Terminal 3D annular character recognition method and device based on OCR technology, equipment and medium

Abstract

The invention discloses a terminal 3D annular character recognition method, device and equipment based on an OCR technology and a medium, and belongs to the technical field of image recognition, and the method comprises the steps: collecting terminal images in a surrounding manner, storing the images according to a character string sequence, and establishing a character string candidate set document; cutting a terminal area in the image to obtain a corresponding terminal image set; enhancing the picture contrast, and identifying the sequentially arranged images one by one; respectively storing the plurality of identification results in a plurality of lists, and performing end-to-end splicing de-duplication on the character strings in each list to obtain one or more complete character strings; and matching the splicing result with a character string format in the candidate set document, and selecting a character string with a matching degree reaching a set threshold value as a final recognition result. According to the method, the existing OCR method is improved, the difficulty in terminal surrounding character and multi-line character recognition is overcome, the recognition rate is improved, the problems of high working intensity and time and labor waste are solved, and the working efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and in particular to a method, device, equipment and medium for recognizing 3D circular characters on a terminal based on OCR technology. Background Art

[0002] In recent years, OCR technology has seen widespread application. It can extract printed or handwritten text from images or scanned documents and convert it into an editable text format. Applications include healthcare, education, business card scanning, and historical document research. For example, in the medical field, OCR can be used to convert doctors' handwritten prescriptions and medical records into electronic text for easier management and retrieval. Overall, OCR technology has broad applications in various fields, improving the efficiency, accuracy, and accessibility of document processing.

[0003] However, existing OCR technology cannot recognize 3D circular text on circular objects, such as the circular character strings on terminals. The main difficulties in recognition are: the character strings surround the circular terminal, and a single camera cannot capture the entire text information; after multiple cameras capture the entire text information, the text information needs to be deduplicated and spliced in order; and the text information on the edges of the circular object will be blurred and distorted, making recognition even more difficult. Therefore, in the development process of existing terminal products, the circular characters on their surfaces have long relied on manual recognition and manual data entry, resulting in low work efficiency, unstable work quality, and a large consumption of human resources.

[0004] In the prior art, patent CN116030470A discloses a method, device, system and readable storage medium for recognizing round seal curved text. The method includes: obtaining a round seal curved text image, detecting the round seal curved text image through a text detection model, and obtaining text character box coordinates and string area coordinates; processing each single character in the round seal curved text image according to the text character box coordinates and string area coordinates to obtain a single-character positive-sequence splicing sub-graph and a single-character reverse-sequence splicing sub-graph; and identifying the text content in the round seal curved text image based on the single-character positive-sequence splicing sub-graph and the single-character reverse-sequence splicing sub-graph.

[0005] The aforementioned patent applies to the splicing of circular character strings that can be covered by a single image, which can be understood as a 2D circular character string. However, for text wrapped around a circular object, it is a 3D circular character string. A single image cannot cover the entire circular character string, so multiple images are required. However, the method of sorting each character box according to the center coordinates and center coordinates of the aforementioned patent cannot be applied to multiple images. This is because the 3D circular character string in a single image is not a standard circular or fan-shaped one, and the center position of the character box cannot be determined, and thus the sorting method in the aforementioned patent cannot be used. Summary of the Invention

[0006] The present invention aims to solve the problem that the existing technology cannot recognize 3D circular characters on circular objects, and proposes a terminal 3D circular character recognition method, device, equipment and medium based on OCR technology, which can effectively improve the accuracy of 3D circular character recognition in terminals and improve recognition efficiency.

[0007] In order to achieve the above-mentioned object of the invention, the technical solution of the present invention is as follows: A terminal 3D circular character recognition method based on OCR technology, characterized by comprising the following steps: Step S1, collecting terminal images in 360° surround, storing the images in character string order, and creating a character string candidate set document; Step S2: cropping the terminal area in the image to obtain a corresponding terminal image set; Step S3: enhancing the image contrast and performing OCR recognition on the terminal images arranged in sequence one by one; Step S4: storing the multiple recognition results in multiple lists, concatenating the character strings in each list and removing duplicates, to finally obtain one or more complete character strings; Step S5: Match the concatenated result with the preset character string format in the candidate set document, and select the character string whose matching degree reaches the set threshold as the final recognition result; otherwise, the concatenated result is used as the final result.

[0008] Furthermore, the multiple recognition results are stored in multiple lists, and the strings in each list are concatenated and duplicated, ultimately obtaining one or more complete strings, including: Step S401: Obtain the y coordinates of multiple character string positions identified in multiple images; Step S402: Compare the y coordinates of multiple strings in multiple images, add the strings that are greater than a set threshold to the same list, and store the others in another list; Step S403: for multiple character strings in the same list, determine the x-coordinates of each character string text box, and determine whether to delete the first and last characters based on the x-coordinates. Step S404: update the list accordingly according to the result of step S403, and remove duplicates at the beginning and end of the character strings in the same list to obtain a spliced result.

[0009] Furthermore, when performing OCR recognition on the images arranged in sequence one by one, the images are divided into two categories, one is the images that have been cropped and contrast enhanced, and the other is the images that have not been cropped and the contrast is directly enhanced; OCR recognition is performed on the two categories of images respectively, and finally the group of results with a higher recognition rate in the two categories of images is returned as the recognition result.

[0010] Furthermore, a method for removing duplicates and concatenating the strings in the same list is as follows: the similarity between the two strings is calculated using the edit distance, and then a function is designed to use a for loop to iterate the two strings from the beginning and end respectively to compare their local similarities. If the local similarity is greater than a set threshold, the first string is concatenated with the second string with the local part removed.

[0011] Furthermore, in step S404, the character string after the last null character in the same list is used as the starting position for splicing.

[0012] Furthermore, in step S403, the strings in the same list are represented as multiple segments of the same complete string. It is determined whether the x-left coordinate of each string in the list is 0. If it is 0, it is considered that the first character in the string is not accurately recognized and is deleted. It is also determined whether the x-right coordinate of each string in the list is equal to the width of the image where the string is located. If it is, it is considered that the last character in the string is not accurately recognized and is deleted.

[0013] The present invention also proposes a terminal 3D circular character recognition device based on OCR technology, comprising: Image acquisition module, used for 360° surround acquisition of terminal images; The image storage module is used to store the collected terminal images in the order of character strings and establish a character string candidate set document; The image cropping module crops the terminal area in the image to obtain the corresponding terminal image set; Image processing module, used to enhance the contrast of the image and perform OCR recognition on the terminal images arranged in sequence one by one; The concatenation and deduplication module is used to store multiple recognition results in multiple lists, concatenate the strings in each list and remove duplicates, and finally obtain one or more complete strings; The result output module is used to match the splicing result with the string format in the candidate set document, and select the string whose matching degree reaches the set threshold as the final recognition result, otherwise the splicing result is used as the final result.

[0014] Furthermore, the splicing and deduplication module includes: A first recognition unit is used to obtain the y coordinates of the positions of multiple character strings recognized from multiple images; a first data processing unit, configured to compare the y-coordinates of multiple strings in multiple images, add strings whose values are greater than a set threshold to the same list, and store the others in another list; The second data processing unit is used to determine the x-coordinates of each string text box for multiple strings in the same list, and determine whether to delete the first and last characters according to the x-coordinates. The first splicing unit is used to update the list according to the processing result of the second data processing unit, and to remove the duplication at the beginning and end of the character strings in the same list and splice them to obtain a spliced result.

[0015] The present invention further proposes a terminal 3D circular character recognition device based on OCR technology, comprising: memory for storing computer programs; The processor is configured to implement the steps of the above-mentioned terminal 3D circular character recognition method based on OCR technology when executing the computer program.

[0016] The present invention further proposes a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the terminal 3D circular character recognition method based on OCR technology as described above are implemented.

[0017] In summary, the present invention has the following advantages: (1) The present invention improves the existing OCR recognition method, thereby improving the recognition rate and work efficiency of terminal surrounding characters and multi-line characters, saving time and manpower costs in the process of terminal information input, and greatly improving its efficiency and accuracy, and has good promotion and application value; (2) The present invention uses multiple cameras to capture multiple terminal images around 360°. Through multi-view image stitching and string matching algorithms, it ensures the correct recognition and merging of the same string from multiple perspectives, and achieves accurate string reconstruction and recognition. (3) In order to optimize the string splicing process, the present invention rationally classifies the identified strings, thereby simplifying the string splicing process from different perspectives and improving the overall processing efficiency; (4) When identifying different character strings, the present invention selects the first non-empty character string identified as a benchmark for subsequent comparison, and classifies and aggregates the character strings based on the comparison results, which simplifies the subsequent multi-view splicing operation and improves the overall processing efficiency; (5) When the present invention performs head-to-tail splicing and deduplication, the empty string is regarded as a separator between complete strings to identify the start and end positions of the string. The use of the empty string as a separator can help the system accurately identify and extract each substring, thereby achieving effective splicing and reorganization of the complete string; (6) The present invention presets specific formats and specifications based on the regularity of character strings to optimize the character string recognition process and improve recognition accuracy. These customized formats and specifications help the system to more accurately identify and classify target character strings and reduce the possibility of misidentification or incorrect splicing; (7) The present invention still outputs the splicing results that do not meet the set threshold. This processing method ensures the flexibility of the system, can adapt to diverse input data and provide complete recognition results. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the overall process of the present invention; Figure 2 Schematic diagram of the image acquisition device of the present invention; Figure 3 It is a schematic diagram of the overall process of the present invention; Figure 4 are terminal images at different viewing angles in Example 1 of the present invention, Figure 4 (a) is the image taken by camera a from the first perspective, Figure 4 (b) is the image taken by camera b from the second perspective. DETAILED DESCRIPTION

[0019] In order to more clearly illustrate the present invention, the present invention is further described below in conjunction with preferred embodiments and drawings. Those skilled in the art should understand that the following specific description is illustrative rather than restrictive and should not be used to limit the scope of protection of the present invention.

[0020] Example 1 The present invention provides a terminal 3D circular character recognition method based on OCR technology, such as Figures 1 to 3 As shown, the following steps are included: Step S1: Place the terminal into Figure 2 In the image acquisition device shown, the image acquisition device uses four cameras to capture images in a 360° surround; then the images are stored in string order, a string candidate set document is established, and the terminal string format and specifications are entered into the document in advance for subsequent string matching.

[0021] Preferably, in this step, when the terminal is placed in the image acquisition device, the front end displays the terminal image. In order to reduce the error rate, the direction classifier in the OCR technology is not used, and the inverted character terminal image is directly rotated 180° and the image is stored in the order of the character string.

[0022] In this step, the terminal string format specification is a customized rule and standard set by this solution based on specific application scenarios. For example, some strings may begin with a specific character sequence (such as "aBC"). The strings recognized by this solution exhibit certain regularities, so specific formats and specifications can be preset based on these regularities to optimize the string recognition process and improve recognition accuracy. These customized formats and specifications help the system more accurately identify and classify target strings, reducing the possibility of misidentification or incorrect splicing.

[0023] Step S2: pre-process the image.

[0024] Specifically include: Step S201: cropping the terminal area in the image to obtain a corresponding terminal image set; Step S202: Enhance the image contrast.

[0025] Step S3: Perform OCR recognition on the images arranged in sequence one by one.

[0026] Specifically include: The images are divided into two categories: cropped and contrast-enhanced images, and uncropped, contrast-enhanced images. OCR is performed on each category, and the result from the group with the higher recognition rate is returned. It should be noted that OCR can automatically obtain the coordinates of text boxes within a string.

[0027] Step S4: store the multiple recognition results in multiple lists respectively, concatenate the character strings in each list and remove duplicates, and finally obtain one or more complete character strings.

[0028] The splicing technology used in this solution can identify two different character strings on one terminal. Due to the ring structure of the terminal, a single camera cannot capture all the character strings on the entire terminal at once, so multiple cameras are required to shoot from different angles. Figure 4 (a) shows the image captured by camera a from the first perspective. Figure 4 (b) shows the image captured by camera b from a second perspective. By combining the image data from these two different perspectives, the system can fully identify all the string information on the terminal, thereby achieving accurate recognition and splicing of the entire terminal string.

[0029] In addition, there are two character strings on the terminal, one of which is photographed from two different perspectives. Although the character string appears separately from the two perspectives, the system is still able to use stitching technology to accurately align and regularize the same character string from the two perspectives. Therefore, the system can correctly identify the two character strings: "aBC123 dEF456 gHI789" and "ABC123 DEF456 GHI789", rather than mistakenly splicing them into "aBC123 dEF456 gHI789 ABC123DEF456" or "GHI789". Through multi-perspective image stitching and string matching algorithms, the system ensures that the same character string from multiple perspectives is correctly identified and merged, thereby achieving accurate character string reconstruction and recognition.

[0030] Specifically, step S4 includes the following steps: Step S401: Obtain the y coordinates of multiple character string positions identified in multiple images; Step S402: Compare the y coordinates of multiple strings in multiple images, specifically comparing the y coordinate of each string with the y coordinate of the first non-empty string; add strings greater than a set threshold to the same list, and store the others in another list. Strings in the same list are represented as multiple segments of the same complete string; Step S403: for multiple character strings in the same list, determine the x left and right coordinates of each character string text box; Step S404: update the list accordingly according to the result of step S403, and remove duplicates at the beginning and end of the character strings in the same list to obtain a spliced result.

[0031] In the above step S402, considering that the PaddleOCR technology used will recognize empty strings and set the y coordinate of the empty string to 0 by default, in order to optimize the subsequent string splicing processing steps, the recognized strings can be reasonably classified through step S402, thereby simplifying the string splicing process under different perspectives. For example, under the same perspective, OCR may recognize multiple strings, such as "aBC123", "dEF456", "gHI789" and "ABC123", "DEF456". If all strings are compared with the y coordinate of the empty string (that is, 0), the system needs to process a large amount of redundant data, which increases the complexity of subsequent splicing work. In order to improve efficiency, the system chooses the first non-empty string (assuming it is "aBC123") as the benchmark for subsequent comparisons. Figure 4As shown in (a), the y-coordinates of the two complete strings are significantly different, while the y-coordinates of the substrings in the same complete string are usually close together (generally less than 250 pixels). Therefore, the system sets a threshold (250 pixels) to determine whether the strings belong to different parts of the same complete string. When the y-coordinate distance between the strings "aBC123" and "dEF456" is less than the threshold, the system assumes that they belong to the same complete string and classifies them into the same list. For those strings whose y-coordinate difference is greater than the threshold, such as "ABC123", the system will determine that they are part of another complete string and classify them into different lists. In this way, the system can effectively classify and aggregate strings, simplify subsequent multi-perspective splicing operations, and improve overall processing efficiency.

[0032] Furthermore, step S403 specifically includes: determining whether the x-left coordinate of each string in the list is 0; if so, the first character in the string is considered inaccurately recognized and deleted; determining whether the x-right coordinate of each string in the list is equal to the width of the image containing the string; if so, the last character in the string is considered inaccurately recognized and deleted; at this point, the list is updated to the string field with improved accuracy for each image. The aforementioned coordinate processing method can distinguish whether each string field is from the same complete string and improve the accuracy of string concatenation.

[0033] Furthermore, in step S404, the strings in the same list are concatenated to remove duplicates at the beginning and end. The method involves using the edit distance to calculate the similarity between the two strings. A function is then designed that uses a for loop to iteratively compare the local similarities of the two strings, starting from the beginning and ending respectively. If the local similarity exceeds a set threshold, the first string is concatenated with the second string, minus the local portion, to achieve the desired deduplication effect. In this embodiment, the similarity threshold is set to 0.7, and the string following the last null character in the same list that is not null at the beginning or end is used as the starting position for concatenation. This solution treats the empty string as a separator between complete strings, used to identify the starting and ending positions of the string. For example, in the list ["dEF456", "gHI789", " ", "aBC123"], the presence of the empty string " " indicates that "aBC123" is the beginning of a complete string. Therefore, the system uses the empty string as a marker and "aBC123" as the starting position for the concatenation process. The use of an empty string as a separator can help the system accurately identify and extract each substring, thereby achieving effective concatenation and reorganization of the complete string.

[0034] Step S5: Match the concatenated result with the character string format in the candidate set, and select the character string whose matching degree reaches the set threshold as the final recognition result; otherwise, the concatenated result is used as the final result.

[0035] It's important to note that failing to meet the threshold doesn't necessarily mean the recognition result doesn't meet the requirements. It could be that the string's format doesn't match the predefined format and specifications for this solution, meaning it represents a new or undefined string type. Therefore, even if a string doesn't meet the threshold, the system will still consider it a valid recognition result and display it as the final output in the user interface. This approach ensures the system's flexibility, allowing it to adapt to diverse input data and provide complete recognition results.

[0036] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention shall fall within the scope of protection of the present invention.

[0037] Example 2 Based on the same inventive concept, an embodiment of the present invention provides a terminal 3D circular character recognition device based on OCR technology, including: an image acquisition module, an image storage module, an image cropping module, an image processing module, a splicing and deduplication module and a result output module.

[0038] Among them, the image acquisition module is used to capture 360° surround terminal images.

[0039] The image storage module is used to store the collected terminal images in the order of character strings and establish a character string candidate set document.

[0040] The image cropping module is used to crop the terminal area in the image to obtain the corresponding terminal image set.

[0041] The image processing module is used to enhance the contrast of the image and perform OCR recognition on the terminal images arranged in sequence one by one.

[0042] The concatenation and deduplication module is used to store multiple recognition results in multiple lists, concatenate the strings in each list and remove duplicates, and finally obtain one or more complete strings.

[0043] Specifically, the splicing and deduplication module includes a first identification unit, a first data processing unit, a second data processing unit and a first splicing unit.

[0044] The first recognition unit is used to obtain the y coordinates of the positions of multiple character strings recognized from multiple pictures.

[0045] The first data processing unit is used to compare the y coordinates of multiple strings in multiple images, specifically comparing the y coordinate of each string with the y coordinate of the first non-empty string; adding strings greater than a set threshold to the same list, and storing the others in another list. The strings in the same list are represented as multiple segments of the same complete string.

[0046] The second data processing unit is used to judge the x-left and x-right coordinates of each string text box for multiple strings in the same list, and judge whether to delete the first and last characters according to the x-left and x-right coordinate values; specifically, the second data processing unit makes judgments according to the following method: judging whether the x-left coordinate of each string in the same list is 0, if it is 0, it is considered that the first character in the string is not accurately recognized and is deleted; judging whether the x-right coordinate of each string in the same list is equal to the width of the picture where the string is located, if it is equal, it is considered that the last character in the string is not accurately recognized and is deleted; at this point, the list is updated to the string field with improved accuracy for each picture.

[0047] The first splicing unit is used to remove duplicates from the beginning and end of the strings in the same list after the update, and obtain the spliced result. The first splicing unit performs the head-to-tail deduplication splicing according to the following method: the edit distance is used to calculate the similarity between the two strings, and then a function is designed to use a for loop to iterate the two strings from the beginning and the end respectively to compare their local similarities. If the local similarity is greater than the set threshold, the first string is spliced with the second string with the local part removed to achieve the deduplication effect. The similarity threshold in this embodiment is set to 0.7; and the string after the last empty character in the same list that is not empty at the beginning and the end is used as the starting position for splicing.

[0048] The result output module is used to match the splicing result with the string format in the candidate set document, and select the string whose matching degree reaches the set threshold as the final recognition result, otherwise the splicing result is used as the final result.

[0049] The functions of the above functional modules correspond one-to-one to the steps in the terminal 3D circular character recognition method in Example 1, and the parts not fully described will be omitted for brevity.

[0050] Example 3 Based on the same inventive concept, an embodiment of the present invention further provides a terminal recognition device based on OCR technology, comprising: memory for storing computer programs; The processor is configured to implement the steps of the terminal 3D circular character recognition method based on OCR technology as described in the above embodiment 1 when executing the computer program.

[0051] Preferably, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units can be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in the device.

[0052] The processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor can also be any conventional processor. The processor is the control center of the device, and various parts of the device are connected using various interfaces and lines.

[0053] The memory mainly includes a program storage area and a data storage area. The program storage area can store the operating system, at least one application required for a function, etc., and the data storage area can store related data, etc. In addition, the memory can be a high-speed random access memory, or a non-volatile memory such as a plug-in hard disk, a smart memory card, a secure digital card, and a flash memory card, or the memory can also be other volatile solid-state memory devices.

[0054] Example 4 Based on the same inventive concept, an embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the terminal 3D circular character recognition method based on OCR technology as described in the above embodiment 1 are implemented.

[0055] Computer storage media may be tangible media that can contain or store programs for use by or in connection with an instruction execution system, apparatus, or device.

Claims

1. A terminal 3D circular character recognition method based on OCR technology, characterized in that: The steps include: Step S1, collecting terminal images in 360° surround, storing the images in character string order, and creating a character string candidate set document; Step S2: cropping the terminal area in the image to obtain a corresponding terminal image set; Step S3: enhancing the image contrast and performing OCR recognition on the terminal images arranged in sequence one by one; Step S4: storing the multiple recognition results in multiple lists, concatenating the character strings in each list and removing duplicates, to finally obtain one or more complete character strings; Step S5: Match the concatenated result with the preset character string format in the candidate set document, and select the character string whose matching degree reaches the set threshold as the final recognition result; otherwise, the concatenated result is used as the final result.

2. The terminal 3D circular character recognition method based on OCR technology according to claim 1, characterized in that: Store multiple recognition results in multiple lists, concatenate the strings in each list and remove duplicates, and finally obtain one or more complete strings, including: Step S401: Obtain the y coordinates of multiple character string positions identified in multiple images; Step S402: Compare the y coordinates of multiple strings in multiple images, add the strings that are greater than a set threshold to the same list, and store the others in another list; Step S403: for multiple character strings in the same list, determine the x-coordinates of each character string text box, and determine whether to delete the first and last characters based on the x-coordinates. Step S404: update the list accordingly according to the result of step S403, and remove duplicates at the beginning and end of the character strings in the same list to obtain a spliced result.

3. The terminal 3D circular character recognition method based on OCR technology as claimed in claim 1, characterized in that: When performing OCR recognition on images arranged in sequence one by one, the images are divided into two categories: one is the image that has been cropped and contrast enhanced, and the other is the image that has not been cropped and the contrast is directly enhanced; OCR recognition is performed on the two categories of images separately, and finally the group of results with the higher recognition rate in the two categories of images is returned as the recognition result.

4. The terminal 3D circular character recognition method based on OCR technology as claimed in claim 2, characterized in that: The method for removing duplicates from the beginning and end of strings in the same list is to use the edit distance to calculate the similarity between the two strings, and then design a function using a for loop to iterate the two strings from the beginning and end respectively to compare their local similarities. If their local similarity is greater than the set threshold, the first string is spliced with the second string with the local part removed.

5. A terminal 3D circular character recognition method based on OCR technology as described in claim 2 or 4, characterized in that: In step S404, the string after the last null character in the same list is used as the starting position for splicing.

6. The terminal 3D circular character recognition method based on OCR technology as claimed in claim 2, characterized in that: In step S403, the strings in the same list are represented as multiple segments of the same complete string. It is determined whether the x-left coordinate of each string in the list is 0. If it is 0, it is considered that the first character in the string is not accurately recognized and is deleted. It is also determined whether the x-right coordinate of each string in the list is equal to the width of the image containing the string. If it is, it is considered that the last character in the string is not accurately recognized and is deleted.

7. A terminal 3D circular character recognition device based on OCR technology, characterized in that: include: Image acquisition module, used for 360° surround acquisition of terminal images; The image storage module is used to store the collected terminal images in the order of character strings and establish a character string candidate set document; The image cropping module crops the terminal area in the image to obtain the corresponding terminal image set; Image processing module, used to enhance the contrast of the image and perform OCR recognition on the terminal images arranged in sequence one by one; The concatenation and deduplication module is used to store multiple recognition results in multiple lists, concatenate the strings in each list and remove duplicates, and finally obtain one or more complete strings; The result output module is used to match the splicing result with the string format in the candidate set document, and select the string whose matching degree reaches the set threshold as the final recognition result, otherwise the splicing result is used as the final result.

8. The terminal 3D circular character recognition device based on OCR technology as claimed in claim 7, characterized in that: The splicing and deduplication module includes: A first recognition unit is used to obtain the y coordinates of the positions of multiple character strings recognized from multiple images; a first data processing unit, configured to compare the y-coordinates of multiple strings in multiple images, add strings whose values are greater than a set threshold to the same list, and store the others in another list; The second data processing unit is used to determine the x-coordinates of each string text box for multiple strings in the same list, and determine whether to delete the first and last characters according to the x-coordinates. The first splicing unit is used to update the list according to the processing result of the second data processing unit, and to remove the duplication at the beginning and end of the character strings in the same list and splice them to obtain a spliced result.

9. A terminal 3D circular character recognition device based on OCR technology, characterized in that: include: memory for storing computer programs; A processor is configured to implement the steps of a terminal 3D circular character recognition method based on OCR technology as described in any one of claims 1 to 6 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the terminal 3D circular character recognition method based on OCR technology as described in any one of claims 1 to 6.

Images