Text recognition method, device and equipment for bills and storage medium

Abstract

The invention discloses a bill character recognition method and device, equipment and a storage medium. The method comprises the steps of obtaining a to-be-recognized bill image; performing anchor point detection and keyword detection on the bill image to obtain an anchor point detection result and a keyword detection result; performing character recognition on the anchor point detection result to obtain an anchor point character recognition result, and performing character recognition on the keyword detection result to obtain a keyword recognition result; and fusing the anchor point character recognition result and the keyword recognition result, and performing character extraction on a fusion result. According to the method and the device, the anchor points and the keywords are separately detected and recognized, so that the conditions that the bills in various places are large in type difference and diversified in style or the existing characters are overlapped and are difficult to recognize can be effectively compatible, a model does not need to be independently trained for the bills in each area, and the bills in multiple areas can be accurately recognized at the same time.

Description

Technical Field

[0001] The present invention relates to the technical field of bill data processing, and in particular to a bill text recognition method, device, equipment and storage medium. Background Art

[0002] Invoice text detection refers to detecting the location information of the key fields that the user wants to extract from an invoice image. With the development of the Internet and financial technology, the number of various types of invoices has grown rapidly, such as train tickets, air ticket itineraries, VAT invoices, bank checks, etc. There are a large number of information fields related to financial services in various types of invoices. In order to improve the management efficiency of information in invoices, invoices need to be digitized. However, invoices often have problems such as text overlap and seal covering. The OCR recognition model trained after annotating the training data using conventional annotation methods has low recognition accuracy for such invoices. In addition, the types of invoices in different places vary greatly. For multiple types of invoices, a recognition model needs to be trained for each type of invoice. There are many recognition models that need to be trained, which is relatively cumbersome. Summary of the invention

[0003] Based on this, it is necessary to provide a method, device, equipment and storage medium for bill text recognition in order to solve at least one of the above technical problems.

[0004] The present invention provides a method for recognizing text on a bill, comprising:

[0005] Obtaining a bill image to be identified;

[0006] Performing anchor point detection and keyword detection on the bill image to obtain anchor point detection results and keyword detection results;

[0007] Performing text recognition on the anchor point detection results to obtain anchor point text recognition results, and performing text recognition on the keyword detection results to obtain keyword recognition results;

[0008] The anchor point text recognition result and the keyword recognition result are fused, and text extraction is performed on the fused result.

[0009] Optionally, according to a method for recognizing text on a bill provided by the present invention, the step of fusing the anchor text recognition result and the keyword recognition result comprises:

[0010] The anchor point character recognition result and the keyword recognition result are fused according to the anchor point coordinate information in the anchor point detection result and the keyword coordinate information in the keyword detection result.

[0011] Optionally, according to a method for text recognition of bills provided by the present invention, before performing anchor point detection and keyword detection on the bill image and obtaining anchor point detection results and keyword detection results, the method further comprises:

[0012] Acquire different types of bill sample sets, wherein any bill sample in the bill sample set is associated with a sample label;

[0013] Input each of the bill samples into a preset detection model to obtain a detection result output by the detection model, wherein the detection result includes an anchor point detection area and a keyword detection area;

[0014] Based on the detection results corresponding to the bill samples and their associated sample labels, the detection model is iteratively trained, wherein the detection model is used to perform anchor point detection and keyword detection on the bill images.

[0015] Optionally, according to a method for text recognition of bills provided by the present invention, the performing of anchor point detection and keyword detection on the bill image to obtain anchor point detection results and keyword detection results includes:

[0016] According to the trained detection model, anchor point detection and keyword detection are performed on the bill image to obtain the position information of the anchor point and the keyword in the bill image;

[0017] The position information corresponding to the anchor point is used as the anchor point detection result, and the position information corresponding to the keyword is used as the keyword detection result.

[0018] Optionally, according to a method for text recognition of bills provided by the present invention, after inputting each bill sample into a preset detection model and obtaining the detection result output by the detection model, the method further comprises:

[0019] Inputting each of the anchor point detection areas into a preset anchor point text recognition model to obtain an anchor point recognition result to be trained, and inputting each of the keyword detection areas into a preset keyword text recognition model to obtain a keyword recognition result to be trained;

[0020] Iteratively training the anchor point character recognition model based on each of the anchor point recognition results to be trained, and iteratively training the keyword character recognition model based on each of the keyword recognition results to be trained;

[0021] Correspondingly, performing text recognition on the anchor point detection results to obtain anchor point text recognition results, and performing text recognition on the keyword detection results to obtain keyword recognition results, including:

[0022] The anchor point text recognition model is used to perform text recognition on the anchor point detection result to obtain the anchor point text recognition result, and the keyword text recognition model is used to perform text recognition on the keyword detection result to obtain the keyword recognition result.

[0023] Optionally, according to a method for text recognition of bills provided by the present invention, after acquiring different types of bill sample sets, the method further includes:

[0024] Preprocessing and data enhancement processing are performed on each of the bill samples.

[0025] The present invention also provides a text recognition device for bills, comprising:

[0026] An acquisition module, used for acquiring a bill image to be identified;

[0027] A detection module, used to perform anchor point detection and keyword detection on the bill image to obtain anchor point detection results and keyword detection results;

[0028] A recognition module, used to perform text recognition on the anchor point detection results to obtain anchor point text recognition results, and to perform text recognition on the keyword detection results to obtain keyword recognition results;

[0029] The fusion module is used to fuse the anchor point text recognition result and the keyword recognition result, and perform text extraction on the fusion result.

[0030] The present invention also provides a computer device, comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, wherein the processor implements the above-mentioned bill text recognition method when executing the computer-readable instructions.

[0031] The present invention also provides one or more readable storage media storing computer-readable instructions, and when the computer-readable instructions are executed by a processor, the above-mentioned bill text recognition method is implemented.

[0032] The above-mentioned bill text recognition method, device, equipment and storage medium include: obtaining a bill image to be recognized; performing anchor point detection and keyword detection on the bill image to obtain anchor point detection results and keyword detection results; performing text recognition on the anchor point detection results to obtain anchor point text recognition results, and performing text recognition on the keyword detection results to obtain keyword recognition results; fusing the anchor point text recognition results and the keyword recognition results, and performing text extraction on the fused results. By adopting separate detection and recognition of anchor points and keywords, the present invention can effectively accommodate the current situation where bills in various places have large differences in types, diverse styles, or overlapping text that is difficult to recognize. There is no need to train a model for bills in each region separately, and bills in multiple regions can be accurately recognized at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0034] Figure 1 It is a flow chart of a method for recognizing text in bills according to an embodiment of the present invention;

[0035] Figure 2 It is a structural schematic diagram of a bill character recognition device in one embodiment of the present invention;

[0036] Figure 3 is a schematic diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] The terms used in one or more embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of the present invention. The singular forms of "a", "said" and "the" used in one or more embodiments of the present invention are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of the present invention refers to and includes any or all possible combinations of one or more associated listed items.

[0039] Invoice text detection refers to detecting the location information of the key fields that the user wants to extract from an invoice image. With the development of the Internet and financial technology, the number of various types of invoices has grown rapidly, such as train tickets, air ticket itineraries, VAT invoices, bank checks, etc. There are a large number of information fields related to financial services in various types of invoices. In order to improve the management efficiency of information in invoices, invoices need to be digitized. However, invoices often have problems such as text overlap and seal covering. The OCR recognition model trained after annotating the training data using conventional annotation methods has low recognition accuracy for such invoices. In addition, the types of invoices in different places vary greatly. For multiple types of invoices, a recognition model needs to be trained for each type. There are many recognition models that need to be trained, which is relatively cumbersome.

[0040] In view of the above problems, the present invention provides the following embodiments. Specifically, in one embodiment, Figure 1 As shown, Figure 1 1 is a flow chart of a method for recognizing text on a bill in an embodiment of the present invention. The present invention provides a method for recognizing text on a bill, comprising the following steps:

[0041] Step S11, obtaining a bill image to be identified;

[0042] It should be noted that the bill image generally refers to a digitized image of various financial bills, transaction vouchers or other paper documents with monetary value. These bills may include but are not limited to checks, bills of exchange, invoices, receipts, stocks, bonds, etc. Specifically, the bill image to be identified is obtained by a scanner, camera or other image capture device.

[0043] Step S12, performing anchor point detection and keyword detection on the bill image to obtain anchor point detection results and keyword detection results;

[0044] It should be noted that in the bill image, the anchor point can be a specific mark, signature, or other unique feature point on the bill. Keyword detection refers to identifying and extracting keywords or text representing bill information in the image. These keywords may be the bill number, date, amount, etc.

[0045] In one embodiment, different types of bill sample sets are obtained, wherein any bill sample in the bill sample set is associated with a sample label. In one embodiment, in order to improve the accuracy of image detection and recognition, after obtaining the bill sample, each of the bill samples is preprocessed and data enhanced. The preprocessing includes grayscale conversion, normalization, denoising and other processing methods, and the data enhancement processing includes rotation, flipping, cropping, scaling and color conversion and other processing methods. Then, each of the bill samples is input into a preset detection model to obtain the detection result output by the detection model, wherein the detection result includes an anchor point detection area and a keyword detection area, wherein the anchor point detection area refers to the text position area of ​​the anchor point in the bill sample, and the keyword detection area refers to the text position area of ​​the keyword in the bill sample. Then, based on the detection results corresponding to the bill samples and the sample labels associated therewith, the detection model is iteratively trained, wherein the detection model is used to perform anchor point detection and keyword detection on the bill image. Then, according to the trained detection model, anchor point detection and keyword detection are performed on the bill image to obtain the position information of the anchor point and the keyword in the bill image, and then the position information corresponding to the anchor point is directly used as the anchor point detection result, and the position information corresponding to the keyword is used as the keyword detection result. In other embodiments, two detection models can also be trained to perform anchor point detection and keyword detection on the bill image respectively.

[0046] Step S13, performing text recognition on the anchor point detection results to obtain anchor point text recognition results, and performing text recognition on the keyword detection results to obtain keyword recognition results;

[0047] It should be noted that, in the process of iteratively training the detection model, each of the anchor point detection areas output by the detection model is input into a preset anchor point text recognition model to obtain the anchor point recognition result to be trained, and each of the keyword detection areas is input into a preset keyword text recognition model to obtain the keyword recognition result to be trained; then, based on each of the anchor point recognition results to be trained and the text labels corresponding to each bill sample, the anchor point text recognition model is iteratively trained, and based on each of the keyword recognition results to be trained and the text labels corresponding to each bill sample, the keyword text recognition model is iteratively trained. Specifically, the anchor point detection result is subjected to text recognition using the anchor point text recognition model to obtain the anchor point text recognition result, and the keyword text recognition model is used to perform text recognition on the keyword detection result to obtain the keyword recognition result. Therefore, by adopting the separate detection and recognition of anchor points and keywords, it is possible to effectively accommodate the current situation where the bill types in various places are very different, the styles are diverse, or the text overlaps and is difficult to recognize, without the need to train a separate model for the bills in each region.

[0048] Step S14, fusing the anchor point text recognition result and the keyword recognition result, and performing text extraction on the fused result.

[0049] In one embodiment, the anchor point detection result is associated with the corresponding text recognition result, that is, the recognized text content and its coordinate information are recorded, and these coordinate information represent the position of the text in the original bill image. Keyword detection usually recognizes important information on the bill, such as the amount, date, etc. The recognition results of these key information also need to be associated with the position information in the original image. Then, according to the anchor point coordinate information in the anchor point detection result and the keyword coordinate information in the keyword detection result, the anchor point text recognition result and the keyword recognition result are fused. More specifically, during the fusion process, the recognized text content can be displayed according to its position in the original bill image. For example, if the recognized amount text is located in the upper right corner of the bill, then the text is displayed in the upper right corner of the image, thereby obtaining a bill image fused with the recognized text. This image not only contains the original bill image, but also displays the recognized text content and its position. Further, the fusion result is subjected to text extraction to extract the required text information in a structured manner.

[0050] The embodiment of the present invention, through the above scheme, includes: obtaining a bill image to be identified; performing anchor point detection and keyword detection on the bill image to obtain anchor point detection results and keyword detection results; performing text recognition on the anchor point detection results to obtain anchor point text recognition results, and performing text recognition on the keyword detection results to obtain keyword recognition results; fusing the anchor point text recognition results and the keyword recognition results, and performing text extraction on the fused results. The embodiment of the present invention, through the use of separate detection and recognition of anchor points and keywords, can effectively accommodate the current situation where bills in various places have large differences in types, diverse styles, or overlapping text that is difficult to identify. There is no need to train a model for bills in each region separately, and bills in multiple regions can be accurately identified at the same time.

[0051] In one embodiment of the present invention, before performing anchor point detection and keyword detection on the bill image and obtaining the anchor point detection result and the keyword detection result, the method further includes:

[0052] Obtain different types of bill sample sets, wherein any bill sample in the bill sample set is associated with a sample label; input each of the bill samples into a preset detection model to obtain a detection result output by the detection model, wherein the detection result includes an anchor point detection area and a keyword detection area; iteratively train the detection model based on the detection results corresponding to the bill samples and their associated sample labels, wherein the detection model is used to perform anchor point detection and keyword detection on the bill image.

[0053] In one embodiment, after inputting each of the bill samples into a preset detection model and obtaining the detection result output by the detection model, the method further includes:

[0054] Input each of the anchor point detection areas into a preset anchor point text recognition model to obtain an anchor point recognition result to be trained, and input each of the keyword detection areas into a preset keyword text recognition model to obtain a keyword recognition result to be trained; based on each of the anchor point recognition results to be trained, iteratively train the anchor point text recognition model, and based on each of the keyword recognition results to be trained, iteratively train the keyword text recognition model.

[0055] Specifically, different types of bill samples are collected, such as invoices, checks, receipts, etc. Each bill sample should have a sample label so as to provide correct output information during the training process. The sample label can indicate the location of the key information of the bill, etc. Each of the bill samples is input into the detection model, and the detection model should be able to output two types of detection areas: anchor detection areas and keyword detection areas. The detection model is trained using the detection results of the bill samples and the corresponding sample labels, and the training process includes adjusting the parameters of the model to reduce the error between the predicted results and the true labels. In addition, each of the anchor detection areas is input into a preset anchor text recognition model to obtain the anchor recognition results to be trained, and each of the keyword detection areas is input into a preset keyword text recognition model to obtain the keyword recognition results to be trained; based on the anchor recognition results to be trained and the text labels of each of the bill samples, the anchor text recognition model is iteratively trained, and based on the keyword recognition results to be trained and the text labels of each of the bill samples, the keyword text recognition model is iteratively trained, wherein the text label is used to indicate the text information of the bill. Therefore, by adopting the separate detection and recognition of anchor points and keywords, we can effectively accommodate the current situation where the types of bills in various places vary greatly, the styles are diverse, or there are overlapping texts that are difficult to recognize, without the need to train a separate model for the bills in each region.

[0056] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0057] In one embodiment, a text recognition device for bills is provided, and the text recognition device for bills corresponds to the text recognition method for bills in the above embodiment. Figure 2 As shown, Figure 2 1 is a schematic diagram of a structure of a text recognition device for bills in one embodiment of the present invention, the text recognition device for bills comprises:

[0058] An acquisition module 21 is used to acquire a bill image to be identified;

[0059] A detection module 22, used to perform anchor point detection and keyword detection on the bill image to obtain anchor point detection results and keyword detection results;

[0060] The recognition module 23 is used to perform text recognition on the anchor point detection results to obtain anchor point text recognition results, and to perform text recognition on the keyword detection results to obtain keyword recognition results;

[0061] The fusion module 24 is used to fuse the anchor point text recognition result and the keyword recognition result, and perform text extraction on the fusion result.

[0062] The fusion module 24 is also used for:

[0063] The anchor point character recognition result and the keyword recognition result are fused according to the anchor point coordinate information in the anchor point detection result and the keyword coordinate information in the keyword detection result.

[0064] The text recognition device of the bill comprises:

[0065] A sample acquisition module, used to acquire different types of bill sample sets, wherein any bill sample in the bill sample set is associated with a sample label;

[0066] A first input module, used for inputting each of the bill samples into a preset detection model to obtain a detection result output by the detection model, wherein the detection result includes an anchor point detection area and a keyword detection area;

[0067] The first training module is used to iteratively train the detection model based on the detection results corresponding to the bill samples and the sample labels associated therewith, wherein the detection model is used to perform anchor point detection and keyword detection on the bill images.

[0068] The detection module 22 is also used for:

[0069] According to the trained detection model, anchor point detection and keyword detection are performed on the bill image to obtain the position information of the anchor point and the keyword in the bill image;

[0070] The position information corresponding to the anchor point is used as the anchor point detection result, and the position information corresponding to the keyword is used as the keyword detection result.

[0071] The text recognition device of the bill comprises:

[0072] A second input module, used to input each of the anchor point detection areas into a preset anchor point text recognition model to obtain an anchor point recognition result to be trained, and to input each of the keyword detection areas into a preset keyword text recognition model to obtain a keyword recognition result to be trained;

[0073] The second training module is used to iteratively train the anchor point text recognition model based on each of the anchor point recognition results to be trained, and iteratively train the keyword text recognition model based on each of the keyword recognition results to be trained.

[0074] The identification module 23 is also used for:

[0075] The anchor point text recognition model is used to perform text recognition on the anchor point detection result to obtain the anchor point text recognition result, and the keyword text recognition model is used to perform text recognition on the keyword detection result to obtain the keyword recognition result.

[0076] The text recognition device of the bill comprises:

[0077] The processing module is used to perform preprocessing and data enhancement processing on each of the bill samples.

[0078] The specific definition of the text recognition device for bills can be found in the definition of the text recognition method for bills above, which will not be repeated here. Each module in the above-mentioned text recognition device for bills can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0079] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, Figure 3 : is a schematic diagram of a computer device in an embodiment of the present invention. The computer device includes a processor, a memory, a network interface and a database connected by a device bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a readable storage medium and an internal memory. The readable storage medium stores an operating device, a computer-readable instruction and a database. The internal memory provides an environment for the operation of the operating device and the computer-readable instructions in the readable storage medium. The database of the computer device is used to store data involved in the text recognition method of the bill. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer-readable instruction is executed by the processor, a text recognition method for the bill is implemented. The readable storage medium provided in this embodiment includes a non-volatile readable storage medium and a volatile readable storage medium.

[0080] In one embodiment, a computer device is provided. The computer device may be a terminal device, and its internal structure diagram may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, and a network interface connected through a device bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a readable storage medium. The readable storage medium stores computer-readable instructions. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer-readable instructions are executed by the processor, a method for text recognition of a bill is implemented. The readable storage medium provided in this embodiment includes a non-volatile readable storage medium and a volatile readable storage medium.

[0081] In one embodiment, a computer device is provided, comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, wherein when the processor executes the computer-readable instructions, the steps of the above-mentioned bill text recognition method are implemented. The method comprises: obtaining a bill image to be recognized; performing anchor point detection and keyword detection on the bill image to obtain an anchor point detection result and a keyword detection result; performing text recognition on the anchor point detection result to obtain an anchor point text recognition result, and performing text recognition on the keyword detection result to obtain a keyword recognition result; fusing the anchor point text recognition result and the keyword recognition result, and performing text extraction on the fused result.

[0082] In one embodiment, a readable storage medium is provided, the readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the text recognition method of the above-mentioned bill are implemented. A person of ordinary skill in the art can understand that the implementation of all or part of the process in the above-mentioned embodiment method can be completed by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a non-volatile readable storage medium or a volatile readable storage medium. When the computer-readable instructions are executed, they can include the process of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0083] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0084] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for recognizing text on a bill, characterized in that: include: Obtaining a bill image to be identified; Performing anchor point detection and keyword detection on the bill image to obtain anchor point detection results and keyword detection results; Performing text recognition on the anchor point detection results to obtain anchor point text recognition results, and performing text recognition on the keyword detection results to obtain keyword recognition results; The anchor point text recognition result and the keyword recognition result are fused, and text extraction is performed on the fused result.

2. The method for recognizing text on a bill according to claim 1, characterized in that: The fusing the anchor character recognition result and the keyword recognition result includes: The anchor point character recognition result and the keyword recognition result are fused according to the anchor point coordinate information in the anchor point detection result and the keyword coordinate information in the keyword detection result.

3. The method for recognizing text on a bill according to claim 1, characterized in that: Before performing anchor point detection and keyword detection on the bill image and obtaining anchor point detection results and keyword detection results, the method further includes: Acquire different types of bill sample sets, wherein any bill sample in the bill sample set is associated with a sample label; Input each of the bill samples into a preset detection model to obtain a detection result output by the detection model, wherein the detection result includes an anchor point detection area and a keyword detection area; Based on the detection results corresponding to the bill samples and their associated sample labels, the detection model is iteratively trained, wherein the detection model is used to perform anchor point detection and keyword detection on the bill images.

4. The method for recognizing text on a bill according to claim 3, characterized in that: The performing anchor point detection and keyword detection on the bill image to obtain anchor point detection results and keyword detection results includes: According to the trained detection model, anchor point detection and keyword detection are performed on the bill image to obtain the position information of the anchor point and the keyword in the bill image; The position information corresponding to the anchor point is used as the anchor point detection result, and the position information corresponding to the keyword is used as the keyword detection result.

5. The method for recognizing text on a bill according to claim 3, characterized in that: After inputting each of the bill samples into a preset detection model and obtaining the detection result output by the detection model, the method further includes: Inputting each of the anchor point detection areas into a preset anchor point text recognition model to obtain an anchor point recognition result to be trained, and inputting each of the keyword detection areas into a preset keyword text recognition model to obtain a keyword recognition result to be trained; Iteratively training the anchor point character recognition model based on each of the anchor point recognition results to be trained, and iteratively training the keyword character recognition model based on each of the keyword recognition results to be trained; Correspondingly, performing text recognition on the anchor point detection results to obtain anchor point text recognition results, and performing text recognition on the keyword detection results to obtain keyword recognition results, include: The anchor point text recognition model is used to perform text recognition on the anchor point detection result to obtain the anchor point text recognition result, and the keyword text recognition model is used to perform text recognition on the keyword detection result to obtain the keyword recognition result.

6. The method for recognizing text on a bill according to claim 1, characterized in that: After obtaining different types of bill sample sets, the method further includes: Preprocessing and data enhancement processing are performed on each of the bill samples.

7. A text recognition device for bills, characterized in that: include: An acquisition module, used for acquiring a bill image to be identified; A detection module, used to perform anchor point detection and keyword detection on the bill image to obtain anchor point detection results and keyword detection results; A recognition module, used to perform text recognition on the anchor point detection results to obtain anchor point text recognition results, and to perform text recognition on the keyword detection results to obtain keyword recognition results; The fusion module is used to fuse the anchor point text recognition result and the keyword recognition result, and perform text extraction on the fusion result.

8. The bill character recognition device according to claim 7, characterized in that: The fusion module is also used for: The anchor point character recognition result and the keyword recognition result are fused according to the anchor point coordinate information in the anchor point detection result and the keyword coordinate information in the keyword detection result.

9. A computer device comprising a memory, a processor, and computer-readable instructions stored in the memory and executed on the processor, characterized in that: When the processor executes the computer-readable instructions, the method for recognizing text on a bill as described in any one of claims 1 to 6 is implemented.

10. A readable storage medium having computer readable instructions stored thereon, characterized in that: When the computer-readable instructions are executed by a processor, the method for text recognition of a bill as described in any one of claims 1 to 6 is implemented.

Images