CN101329730B - Method for identifying symbol image - Google Patents
Method for identifying symbol image Download PDFInfo
- Publication number
- CN101329730B CN101329730B CN2007101094627A CN200710109462A CN101329730B CN 101329730 B CN101329730 B CN 101329730B CN 2007101094627 A CN2007101094627 A CN 2007101094627A CN 200710109462 A CN200710109462 A CN 200710109462A CN 101329730 B CN101329730 B CN 101329730B
- Authority
- CN
- China
- Prior art keywords
- image
- symbol
- identified
- pixel density
- circle
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a symbolic image recognition method. The pixel density values and aspect ratios of a plurality of symbolic images are captured; a plurality of center values and corresponding radius values are calculated by a clustering and splitting algorithm. Then an image to be recognized is captured, and the pixel density value and the aspect ratio of the image are obtained; a comparison is carried out to judge whether the distance between the pixel density value and the aspect ratio and the center value is smaller than the corresponding radius value so as to judge that the image to be recognized is a single symbolic image or a conjoint symbolic image. Contrast data is obtained through the clustering and splitting algorithm; the judgment of the image to be recognized is carried out; contrast complexity can be reduced, thereby greatly increasing contrast speed and the speed of the whole symbolic image recognition process and reducing hardware resources essentially used. In another embodiment, the processes of symbol edge detection and image splitting are added to promote the accuracy of the whole recognition process.
Description
Technical field
The present invention relates to a kind of image recognition method, relate in particular to a kind of method for identifying symbol image.
Background technology
How the symbol data on the Hard copy is sent on the computing machine quickly and accurately, and edited, is the important topic in digital information epoch.Traditionally utilize scanner or other image capture unit more, the Hard copy image is sent to computing machine after, utilize computing machine to carry out identification again, make the symbol data in the image convert the accessible character combination of computing machine to.
But each symbol image in the Hard copy image must be handled in look-ahead analysis, just can carry out identification, otherwise the accuracy of identification will decline to a great extent.The fundamental purpose of this analyzing and processing process is to determine that each symbol image all is single symbol, and the symbol image that comprises a plurality of symbols is cut.Just directly carry out identification if comprise a plurality of continuous symbols in the symbol image, may have the situation of differentiating error and take place.For example the English words www that connects of lines is if directly carry out identification after the scanning, may be because three w hypotelorisms, and cause the situation of identification mistake.Therefore go up in the example and must earlier symbol image be cut into three, in addition identification again is with the accuracy rate of increase identification.
Though have many identifications and dividing method to come the distinguished symbol image whether only to comprise single symbol now, and the symbol image of a plurality of symbols cut.Yet various common image analysing computer processes are accompanied by very complicated calculating usually, and consume too much hardware resource, more are unfavorable for using on the lower portable electronic devices of some processing poweies.
Therefore, how to reduce the complexity of identifying symbol image, and then increase the speed of coded identification process, and reduce the hardware resource that must use, be the target that each manufacturer developed now.
Summary of the invention
Technical matters to be solved by this invention is to provide a kind of method for identifying symbol image, in order to accelerate the speed of whole coded identification process.
According to above-mentioned purpose of the present invention, a kind of method for identifying symbol image is proposed, increase computer identification symbol image speed, reduce the hardware resource that computing machine must use.At first, capture a plurality of symbol images.Calculate the pixel density value and the length breadth ratio of each symbol image.Then the pixel density value and the length breadth ratio of each symbol image are calculated, to form a plurality of circle-center values and corresponding radius value according to the partitioning algorithm that hives off.After above-mentioned steps is finished, capture image to be identified again, calculate the pixel density value and the length breadth ratio of image to be identified.At last, whether the distance that contrasts the pixel density value of image to be identified and length breadth ratio and arbitrary circle-center values is less than the corresponding radius value of circle-center values, if judge that then image to be identified is the image of monadic symbols, if not, judge the image of image to be identified for the symbol that links to each other.
According to another embodiment of the present invention, be used for increasing the speed of computer identification symbol image, reduce the hardware resource that computing machine must use.Its step at first captures a plurality of symbol images, calculates the pixel density value and the length breadth ratio of each symbol image again.Then, according to the partitioning algorithm that hives off the pixel density value and the length breadth ratio of each symbol image are calculated, to produce into a plurality of circle-center values and corresponding radius value.After above-mentioned steps is finished, capture image to be identified again, calculate the pixel density value and the length breadth ratio of image to be identified, whether the distance that then contrasts the pixel density value of image to be identified and length breadth ratio and arbitrary circle-center values is less than the corresponding radius value of circle-center values.If less than the corresponding radius value of circle-center values, then judge the image that this image to be identified is a monadic symbols, if greater than the corresponding radius value of circle-center values, then utilize symbol edge detection analysis image to be identified.After the symbol edge detection analysis, if this image to be identified of symbol rim detection test is the image of the symbol that links to each other, then the image to the symbol that links to each other cuts.
Obtain correlation data by the partitioning algorithm that hives off, and carry out the judgement of image to be identified, can reduce the contrast complexity, and then increase substantially versus speed, increase the speed of whole coded identification process, reduce the hardware resource that must use.Among another embodiment, and add the accuracy that the process of symbol rim detection and image cutting improves whole identification process.
Description of drawings
For above and other objects of the present invention, feature, advantage and embodiment can be become apparent, being described in detail as follows of appended accompanying drawing:
Fig. 1 is the process flow diagram of first embodiment of the invention;
Fig. 2 is a picked image synoptic diagram in the embodiment of the invention;
Fig. 3 is the correlation data synoptic diagram that hives off in the embodiment of the invention;
Fig. 4 is the process flow diagram of second embodiment of the invention;
Fig. 5 is among second embodiment, symbol edge detection analysis process flow diagram;
Fig. 5 is that image is cut apart process flow diagram among second embodiment.
Wherein, Reference numeral:
102～118: step
410～450: step
Embodiment
The various embodiments of the present invention utilization partitioning algorithm that hives off is obtained the symbol image data, and image data to be identified and symbol image data are contrasted, and improves the speed of whole coded identification process, and reduces the loss of hardware resource.Those skilled in the art when variable required parameter, to cooperate various application scenarios, obtain required balance on speed and degree of accuracy in not breaking away from these novel spirit and scope.
First embodiment
Please be simultaneously with reference to Fig. 1, Fig. 2 and Fig. 3.Fig. 1 is the process flow diagram of first embodiment of the invention.Fig. 2 is a picked image synoptic diagram in the embodiment of the invention.Fig. 3 then is the correlation data synoptic diagram that hives off among the embodiment.By the coded identification method of present embodiment, when utilizing the computer identification symbol image, can increase the speed of computer identification symbol, reduce the hardware resource that this computing machine must use.In step 102, capture a plurality of symbol images earlier.Step 104 is for obtaining the pixel density value and the length breadth ratio of these symbol images.Come to be step 106 again, the pixel density value and the length breadth ratio of symbol image are calculated, to produce a plurality of circle-center values (the clustering center that partitioning algorithm was drawn of also promptly hiving off) and corresponding radius value according to the partitioning algorithm that hives off.
Till step 106, for setting up the correlation data of hiving off of a plurality of image symbols.If certain symbol image of hypothesis as shown in Figure 2, this symbol is alphabetical A, and the pixel density value of then symbol image is 16/49, and length breadth ratio is 7/7.Certainly, the pixel density value of distinct symbols image and length breadth ratio can change to some extent with parameter such as its image size and resolution or distinct symbols, only lift a simple case explanation at this.
In order to illustrate that simply how pixel density value and length breadth ratio set up out correlation data, please refer to Fig. 3.Length breadth ratio is set at transverse axis, and density value is set at the longitudinal axis, and then the symbol image of all acquisitions is all signable on this coordinate system.Then handle these data by the partitioning algorithm that hives off again, can obtain a plurality of circle-center values A and B and corresponding radius value R and D.And each point that the circle inside that circle-center values A and B and radius R and D are surrounded is comprised then is the symbol image with similar characteristics.
After setting up out correlation data, step 108 is acquisition image to be identified, and step 110 is for calculating the pixel density value and the length breadth ratio of image to be identified.Then step 112, whether the distance that then is the pixel density value of contrast image to be identified and length breadth ratio and arbitrary circle-center values is less than the corresponding radius value of circle-center values.If, then enter step 114, judge that image to be identified is a continuous symbol image.If not, then enter step 116, judge that image to be identified is the monadic symbols image.
If last judgement image to be identified is the monadic symbols image, can enter step 118 again, image to be identified is carried out the identification action.In present embodiment, whether the pixel density value that contrasts image to be identified and the distance of length breadth ratio and arbitrary circle-center values are 0.7 times less than corresponding radius value, if will improve the accuracy of last identification, can do more harsh restriction to the standard of judging.In addition, the employed partitioning algorithm that hives off is that K organizes average grouping method (K-means grouping method) among the embodiment.
After setting up out the correlation data of hiving off by the partitioning algorithm that hives off, only need image to be identified is obtained pixel density value and length breadth ratio, just can compare, and correlation data quantity is simplified, do not need complex calculation, whether be link to each other symbol image or monadic symbols image, accelerated the process of whole coded identification, and alleviated load of hardware resources if therefore can differentiate image to be identified fast.
Second embodiment
The mentioned control methods of hiving off in the foregoing description also can combine with common coded identification process, as a step of differentiating monadic symbols image and the symbol image that links to each other fast, improves the speed of whole coded identification process.
Please refer to Fig. 4, this is the flow chart of steps of second embodiment.In the step 410, capture image to be identified at the beginning earlier.Step 420 judges for the way of contrast that hives off that uses the foregoing description whether image to be identified is the symbol image that links to each other.If judge not for the symbol image that links to each other, then be the monadic symbols image, enter step 450, carry out the identification action.If be judged to be continuous symbol image, then enter step 430, use whether symbol edge detection analysis image to be identified is the symbol image that links to each other.In the step 430,, then be the monadic symbols image, enter step 450, carry out the identification action if judge not for the symbol image that links to each other.If be judged to be continuous symbol image, then enter step 440, image to be identified is cut.
Wherein, the way of contrast that hives off mentioned in the step 420 describes in detail in the foregoing description, and the rim detection contrast is for existing technology, so its principle and details will not added to give unnecessary details at this.And the process flow diagram of the symbol edge detection analysis in the step 430 please refer to Fig. 5.In step 432, can capture a plurality of symbol images.Step 434 calculates the rim detection correlation data of each symbol image.Step 432 is being set up the rim detection correlation data of each symbol with step 434 purpose.After the rim detection data are set up, enter step 436, obtain the rim detection correlation data of image to be identified.Step 438 is for to compare the rim detection correlation data of image to be identified and the rim detection correlation data of arbitrary symbol image.If judge not for the symbol image that links to each other, then be the monadic symbols image, enter step 450, carry out the identification action.If be judged to be continuous symbol image, then enter step 440, image to be identified is cut.
Because containing each sideline area, the rim detection correlation data manys a feature correlation data than data and length breadth ratio data etc., therefore, if the image edge to be identified of specific ratios detects correlation data contrast back error less than a critical value, judge that then image to be identified is the image of a monadic symbols.If the image edge to be identified of specific ratios detects correlation data contrast back error greater than above-mentioned critical value, judge that then this image to be identified is the image of a continuous symbol.In present embodiment, critical value setting is 10%, that is to say, if after the contrast of the rim detection correlation data of specific ratios, finds that error rate more than 10%, will be judged to be continuous symbol image, if be the monadic symbols image less than 10%.Certainly, critical value can be done other setting in other embodiment, to cooperate different application.
Fig. 6 is in the present embodiment, image cutting process flow diagram.In the step 442, obtain continuous symbol image (being judged as the image to be identified of the symbol that links to each other) pixel projection amount of a plurality of subpoints on X-axis.Step 444 is for to utilize the cut point algorithm computation to go out the cut value of each subpoint.Step 446 is item for to be made as cut point cutting to the corresponding subpoint of maximal value in the cut value.Step 450 is carried out the identification action for the continuous symbol image after will cutting.Wherein, the cut point algorithm is existing technology, so do not given unnecessary details at this.
The first embodiment of the present invention utilization partitioning algorithm that hives off is set up comparison database, and image to be identified is contrasted fast with the way of contrast that hives off, and accelerates the speed of coded identification.Method with first embodiment in second embodiment combines with edge detection analysis, earlier symbol image is done to judge for the first time by the way of contrast that hives off, and determines whether to carry out complexity and the higher edge detection analysis of degree of accuracy at result of determination more afterwards.So when when utilizing computing machine to carry out identifying symbol image, not only taken into account the accuracy of coded identification, more, reduce the computational burden of computing machine, and accelerated the speed of overall symbol identification process and reduced employed hardware resource because reduced the utilization rate of edge detection analysis.
Certainly; the present invention also can have other various embodiments; under the situation that does not deviate from spirit of the present invention and essence thereof; being familiar with those of ordinary skill in the art ought can make various corresponding changes and distortion according to the present invention, but these corresponding changes and distortion all should belong to the protection domain of the appended claim of the present invention.
Claims (9)
1. a method for identifying symbol image in order to increase the speed of a computer identification symbol, reduces the hardware resource that this computing machine must use, and it is characterized in that, comprises at least:
(a) capture a plurality of symbol images;
(b) calculate the respectively pixel density value and the length breadth ratio of this symbol image;
(c) according to the partitioning algorithm that hives off the respectively pixel density value and the length breadth ratio of this symbol image are calculated, to produce into a plurality of circle-center values and corresponding radius value;
(d) acquisition one image to be identified;
(e) calculate the pixel density value and the length breadth ratio of this image to be identified; And
(f) whether contrast the distance of the pixel density value of this image to be identified and length breadth ratio and arbitrary circle-center values less than the corresponding radius value of this circle-center values;
If, then judge the image that this image to be identified is a monadic symbols, and this image to be identified is carried out identification action, and
If not, judge that then this image to be identified is the image of a continuous symbol.
2. method for identifying symbol image according to claim 1, it is characterized in that the contrast step of this step (f) is that the distance of the contrast pixel density value of this image to be identified and length breadth ratio and arbitrary circle-center values is whether less than 0.7 times of the corresponding radius value of this circle-center values.
3. method for identifying symbol image according to claim 1 is characterized in that, this partitioning algorithm that hives off is that a K organizes average grouping method.
4. a method for identifying symbol image in order to increase the speed of a computer identification symbol, reduces the hardware resource that this computing machine must use, and it is characterized in that, comprises at least:
(a) capture a plurality of symbol images;
(b) calculate the respectively pixel density value and the length breadth ratio of this symbol image;
(c) according to the partitioning algorithm that hives off the respectively pixel density value and the length breadth ratio of this symbol image are calculated, to produce into a plurality of circle-center values and corresponding radius value;
(d) acquisition one image to be identified;
(e) calculate the pixel density value and the length breadth ratio of this image to be identified;
(f) whether contrast the distance of the pixel density value of this image to be identified and length breadth ratio and arbitrary circle-center values less than the corresponding radius value of this circle-center values;
(g) if less than the corresponding radius value of this circle-center values, then judge the image that this image to be identified is a monadic symbols, and this image to be identified is carried out identification action, and
If greater than the corresponding radius value of this circle-center values, then utilize this image to be identified of a symbol edge detection analysis; And
(h) if this image to be identified of this symbol rim detection test is the image of a continuous symbol, then the image to this continuous symbol cuts, and the image of the symbol that should link to each other after the cutting is carried out identification action.
5. method for identifying symbol image according to claim 4, it is characterized in that the contrast step of this step (f) is that the distance of the contrast pixel density value of this image to be identified and length breadth ratio and arbitrary circle-center values is whether less than 0.7 times of the corresponding radius value of this circle-center values.
6. method for identifying symbol image according to claim 4 is characterized in that, this partitioning algorithm that hives off is that a K organizes average grouping method.
7. method for identifying symbol image according to claim 4 is characterized in that, in this step (g), utilizes this image to be identified of a symbol edge detection analysis to comprise following substep:
(1) calculates the respectively rim detection correlation data of this symbol image;
(2) calculate the rim detection correlation data of this image to be identified; And
(3) contrast the rim detection correlation data of this image to be identified and the rim detection correlation data of arbitrary this symbol image.
8. method for identifying symbol image according to claim 7, it is characterized in that, respectively the rim detection correlation data of this symbol image comprise this symbol image respectively each sideline area than data and length breadth ratio data, the rim detection correlation data of this image to be identified comprises the sideline area of this image to be identified than data and length breadth ratio data.
9. method for identifying symbol image according to claim 4 is characterized in that, in this step (h), the image of this continuous symbol cut also comprises following substep:
(4) obtain the pixel projection amount of image a plurality of subpoints on X-axis of this continuous symbol;
(5) utilize all cut-vertex algorithms to calculate the respectively cut value of this subpoint; And
(6) the corresponding subpoint of getting in this cut value of maximal value is that cut point cuts.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2007101094627A CN101329730B (en) | 2007-06-21 | 2007-06-21 | Method for identifying symbol image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2007101094627A CN101329730B (en) | 2007-06-21 | 2007-06-21 | Method for identifying symbol image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101329730A CN101329730A (en) | 2008-12-24 |
| CN101329730B true CN101329730B (en) | 2010-09-15 |
Family
ID=40205532
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2007101094627A Expired - Fee Related CN101329730B (en) | 2007-06-21 | 2007-06-21 | Method for identifying symbol image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101329730B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1157975A (en) * | 1995-12-21 | 1997-08-27 | 塞克斯丹航空电子公司 | Method and system for display of symbolic images |
| CN1595425A (en) * | 2004-07-13 | 2005-03-16 | 清华大学 | Method for identifying multi-characteristic of fingerprint |
-
2007
- 2007-06-21 CN CN2007101094627A patent/CN101329730B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1157975A (en) * | 1995-12-21 | 1997-08-27 | 塞克斯丹航空电子公司 | Method and system for display of symbolic images |
| CN1595425A (en) * | 2004-07-13 | 2005-03-16 | 清华大学 | Method for identifying multi-characteristic of fingerprint |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101329730A (en) | 2008-12-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20190188528A1 (en) | Text detection method and apparatus, and storage medium | |
| CN111583180B (en) | Image tampering identification method and device, computer equipment and storage medium | |
| CN108460346B (en) | Fingerprint identification method and device | |
| US9176043B2 (en) | Cell image analysis apparatus, cell image analysis method, and program | |
| CN111967545B (en) | Text detection method and device, electronic equipment and computer storage medium | |
| CN115294332B (en) | Image processing method, device, equipment and storage medium | |
| CN113409555B (en) | Real-time alarm linkage method and system based on Internet of things | |
| CN111027343A (en) | Bar code area positioning method and device | |
| KR101836811B1 (en) | Method, apparatus and computer program for matching between the images | |
| CN112926621A (en) | Data labeling method and device, electronic equipment and storage medium | |
| WO2015031350A1 (en) | Systems and methods for memory utilization for object detection | |
| JP4967045B2 (en) | Background discriminating apparatus, method and program | |
| CN101329730B (en) | Method for identifying symbol image | |
| CN115080745A (en) | Multi-scene text classification method, device, equipment and medium based on artificial intelligence | |
| CN115700758A (en) | Sperm activity detection method, device, equipment and storage medium | |
| CN115098679A (en) | Method, device, equipment and medium for detecting abnormality of text classification labeling sample | |
| CN112016514B (en) | Traffic sign recognition method, device, equipment and storage medium | |
| CN106326882A (en) | Fingerprint identification system and fingerprint identification method based on image quality assessment technology | |
| CN113283396A (en) | Target object class detection method and device, computer equipment and storage medium | |
| CN115311649A (en) | Card type identification method and device, electronic equipment and storage medium | |
| CN111737371A (en) | Data flow detection classification method and device capable of dynamically predicting | |
| CN112733670A (en) | Fingerprint feature extraction method and device, electronic equipment and storage medium | |
| CN106980494B (en) | Method and device for achieving browser capable of intelligently recognizing zoom gesture instructions | |
| CN110991296B (en) | Video annotation method and device, electronic equipment and computer-readable storage medium | |
| CN113435320B (en) | Human body posture estimation method with multiple models configured in self-adaption mode |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20100915 Termination date: 20160621 |