峰科技有限公司 | A kind of instrument character wheel recognition methods and equipment |
| CN109285333A (en) * | 2018-09-05 | 2019-01-29 | 江苏爱尔数字科技有限公司 | A kind of multidimensional data independence wireless acquisition terminal based on Internet of Things |
| CN109147297A (en) * | 2018-09-05 | 2019-01-04 | 江苏爱尔数字科技有限公司 | A kind of local area radio table acquisition terminal based on Internet of Things |
| CN111357007B (en) * | 2018-10-26 | 2024-01-19 | 合刃科技（深圳）有限公司 | Character acquisition method and device |
| CN109714546B (en) * | 2018-12-26 | 2020-10-09 | 呈像科技(北京)有限公司 | Image processing method and device |
| 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 |
| CN109886276B (en) * | 2019-02-18 | 2023-05-09 | 福州视驰科技有限公司 | Half-word judging method for dial rolling digital character |
| 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 |
| CN111507893A (en) * | 2020-04-15 | 2020-08-07 | 广西科技大学 | Image thinning method and system based on synchronous deletable principle |
| CN111461123B (en) * | 2020-06-17 | 2020-12-22 | 台州傲京厨卫有限公司 | Color recognition instrument monitoring device |
| CN116129435B (en) * | 2023-04-14 | 2023-08-08 | 歌尔股份有限公司 | Character defect detection method, device, equipment and storage medium |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100394437C (en) * | 2006-02-28 | 2008-06-11 | 浙江工业大学 | Dynamic character discriminating method of digital instrument based on BP nerve network |
| CN1916933A (en) * | 2006-08-31 | 2007-02-21 | 合肥新思源智能电子有限公司 | Video direct reading type automatic meter reading system, and image processing method |
| CN101055619B (en) * | 2007-05-11 | 2010-12-08 | 刘军海 | Image identification method and device for mechanical instrument window display figure |
-
2010
- 2010-07-06 CN CN2010102203746A patent/CN101916364B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN101916364A (en) | 2010-12-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101916364B (en) | Adaptive dial digital identification method | |
| CN109711325B (en) | Mango picking point identification method | |
| CN108108761B (en) | Rapid traffic signal lamp detection method based on deep feature learning | |
| CN103235938B (en) | The method and system of car plate detection and indentification | |
| CN104715239B (en) | A kind of vehicle color identification method based on defogging processing and weight piecemeal | |
| CN106067023B (en) | Container number and truck number identification system and method based on image processing | |
| CN106203415B (en) | bank card number automatic identification device based on digital image processing | |
| CN103824091B (en) | A kind of licence plate recognition method for intelligent transportation system | |
| CN108520516A (en) | A kind of bridge pavement Crack Detection and dividing method based on semantic segmentation | |
| CN103177445B (en) | Based on the outdoor tomato recognition methods of fragmentation threshold Iamge Segmentation and spot identification | |
| CN102706274B (en) | System for accurately positioning mechanical part by machine vision in industrially-structured scene | |
| CN101980245B (en) | Adaptive template matching-based passenger flow statistical method | |
| CN110309746A (en) | High-grade information security area list data information extracting method without communication interconnection | |
| CN103971126A (en) | Method and device for identifying traffic signs | |
| CN101584624B (en) | Guideboard recognition blind-guide device and method thereof based on DSP | |
| CN102521565A (en) | Garment identification method and system for low-resolution video | |
| CN103440035A (en) | Gesture recognition system in three-dimensional space and recognition method thereof | |
| CN106529461A (en) | Vehicle model identifying algorithm based on integral characteristic channel and SVM training device | |
| CN107133610B (en) | Visual detection and counting method for traffic flow under complex road conditions | |
| CN112733914B (en) | Underwater target visual identification classification method based on support vector machine | |
| CN105069816B (en) | A kind of method and system of inlet and outlet people flow rate statistical | |
| CN107122775A (en) | A kind of Android mobile phone identity card character identifying method of feature based matching | |
| CN106803087A (en) | A kind of car number automatic identification method and system | |
| CN109147351A (en) | A kind of traffic light control system | |
| CN111695373B (en) | Zebra stripes positioning method, system, medium and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120606 Termination date: 20140706 |
|
| EXPY | Termination of patent right or utility model |