CN101408937B - Character line positioning method and device - Google Patents

Abstract

The invention discloses a character line positioning method and a device, wherein the method comprises the following steps: obtaining candidate character information according to the image data to be processed; obtaining inter-character inclination angle information according to the candidate character information, wherein the inter-character inclination angle information is used for identifying the position relation between a candidate character and a neighborhood candidate character of the candidate character; and determining a boundary box of the candidate character row according to the information of the inclination angle between the characters so as to finish positioning. By adopting the technical scheme in the embodiment of the invention, during positioning, the information of the inclination angle between the characters is obtained according to the information of the candidate characters, so that the character line with a certain inclination can be accurately positioned on the premise of ensuring better noise resistance.

Description

Method and device for character row positioning

technology neighborhood

The invention relates to the field of image processing, in particular to a method and device for character row positioning.

Background technique

In the automatic character recognition technology, it is a difficulty and a key point of the whole recognition technology to accurately locate the character to be recognized. A complete character recognition process, such as the process of character recognition in the magnetic code character line of the bill, usually includes: binarization of the character area to be recognized (such as the magnetic code character line), further positioning of the character area to be recognized, according to The positioning results segment the character area to be recognized, and perform feature extraction and recognition on the segmented character area, as shown in Figure 1. Positioning has a crucial impact on the accuracy of subsequent segmentation, feature extraction, and recognition.

The positioning methods in character recognition in the traditional magnetic code character line are mainly divided into two types: spatial projection method and spatial clustering method. The difference from the positioning of ordinary character recognition is that the characters in the magnetic code character line are arranged in rows, and usually there are only one or several lines, so the positioning range has obvious regional characteristics. The spatial projection method utilizes the feature that the magnetic code character line has a relatively large projection component in the horizontal direction, and selects an appropriate threshold to determine the vertical spatial position of the magnetic code character line; the spatial clustering method is based on connected domain marks. The density of candidate character centers in the vertical y-axis direction is used to locate the magnetic code character row position.

However, in the process of realizing the present invention, the inventors found that although the space projection method has a fast positioning speed and has certain robustness to the detection of inclined magnetic code character lines, its anti-noise ability is not enough, especially when black bands appear in the search area. In the case of a black block or a black block, it is likely to cause positioning failure; although the spatial clustering method has good anti-noise ability, for a magnetic code character row with a certain degree of inclination, since the center position of each character deviates greatly in the vertical direction , so the detection effect is not good.

Contents of the invention

The technical problem to be solved by the present invention is to provide a method and device for character line positioning. It can immediately ensure a good anti-noise ability, and at the same time accurately locate the character line with a certain degree of inclination.

For this reason, on the one hand, the embodiment of the present invention provides a kind of method of character line location, and method comprises:

Obtaining candidate character information according to the image data to be processed, wherein the candidate character information includes regions of the image data whose number of candidate characters reaches a certain number;

Acquire position information of candidate characters in the candidate character information, where the position information includes the center coordinates of the candidate characters;

Obtaining position information of other candidate characters that meet the neighborhood range condition of the candidate character according to the position information of the candidate character, the neighborhood range condition being included within a certain spatial distance range of the center coordinates of the candidate character;

Acquiring the inter-character inclination angle information according to the position information of the candidate characters and the position information of the other candidate characters;

Determine the bounding box of the character line of the candidate character according to the inter-character inclination angle information, so as to complete the positioning.

On the other hand, an embodiment of the present invention provides a device for character line positioning, including:

A character information acquisition unit, configured to obtain candidate character information according to the image data to be processed, the candidate character information including areas of the image data where the number of candidate characters reaches a certain number;

A first position information module, configured to acquire position information of candidate characters in the candidate character information, where the position information includes the center coordinates of the candidate characters;

The second position information module is used to obtain the position information of other candidate characters that meet the neighborhood range conditions of the candidate characters according to the position information of the candidate characters, and the neighborhood range conditions include the coordinates of the center coordinates of the candidate characters within a certain distance;

an inclination information module, configured to acquire the inter-character inclination angle information according to the position information of the candidate characters and the position information of the other candidate characters;

The positioning unit is configured to determine the bounding box of the character line of the candidate character according to the inter-character inclination angle information, so as to complete the positioning.

In the technical solution adopted in the embodiment of the present invention, when positioning, the inclination angle information between characters is also obtained according to the candidate character information, so that the character with a certain inclination can be accurately positioned on the premise of ensuring better noise resistance. character line.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without any creative effort.

Fig. 1 is the schematic flow chart of existing automatic character recognition technology;

Fig. 2 is a specific schematic flow chart of the method for character line positioning in the embodiment of the present invention;

Fig. 3 is another specific flowchart of the method for character line positioning in the embodiment of the present invention;

Fig. 4 is a schematic diagram of an image that needs to be positioned;

Fig. 5 is a schematic diagram of the result of positioning;

FIG. 6 is a schematic diagram of a specific composition of the character line positioning device in the embodiment of the present invention;

Fig. 7 is a schematic diagram of a specific composition of the inclination acquisition unit in Fig. 6;

Fig. 8 is a schematic diagram of a specific composition of the positioning unit in Fig. 6;

FIG. 9 is a schematic diagram of another specific composition of the positioning unit in FIG. 6 .

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

Embodiments of the present invention will be described below with reference to the accompanying drawings. Referring to FIG. 2 , it is a schematic flow chart of a method for character line positioning in an embodiment of the present invention. The process includes:

201. Obtain candidate character information according to image data to be processed. Specifically, it may include performing binarization on the image data to be processed to obtain a binary image; performing connected domain analysis on the binary image to obtain candidate character information, such as performing connected domain marking on the binary image, and selecting qualified Connected components are used as candidate characters, and the center coordinates of the candidate characters are calculated. It is also possible not to perform binarization on the image to be processed, but to directly use information such as grayscale or color on the grayscale image or color image to obtain candidate character information.

Wherein, the candidate characters may include multiple characters. For example, when the number of candidate characters obtained reaches a certain number, it is determined that positioning is required and subsequent steps are performed; otherwise, the image data acquisition area is changed and the image data is acquired again.

202. Obtain inter-character inclination angle information according to the candidate character information, where the inter-character inclination angle information is used to identify a positional relationship between a candidate character and a neighboring candidate character of the candidate character. Specifically, it may include:

A. Obtain the position information of the candidate characters in the candidate character information, such as the center coordinates of the candidate characters;

B. According to the position information of the candidate characters, obtain the position information of other candidate characters that meet the neighborhood range conditions of the candidate characters, such as calculating other neighbors within a certain spatial distance of the coordinates according to the center coordinates of the above-mentioned candidate characters candidate characters. That is, the neighborhood range is a neighborhood range based on spatial measurement distance, such as a circular neighborhood range based on Euclidean distance, or a neighborhood range based on block distance, or a neighborhood range based on other distance calculation methods.

C. Obtain the inter-character inclination angle information according to the position information of the candidate characters and the position information of the other candidate characters.

203. Determine a bounding box of a character line of the candidate character according to the inter-character inclination angle information, so as to complete positioning. Wherein, the inter-character inclination angle information includes inter-character inclination angle; the bounding box may be a rectangular box.

Specifically, the inclination angle can be determined according to a statistical method, such as: obtaining at least one inter-character inclination angle related to the candidate character; quantifying the value range of the inclination angle between each of the characters; statistically obtaining the inclination angle falling into each quantization interval The number of inclination angles between characters within; obtain the said number of quantization intervals according to statistics to determine the character row inclination angle; determine the bounding box of the candidate character according to the character row inclination angle, that is, determine the character row inclination angle according to the above The character space position information corresponding to it determines the boundary of the character line of the candidate character.

It is also possible to determine the inclination angle according to the method of clustering, such as: obtain at least one inter-character inclination angle related to the candidate characters; perform clustering according to each of the inter-character inclination angles, and use the category center obtained by clustering as a determination The inclination angle of the character row; determine the bounding box of the character row of the candidate character according to the determined inclination angle of the character row.

As shown in FIG. 3 , it is a schematic flowchart of a method for locating a character line in an embodiment of the present invention when applied to locating a character line of a magnetic code of a bill. The process includes:

301. Select a partial area in the entire bill image as an area for positioning search, and then perform binarization on the image data in the area to obtain a binary image. Usually the magnetic code character lines in the bill are all under the bill, therefore, the lower part (such as the lower 1/5) area of the bill image can be selected as the search area.

302. Perform connected domain analysis on the binary image obtained in the previous step, and mark the connected domain according to the analysis result, so as to record the position and size of each connected component. For example, connected domain analysis can be performed according to the fast region labeling algorithm.

303. In the binary graph marked with connected domains, select connected components that meet the width and height conditions of the magnetic code characters as candidate characters according to a predetermined width and height (or width and height range). For example, determine a predetermined width and height or a width and height range according to some prior information to determine candidate characters, that is, it is known in advance that the size range of a certain character to be positioned should be between the number of pixels and the number of pixels between the width and height.

304, count whether the total number of candidate characters is greater than the predetermined number, if the number of candidate characters is greater than the predetermined number, then consider that the magnetic code character row exists, and perform step 305; otherwise, then consider that the magnetic code character row does not exist, exit positioning, or Exit this positioning, re-execute step 301, and select the next search area for positioning until the search is completed.

305. Calculate the center coordinates of all candidate characters.

306. Calculate an inter-character inclination angle according to the center coordinates of the candidate characters. For example, calculate the angle of the line connecting the center coordinates of each candidate character and the center coordinates of each candidate character in its neighborhood, where the neighborhood range can be a range based on Euclidean distance, for example, if the center of the candidate character The coordinates are (x ₀ , y ₀ ), then the neighborhood range based on Euclidean distance centered on it can be defined as:

RingNeibour(x ₀ , y ₀ )={(x, y)|T1<f(x, y, x ₀ , y ₀ )<T2} (Formula 1)

$f(x,\mathrm{the\; y},x_0,{\mathrm{the\; y}}_0)=\sqrt{{\left({x-x}_0\right)}^2+{\left({\mathrm{the\; y}-\mathrm{the\; y}}_0\right)}^2}$ (Formula 2)

Wherein, T1 may take three to four times the average width of candidate characters, and T2 may take five to six times the average width of candidate characters. Candidate characters satisfying the above definition of the neighborhood range are selected as neighborhood candidate characters to calculate the tilt angle.

Wherein, the above-mentioned neighborhood range may also be based on other example definition methods, such as block distance and the like. That is, the distance formula in Equation 2 can also be defined as:

Distance formula 1: f(x, y, x ₀ , y ₀ )=|xx ₀ |+|yy ₀ |

Distance formula 2: f(x, y, x ₀ , y ₀ )=Max(|xx ₀ |, yy ₀ |)

Distance formula 3: f(x, y, x ₀ , y ₀ )=Min(|xx ₀ |, yy ₀ |)

...

For each candidate character, or part of the candidate characters selected according to the rules, the inter-character inclination angles between it and each candidate character in the neighborhood are calculated respectively. Since the multiple inclination angles obtained by the above calculation may be different, it is necessary to obtain an inclination angle closest to the real value through statistics. Specifically, various probability and statistics methods can be used.

The statistical method is exemplified as follows: determine the possible range of an inclination angle, such as being determined as [-Max, Max], wherein the value of Max is 10, quantify this angle range, and the quantization step length t value is 0.1, and the above statistics are obtained The number of multiple inclination angles that fall into each quantization step, and determine the angle corresponding to the quantization step with the largest number of inclination angles to be the inclination angle of the magnetic code character row.

For example, an angle counter can be defined with the above-mentioned range and quantization order to count each tilt angle. If the angle value falls within a certain step area of the counter, the count of the corresponding step area will be incremented by one unit, and the angle value will be recorded at the same time The information of the corresponding candidate character (such as the number of the candidate character), count the angle with the largest counting unit, if the count value is less than a certain threshold, it is considered that the magnetic code character line does not exist, and exit; otherwise, the angle is considered to be the magnetic code Character line tilt angle.

Also can adopt the method of clustering to determine inclination angle, as, all angle values are clustered, and the inclination angle corresponding to the center of the largest category obtained by clustering is used as the inclination angle of the magnetic code character line, that is, found by clustering Pick a category with the largest count among the angle categories of , and take the angle value of the center of the category as the final result. . For example, clustering is performed in the following clustering space: the clustering space has two coordinate axes, the horizontal axis reflects the angle value information, and the vertical axis reflects the count corresponding to the angle value. The specific clustering method can use various clustering algorithms such as kmeans (k-means) clustering and ISODATA clustering.

307. According to the determined inclination angle of the character row, use the least squares method to fit the upper boundary line and the lower boundary line of the character row of the candidate character corresponding to the angle value, and complete the positioning.

In the scheme provided by the embodiment of the present invention, according to the spatial arrangement characteristics of the magnetic code character row (the center of the magnetic code character is basically closely arranged in segments on a straight line), the calculated candidate character and the inclination between the adjacent characters are used The angle estimates the inclination angle of the character row of the magnetic code, so as to accurately locate the character row of the magnetic code with a certain inclination angle. Wherein, when determining the inclination angle of the character row of the magnetic code, the inclination angle of the entire character row of the magnetic code is calculated by voting at each angle, which improves the stability of determining the inclination.

At the same time, in order to avoid the fact that the distance between adjacent magnetic code characters is too close, the deflection of the magnetic code character row cannot be truly reflected, so the deflection of each magnetic code character and its ring-shaped adjacent magnetic code characters is investigated. The inclination angle between characters can more truly reflect the inclination degree of the entire magnetic code character row. Improved positioning stability and accuracy.

As shown in Fig. 4 and Fig. 5, it is the situation of positioning the actual bill by adopting the solution in the embodiment of the present invention. Figure 4 shows the images that need to be positioned. The rectangular frame in Fig. 5 is the result of positioning. It can be seen that for the case where the magnetic code character line in the image is tilted, the solution in the embodiment of the present invention can still be accurately positioned.

Correspondingly, the embodiment of the present invention also provides a character line positioning device, as shown in Figure 6, the device includes:

The character information acquisition unit 2 is used to obtain candidate character information according to the image data to be processed; the inclination acquisition unit 4 is used to obtain inter-character inclination angle information according to the candidate character information, and the inter-character inclination angle information is used to identify candidate characters. The positional relationship between the character and the candidate character in the vicinity of the candidate character; the positioning unit 6 is used to determine the bounding box of the character row of the candidate character according to the inclination angle information between the characters, so as to complete the positioning.

Wherein, as shown in FIG. 7 , the inclination acquisition unit 4 includes: a first position information module 40 for acquiring position information of candidate characters in the candidate character information; a second position information module 42 for The position information of the character obtains the position information of other candidate characters that meet the neighborhood range conditions of the candidate characters; the inclination information module 44 is used to obtain the position information of the candidate characters according to the position information of the candidate characters and the position information of the other candidate characters. The inclination angle information between the above characters.

As shown in FIG. 8 , the positioning unit 6 includes: an acquisition module 60 for acquiring at least one inter-character inclination angle related to the candidate character in the inter-character inclination angle information; a quantization module 62 for Quantify the value range of the inclination angle between the characters; the statistical module 64 is used to obtain the number of inclination angles between characters falling into each quantization interval; the angle determination module 66 is used to obtain the said quantization interval according to the statistics The number determines the inclination angle of the character line; the boundary determination module 68 is used to determine the bounding box of the character line of the candidate character according to the determined inclination angle of the character line, so as to complete the positioning.

Or, as shown in FIG. 9, the positioning unit 6 includes: an acquisition module 60 for acquiring at least one inter-character inclination angle related to the candidate character in the inter-character inclination angle information; a clustering module 61 for Carry out clustering according to the slant angle between each described character, take the center of the largest category obtained by clustering as the character line slant angle; Boundary determination module 68 is used for determining the character line of the candidate character according to the determined character line slant angle Bounding box to complete the positioning.

Wherein, other specific details of the above-mentioned modules and units are consistent with those in other embodiments of the present invention, and will not be repeated here.

In the scheme provided by the embodiment of the present invention, according to the spatial arrangement characteristics of the magnetic code character row (the center of the magnetic code character is basically closely arranged in segments on a straight line), the calculated candidate character and the inclination between the adjacent characters are used The angle estimates the inclination angle of the magnetic code character row, so as to accurately locate the magnetic code character row with a certain inclination angle. Wherein, when determining the inclination angle of the character row of the magnetic code, the inclination angle of the entire character row of the magnetic code is calculated by voting at each angle, which improves the stability of determining the inclination.

At the same time, in order to avoid the fact that the distance between adjacent magnetic code characters is too close, the deflection of the magnetic code character row cannot be truly reflected, so the deflection of each magnetic code character and its ring-shaped adjacent magnetic code characters is investigated. The inclination angle between characters can more truly reflect the inclination degree of the entire magnetic code character row. Improved positioning stability and accuracy.

Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general-purpose hardware platform, and of course also by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the prior art can be embodied in the form of software products, and the computer software products can be stored in computer-readable storage media, such as ROM/RAM, magnetic discs, optical discs, etc., including several instructions to make a computer device (which may be a personal computer, server, or network device, etc.) execute the methods described in various embodiments or some parts of the embodiments.

The implementation methods described above do not constitute a limitation to the scope of protection of the technical solution. Any modifications, equivalent replacements and improvements made within the spirit and principles of the above implementation methods shall be included in the protection scope of the technical solution.

Claims (8)

1. the method for character row location is characterized in that described method comprises:

Obtain information of candidate characters according to pending view data, comprise in the described information of candidate characters that the candidate characters number reaches the zone of the described view data of some;

Obtain the positional information of candidate characters in the described information of candidate characters, described positional information comprises the centre coordinate of described candidate characters;

Obtain to meet the positional information of other candidate characters of the neighborhood range of condition of described candidate characters according to the positional information of described candidate characters, described neighborhood range of condition is included in the certain space distance range of described candidate characters centre coordinate;

Obtain described intercharacter angle of inclination information according to the positional information of described candidate characters and the positional information of described other candidate characters;

Determine the bounding box of the character row of described candidate characters according to described intercharacter angle of inclination information, to finish the location.

2. the method for claim 1 is characterized in that, described neighborhood scope is the neighborhood scope based on the spatial measure distance.

3. the method for claim 1 is characterized in that, comprises the intercharacter angle of inclination in the information of described intercharacter angle of inclination, describedly determines that according to described intercharacter angle of inclination information the bounding box of described candidate characters comprises:

Obtain at least one intercharacter angle of inclination relevant with described candidate characters;

Span to each described intercharacter angle of inclination quantizes;

Statistics obtains to fall into the number at the intercharacter angle of inclination in each quantized interval;

The described number that obtains each quantized interval according to statistics is determined the character row angle of inclination;

Determine the bounding box of the character row of described candidate characters according to described character row angle of inclination.

4. the method for claim 1 is characterized in that, comprises the intercharacter angle of inclination in the information of described intercharacter angle of inclination, describedly determines that according to described intercharacter angle of inclination information the bounding box of described candidate characters comprises:

Obtain at least one intercharacter angle of inclination relevant with described candidate characters;

Carry out cluster according to each described intercharacter angle of inclination, other center of maximum kind that obtains with cluster is as definite character row angle of inclination;

According to described character row angle of inclination determine described candidate characters the bounding box of character row.

5. as each described method in the claim 1 to 4, it is characterized in that, describedly obtain information of candidate characters according to pending view data and comprise:

Described pending view data is carried out the connected domain analysis obtain information of candidate characters.

6. the device of character row location is characterized in that described device comprises:

The character information acquiring unit is used for obtaining information of candidate characters according to pending view data, comprises in the described information of candidate characters that the candidate characters number reaches the zone of the described view data of some;

The primary importance information module is used for obtaining the positional information of described information of candidate characters candidate characters, and described positional information comprises the centre coordinate of described candidate characters;

Second place information module, be used for obtaining to meet according to the positional information of described candidate characters the positional information of other candidate characters of the neighborhood range of condition of described candidate characters, described neighborhood range of condition is included in the certain space distance range of described candidate characters centre coordinate;

The inclination information module is used for obtaining described intercharacter angle of inclination information according to the positional information of described candidate characters and the positional information of described other candidate characters;

Positioning unit is used for determining according to described intercharacter angle of inclination information the bounding box of the character row of described candidate characters, to finish the location.

7. device as claimed in claim 6 is characterized in that, described positioning unit comprises:

Acquisition module, at least one the intercharacter angle of inclination relevant that is used for obtaining described intercharacter angle of inclination information with described candidate characters;

Quantization modules is used for the span at each described intercharacter angle of inclination is quantized;

Statistical module is used to add up the number that obtains to fall into the intercharacter angle of inclination in each quantized interval;

The angle determination module is used for determining the character row angle of inclination according to the described number of each quantized interval of statistics acquisition;

The border determination module is used for determining according to described character row angle of inclination the bounding box of the character row of candidate characters, to finish the location.

8. device as claimed in claim 6 is characterized in that, described positioning unit comprises:

Acquisition module, at least one the intercharacter angle of inclination relevant that is used for obtaining described intercharacter angle of inclination information with described candidate characters;

The cluster module is used for carrying out cluster according to each described intercharacter angle of inclination, and other center of maximum kind that obtains with cluster is as definite character row angle of inclination;

The border determination module is used for determining according to described character row angle of inclination the bounding box of the character row of candidate characters, to finish the location.

Images