CN116578706A - A method of labeling the address of the form through the algorithm of text proximity and address collision - Google Patents

Abstract

This application relates to the technical field of data labeling, and in particular to a method for labeling the address of the bill of lading through text proximity and address collision algorithms, including: obtaining the face bill picture, performing text recognition on the face bill picture, and obtaining the text content and each text content in it coordinate information; based on the text content and the coordinate information of each text content, determine the proximity value of each text in the text content; according to the proximity value of each text in the text content, gather each text in the text content into groups to obtain an array of text fragments; Compare the text fragment array with the preset address comparison library, and determine the text fragment array with the highest similarity as the address text; take the four-corner coordinates of the address text as the address label value of the face sheet image. The technical solution in this application can replace manual work and automatically mark the address of the face sheet pictures, thereby greatly improving the marking efficiency.

Description

A method of labeling the address of the form through the algorithm of text proximity and address collision

technical field

The present application relates to the technical field of data labeling, in particular to a method for labeling the address of a bill of lading through text proximity and address collision algorithms.

Background technique

Machine learning data labeling is the process of labeling metadata such as text and images, and the labeled data will be used to train machine learning models. Different data annotation types are suitable for different annotation scenarios, and different annotation scenarios are for different machine learning application scenarios.

Labeling of face-to-face address data belongs to text classification labeling. Text classification and content classification refer to the task of assigning predefined categories to documents. Sentences or paragraphs in documents can be marked by subject, such as face-to-face address data.

When labeling machine learning face sheets, especially for addresses with high information complexity, cross-line content, and not pure numbers or letters, a lot of manpower is often required for labeling. Using labeling tools to circle the address area not only Inefficiency and waste of manpower.

Contents of the invention

In order to overcome at least to a certain extent the problem of low efficiency and waste of manpower in the related technology of labeling the address of the form by manpower, this application provides a method for marking the address of the form by using the algorithm of text proximity and address collision.

The scheme of this application is as follows:

A method for labeling the address of a bill of lading through text proximity and address collision algorithms, comprising:

Obtaining the face sheet picture, performing text recognition in the face sheet image, and obtaining the text content and the coordinate information of each text content therein;

Based on the text content and the coordinate information of each text content, determine the proximity value of each text in the text content;

According to the proximity value of each word in the text content, each word in the text content is gathered into a group, and an array of text fragments is obtained;

Comparing the array of text fragments with a preset address comparison library, and determining the array of text fragments with the highest degree of similarity as the address text;

Taking the four-corner coordinates of the address text as the address label value of the face sheet picture.

Preferably, the method also includes:

Identify the waybill number in the picture of the waybill;

Call the order center service address library as the default address comparison library according to the waybill number.

Preferably, the method also includes:

Call the standard address library as the default address comparison library.

Preferably, the method also includes:

The order center service address library and the standard address library are combined as a preset address comparison library.

Preferably, the method also includes:

The data in the address comparison library is tiled in the form of a one-dimensional array.

Preferably, each word in the text content is grouped into groups according to the proximity value of each word in the text content, including:

Groups words with the same proximity value into groups.

Preferably, each word in the text content is grouped into groups according to the proximity value of each word in the text content, including:

Group the text whose absolute value of the difference between adjacent values is lower than a preset threshold.

Preferably, the text segment array is compared with a preset address comparison library, including:

The text segment array is collided with a preset address comparison library based on an address collision algorithm, and the address collision algorithm adopts a method of weighting the proportion and confidence of the same text.

The technical solution provided by this application may include the following beneficial effects: the method for labeling the address of the face sheet through the text proximity and address collision algorithm in this application includes: obtaining the image of the face sheet, performing text recognition on the image of the face sheet, and obtaining the text content in it and the coordinate information of each text content; based on the text content and the coordinate information of each text content, determine the proximity value of each text in the text content; according to the proximity value of each text in the text content, gather each text in the text content into groups, and obtain Array of text fragments; compare the array of text fragments with the preset address comparison library, and determine the array of text fragments with the highest similarity as the address text; take the four-corner coordinates of the address text as the address of the face sheet image label value. The technical solution in this application can replace manual work and automatically mark the address of the face sheet pictures, thereby greatly improving the marking efficiency.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description serve to explain the principles of the application.

Fig. 1 is a schematic flowchart of a method for labeling a bill of lading address through text proximity and address collision algorithms provided by an embodiment of the present application.

Fig. 2 is a schematic diagram of calculating the font distance between two adjacent characters in a method for labeling the address of the form by using the character proximity and address collision algorithm provided by an embodiment of the present application.

Detailed ways

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with aspects of the present application as recited in the appended claims.

Fig. 1 is a schematic flowchart of a method for marking the address of a bill of lading through a text proximity and address collision algorithm provided by an embodiment of the present application. Referring to Fig. 1 , a method for marking a bill of lading address through a text proximity and address collision algorithm includes :

S11: Obtain the face sheet image, perform text recognition on the face sheet image, and obtain the text content and the coordinate information of each text content;

S12: Based on the text content and the coordinate information of each text content, determine the proximity value of each character in the text content;

S13: According to the proximity value of each character in the text content, gather each character in the text content into groups, and obtain an array of character fragments;

S14: Compare the text segment array with a preset address comparison library, and determine the text segment array with the highest comparison similarity as the address text;

S15: Take the four-corner coordinates of the address text as the address label value of the face sheet picture.

It should be noted that the technical solution in this embodiment relates to the technical field of data labeling, and is specifically applied to the labeling of face-to-face address data in the field of machine learning data labeling.

In practice, you can first use CNN (Convolutional Neural Networks, Convolutional Neural Networks) + RNN (Recurrent Neural Networks, Recurrent Neural Networks) neural network to perform text recognition on face-to-face single pictures. The results obtained in this process are meaningless The text content of the face sheet and the coordinate information of each text content.

In practice, according to the character proximity algorithm, based on the text content and the coordinate information of each text content, the proximity value of each character in the text content is determined. specific:

Set the four-corner coordinates of each character as (x1, y1), (x2, y2), (x3, y3), (x4, y4), then the font spacing between two adjacent characters is:

Wherein, t1 represents the first character among the two adjacent characters, and t2 represents the second character among the two adjacent characters.

After taking the distance from all adjacent characters from t1 to tn, aggregate according to the proximity distance, it can be known that the font spacing will have a large number of convergent values, assuming this value is K1 to Km, from which the appropriate K value, as the proximity value.

It should be noted that in this embodiment, characters with the same proximity value can be grouped, and characters whose absolute value of the difference between proximity values is lower than a preset threshold can also be grouped into groups, and the formed arrays are A1 to Ax.

It should be noted that the method also includes:

Identify the waybill number in the picture of the waybill;

Call the order center service address library as the default address comparison library according to the waybill number.

It should be noted that the method also includes:

Call the standard address library as the default address comparison library.

It should be noted that the method also includes:

Merge the order center service address library and the standard address library as the default address comparison library.

It should be noted that the method also includes:

The data in the address comparison library is flattened in the form of a one-dimensional array.

In practice, there are two options for address library selection:

One is a locally maintained address library (such as an order center service address library), and the other is a general provincial and urban address library (standard address library). The address library is determined by choosing one or two, and the address The library tiles the data in the form of a one-dimensional array, which is convenient for the next collision.

It should be noted that the array of text fragments is compared with the preset address comparison library, including:

Based on the address collision algorithm, the array of text fragments is collided with the preset address comparison library. The address collision algorithm adopts the weighted method of the proportion and confidence of the same text.

In practice, the address collision algorithm is as follows:

Among them, the C(D) function represents the length of the array D; the L(Ai) function represents the length of the field Ai; T(j,k) represents the jth word of the array Ai and the kth word of the address comparison library Dm The accuracy of is 0 or 1; Pj represents the confidence level of the jth word of Ai; using Mi, the address collision result value of the Ai array can be calculated, and then the text segment array with the highest similarity can be determined.

It can be understood that, in this embodiment, the method for labeling the address of the bill of lading through the algorithm of text proximity and address collision includes: obtaining the picture of the bill of lading, performing text recognition on the picture of the bill of lading, and obtaining the text content and the coordinates of each text content information; based on the text content and the coordinate information of each text content, determine the proximity value of each text in the text content; according to the proximity value of each text in the text content, gather each text in the text content into groups to obtain an array of text fragments; The segment array is compared with the preset address comparison library, and the text segment array with the highest similarity is determined as the address text; the four-corner coordinates of the address text are taken as the address label value of the face sheet image. The technical solution in this embodiment can replace manual work and automatically mark the address of the face sheet picture, thereby greatly improving the marking efficiency.

It can be understood that, the same or similar parts in the above embodiments can be referred to each other, and the content that is not described in detail in some embodiments can be referred to the same or similar content in other embodiments.

It should be noted that, in the description of the present application, terms such as "first" and "second" are used for description purposes only, and should not be understood as indicating or implying relative importance. In addition, in the description of the present application, unless otherwise specified, the meaning of "plurality" means at least two.

Any process or method descriptions in flowcharts or otherwise described herein may be understood to represent modules, segments or portions of code comprising one or more executable instructions for implementing specific logical functions or steps of the process , and the scope of preferred embodiments of the present application includes additional implementations in which functions may be performed out of the order shown or discussed, including in substantially simultaneous fashion or in reverse order depending on the functions involved, which shall It should be understood by those skilled in the art to which the embodiments of the present application belong.

It should be understood that each part of the present application may be realized by hardware, software, firmware or a combination thereof. In the embodiments described above, various steps or methods may be implemented by software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented by any one or combination of the following techniques known in the art: Discrete logic circuits, ASICs with suitable combinational logic gates, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

Those of ordinary skill in the art can understand that all or part of the steps carried by the methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. During execution, one or a combination of the steps of the method embodiments is included.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing module, each unit may exist separately physically, or two or more units may be integrated into one module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software function modules. If the integrated modules are implemented in the form of software function modules and sold or used as independent products, they can also be stored in a computer-readable storage medium.

The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, and the like.

In the description of this specification, descriptions referring to the terms "one embodiment", "some embodiments", "example", "specific examples", or "some examples" mean that specific features described in connection with the embodiment or example , structure, material or characteristic is included in at least one embodiment or example of the present application. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limitations on the present application, and those skilled in the art can make the above-mentioned The embodiments are subject to changes, modifications, substitutions and variations.

Claims (8)

1. A method for labeling the address of the form by text proximity and address collision algorithm, characterized in that it comprises: Obtaining the face sheet picture, performing text recognition on the face sheet image, and obtaining the text content and the coordinate information of each text content; Based on the text content and the coordinate information of each text content, determine the proximity value of each text in the text content; According to the proximity value of each word in the text content, each word in the text content is gathered into a group, and an array of text fragments is obtained; Comparing the array of text fragments with a preset address comparison library, and determining the array of text fragments with the highest degree of similarity as the address text; Taking the four-corner coordinates of the address text as the address label value of the face sheet picture. 2. The method according to claim 1, characterized in that the method further comprises: Identify the waybill number in the picture of the waybill; Call the order center service address library as the default address comparison library according to the waybill number. 3. The method according to claim 2, wherein the method further comprises: Call the standard address library as the default address comparison library. 4. method according to claim 3, is characterized in that, described method also comprises: The order center service address library and the standard address library are combined as a preset address comparison library. 5. The method according to claim 1, wherein the method further comprises: The data in the address comparison library is tiled in the form of a one-dimensional array. 6. The method according to claim 1, characterized in that, according to the proximity value of each word in the text content, each word in the text content is gathered into groups, comprising: Groups words with the same proximity value into groups. 7. The method according to claim 1, wherein, according to the proximity value of each character in the text content, each character in the text content is gathered into groups, including: Group the text whose absolute value of the difference between adjacent values is lower than a preset threshold. 8. The method according to claim 1, wherein comparing the text segment array with a preset address comparison library includes: The text segment array is collided with a preset address comparison library based on an address collision algorithm, and the address collision algorithm adopts a method of weighting the proportion and confidence of the same text.