Person information recognition correction method and device, storage medium and electronic equipment

Abstract

The application relates to a character information correction method and device, a storage medium and electronic equipment. The method comprises the following steps: obtaining text content and spatial coordinate information of each text region in a target micro-drama picture, clustering all the text regions based on spatial distance relationships, and obtaining a first character identification result; inputting the target micro-drama picture into a visual large model to perform visual reasoning on the target micro-drama picture, and obtaining a second character identification result; comparing the first character identification result with the second character identification result, and performing verification processing on the second character identification result according to a comparison result, and obtaining a verified character identification result; in the case that the number of second character names is greater than the number of first character names, performing merging and deduplication processing on the verified character identification result according to the first character identification result, and obtaining a final character identification result. The application solves the technical problem that a large model is prone to misdivide single character information, thereby causing character identification errors.

Description

Technical Field

[0001] This application relates to the field of multimedia intelligent processing technology, and in particular to a method, device, storage medium, and electronic device for correcting personal information. Background Technology

[0002] With the continuous development of artificial intelligence and large-scale visual models in video content understanding, automatic character information recognition in micro-drama scenes has gradually become an important means to improve video analysis efficiency. However, existing methods still have serious problems when recognizing character names in micro-drama images, namely, large models are prone to misinterpreting information about a single character as belonging to multiple people. For example, when the character "Zhang San" appears in the scene, and his introduction is "father of Zhang San / Li Si," if other characters are present in the scene, the model may misidentify the other characters as "Li Si," resulting in a hallucination error in character recognition. This type of problem not only affects the accuracy of recognition but may also interfere with subsequent plot analysis, subtitle generation, or character association processing. The root cause of this problem is that large models, during visual reasoning, are prone to generating non-existent names based on local features or semantic associations, i.e., creating hallucinations, and it is difficult to completely eliminate such errors by relying solely on model training. In addition, the character appearance information in micro-drama images is usually relatively compact and diverse, and the distribution of character name text has a complex correspondence with the position in the scene, making it difficult for traditional recognition methods to accurately distinguish the name information of different characters. Summary of the Invention

[0003] This application provides a method, apparatus, storage medium, and electronic device for correcting personal information recognition, in order to solve the technical problem that large models are prone to missegmenting individual personal information, resulting in personal recognition errors.

[0004] In a first aspect, this application provides a method for correcting the identification of person information, comprising: performing optical character recognition processing on a target micro-drama image to obtain the text content and spatial coordinate information of each text region in the target micro-drama image, and clustering all text regions based on spatial distance relationships to generate multiple first person names, thereby obtaining a first person recognition result; inputting the target micro-drama image into a visual big model to enable the visual big model to perform visual reasoning on the target micro-drama image, thereby obtaining a second person recognition result, wherein the second person recognition result includes multiple second person names and their corresponding face regions; comparing the first person recognition result with the second person recognition result, and performing verification processing on the second person recognition result based on the comparison result, thereby obtaining a verified person recognition result, wherein the verification processing includes deletion and / or completion processing; comparing the number of second person names in the verified person recognition result with the number of first person names in the first person recognition result, and if the number of second person names is greater than the number of first person names, performing merging and deduplication processing on the verified person recognition result based on the first person recognition result, thereby obtaining a final person recognition result.

[0005] Secondly, this application provides a correction device for identifying person information, comprising: an identification module, used to perform optical character recognition processing on a target micro-drama image to obtain the text content and spatial coordinate information of each text region in the target micro-drama image, and to cluster all text regions based on spatial distance relationships to generate multiple first person names, thereby obtaining a first person identification result; and an inference module, used to input the target micro-drama image into a visual big data model, so that the visual big data model performs visual inference on the target micro-drama image to obtain a second person identification result, wherein the second person identification result includes multiple second person names and their corresponding persons. The system includes a face region; a comparison module for comparing the first person recognition result with the second person recognition result, and performing verification processing on the second person recognition result based on the comparison result to obtain a verified person recognition result, wherein the verification processing includes deletion and / or completion processing; and a merging module for comparing the number of second person names in the verified person recognition result with the number of first person names in the first person recognition result, and performing merging and deduplication processing on the verified person recognition result based on the first person recognition result when the number of second person names is greater than the number of first person names, to obtain the final person recognition result.

[0006] As an optional example, the above recognition module includes: a detection unit, used to perform text detection processing on the target micro-drama image to locate each text region in the target micro-drama image; an extraction unit, used to perform optical character recognition processing on each located text region to extract the corresponding text content; a first acquisition unit, used to acquire the spatial coordinate information of each text region in the target micro-drama image, wherein the spatial coordinate information includes the bounding box coordinates and / or center point coordinates of the text region; and a first storage unit, used to associate and store the recognized text content with the corresponding spatial coordinate information to generate a text recognition result.

[0007] As an optional example, the above recognition module includes: a calculation unit, used to calculate the spatial distance between any two text regions based on the spatial coordinate information of each text region; a first division unit, used to divide text regions with a spatial distance less than a preset distance threshold into the same text cluster to generate multiple text clusters; and an aggregation unit, used to aggregate the text content of all text regions in each text cluster to obtain the corresponding first person name and spatial location, and output the first person recognition result composed of all first person names.

[0008] As an optional example, the above-mentioned reasoning module includes: a reasoning unit, used to input the above-mentioned target micro-drama image into the above-mentioned visual big model, so that the above-mentioned visual big model can identify the facial region of each character in the above-mentioned target micro-drama image, and assign a corresponding facial region number to each facial region, and perform character name reasoning on the identified facial regions to generate a second character name corresponding to each character; and a second storage unit, used to associate and store the facial region of each character, the corresponding second character name and the assigned facial region number to form the above-mentioned second character recognition result.

[0009] As an optional example, the comparison module includes: a second acquisition unit, used to acquire the second person name corresponding to each person in the second person recognition result; a judgment unit, used to match and judge the target person name with all the first person names in the first person recognition result, wherein the target person name is any second person name in the second person recognition result; and a deletion unit, used to delete the target person name and the corresponding face region from the second person recognition result if the target person name does not appear in any of the first person names in the first person recognition result.

[0010] As an optional example, the comparison module further includes a completion unit, used to perform extended completion processing on the target person name based on the corresponding first person name when there is a partial match between the target person name and any first person name in the first person recognition result.

[0011] As an optional example, the above merging module includes: a mapping unit, used to map each second person name in the above-mentioned verification person recognition result to the first person name corresponding to the above-mentioned first person recognition result, to obtain a mapping result; a second division unit, used to divide multiple second person names mapped to the same first person name into the same person group according to the above-mentioned mapping result; a determination unit, used to, for each person group, select the second person name that is a role name, or the one with the smallest corresponding face region number, or the first second person name as the representative person name of the corresponding person group; and a selection unit, used to count the representative person names of all person groups, and if the total number of representative person names is greater than the number of above-mentioned first person names, sort all representative person names in ascending order according to the face region number, and select the top N representative person names as the above-mentioned final person recognition result, where the above-mentioned N is the number of above-mentioned first person names.

[0012] Thirdly, this application provides a storage medium storing a computer program, wherein the computer program is executed by a processor to perform the above-described method for correcting the identification of personal information.

[0013] Fourthly, this application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the above-described method for correcting the identification of personal information through the computer program.

[0014] The technical solutions provided in this application have the following advantages compared with the prior art: This application employs optical character recognition (OCR) processing on the target micro-drama image to obtain the text content and spatial coordinate information of each text region in the target micro-drama image. Based on spatial distance relationships, all text regions are clustered to generate multiple first character names, resulting in a first character recognition result. The target micro-drama image is then input into a large-scale visual model to perform visual reasoning, yielding a second character recognition result. This second character recognition result includes multiple second character names and their corresponding face regions. The first character recognition result is compared with the second character recognition result, and based on the comparison result, the second character recognition result is verified to obtain a verified character recognition result. This verification process includes deletion and / or completion. Finally, the number of second character names in the verified character recognition result is compared with the number of first character names in the first character recognition result. The method involves merging and deduplicating the verification character recognition results based on the first character recognition result when the number of second character names exceeds the number of first character names, thus obtaining the final character recognition result. This method obtains text and coordinate information by performing optical character recognition on the micro-drama image, clusters the text region based on spatial distance to obtain the first character recognition result, combines this with visual large model reasoning to obtain the second character recognition result, compares the second character recognition result with the first character recognition result, deletes and / or completes the second character recognition result, and performs clustering mapping and deduplication processing when the number of second character names exceeds the number of first character names, ultimately obtaining an accurate character recognition result. This achieves the goal of improving the accuracy and reliability of character recognition in micro-drama images, thereby solving the technical problem of large models easily missplitting single-character information, causing character recognition errors. Attached Figure Description

[0015] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0018] Figure 1 This is a flowchart of an optional method for correcting personal information identification according to an embodiment of this application; Figure 2 This is a flowchart illustrating the implementation of an optional method for correcting personal information identification according to an embodiment of this application. Figure 3 This is a schematic diagram of the structure of an optional correction device for recognizing personal information according to an embodiment of this application; Figure 4 This is a schematic diagram of an optional electronic device according to an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] The following disclosure provides numerous different embodiments or examples for implementing various structures of this application. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the scope of this application. Furthermore, reference numerals and / or letters may be repeated in different examples. Such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0021] According to a first aspect of the embodiments of this application, a method for correcting identification of person information is provided, optionally, as follows: Figure 1 As shown, the above method includes: S102, perform optical character recognition processing on the target micro-drama image to obtain the text content and spatial coordinate information of each text region in the target micro-drama image, and cluster all text regions based on spatial distance relationship to generate multiple first character names and obtain the first character recognition result; S104, Input the target micro-drama image into the visual big model so that the visual big model can perform visual reasoning on the target micro-drama image to obtain the second person recognition result, wherein the second person recognition result includes multiple second person names and their corresponding face regions; S106, compare the first person recognition result with the second person recognition result, and perform verification processing on the second person recognition result based on the comparison result to obtain the verified person recognition result, wherein the verification processing includes deletion and / or completion processing. S108, compare the number of second person names in the verification person recognition result with the number of first person names in the first person recognition result, and if the number of second person names is greater than the number of first person names, merge and deduplicate the verification person recognition result according to the first person recognition result to obtain the final person recognition result.

[0022] Optionally, in this embodiment, a method for correcting character information is provided for the automatic identification and error correction of character names in micro-drama images. Addressing the problem of missegmentation in existing large-scale models when identifying character names in micro-drama images—such as incorrectly associating single character information with other characters, thus producing hallucinatory recognition errors—a correction mechanism based on OCR (Optical Character Recognition) text clustering combined with visual large-scale model inference is proposed to achieve accurate identification and correction of character information.

[0023] like Figure 2 The specific implementation process shown first involves performing OCR recognition on the target micro-drama image to obtain the text content and spatial coordinates of each text region. By obtaining the spatial coordinates of the text regions, the relative positional relationship of the text in the image can be analyzed, and a text spatial model can be established accordingly. Based on this, all text regions are clustered according to their spatial distance relationships, grouping text regions with similar spatial positions into the same cluster, generating multiple text clusters. The text content of all text regions in each text cluster is then aggregated to obtain the corresponding first person's name and spatial location, thus forming the first person recognition result. The first person recognition result reflects the set of text regions corresponding to each potential person in the image, providing constraint information for subsequent person recognition.

[0024] The target micro-drama images are input into a large-scale visual model for visual inference. The model performs face detection on the figures in the images, identifying the facial regions of each person, and inferring the person's name from each facial region to generate a second person identification result. This second person identification result includes the second person's name and its corresponding facial region for each person. A sequence number can also be assigned to each identified facial region for subsequent sorting and merging processing. However, because existing large-scale models may produce hallucination recognition in complex scenes, such as mistakenly splitting information from a single person into multiple persons, relying solely on the inference results of the large-scale model cannot guarantee accurate identification.

[0025] To address this, a verification mechanism based on OCR text clustering is introduced to compare and process the second person identification results. Specifically, each second person name in the second person identification result is matched with the first person name in the first person identification result. If the second person name does not appear in the first person identification result, the corresponding person information is deleted to eliminate the illusion of a corrected name. If there is a partial match between the second person name and the first person identification result, the second person name is completed by replacing it with the full name of the first person in the first person identification result, thereby correcting incomplete or truncated name information. After this verification process, the verified person identification result is obtained.

[0026] Subsequently, the number of second-person names in the verified person recognition results is compared with the number of first-person names in the verified person recognition results. When the number of second-person names is greater than the number of first-person names, it indicates a mis-splitting phenomenon, requiring merging and deduplication. Specifically, each person in the verified person recognition results is mapped to a corresponding first-person name, and multiple second-person names mapped to the same first-person name are grouped into the same person group based on the mapping relationship. For each group, a representative person name is selected according to the following rules: Prioritize character names that are role names; if all character names are role names or none are role names, select the character name with the smallest avatar frame number; if the number cannot be determined, select the first character name in the group. If the total number of merged character names still exceeds the number of first-person names, sort them according to their avatar frame numbers from smallest to largest, and select the N character names with the smallest numbers as the final output character names, where N is the number of first-person names. Finally, the representative character names for each group are output, forming the final person recognition result.

[0027] Optionally, this embodiment can effectively correct the illusionary character recognition errors generated by the large model in micro-drama images, solving problems such as mis-segmentation of single-character information, incorrect association of multiple characters, and incomplete names. By comprehensively utilizing OCR text clustering constraints and visual large model inference results, it can achieve character name deletion, completion, and deduplication, thereby obtaining stable, accurate character recognition results consistent with the actual screen content. This improves the reliability and usability of automatic character recognition in micro-drama images, while providing accurate data support for subsequent plot analysis, subtitle generation, and character relationship processing.

[0028] As an optional example, optical character recognition (OCR) processing is performed on the target micro-drama image to obtain the text content and spatial coordinate information of each text region in the target micro-drama image, including: Text detection processing is performed on the target micro-drama images to locate various text regions within them; Optical character recognition (OCR) is performed on each text region located to extract the corresponding text content; Obtain the spatial coordinate information of each text region in the target micro-drama image, wherein the spatial coordinate information includes the bounding box coordinates and / or center point coordinates of the text region; The recognized text content is associated with and stored with the corresponding spatial coordinate information to generate text recognition results.

[0029] Optionally, in this embodiment, optical character recognition (OCR) processing is performed on the target micro-drama image to obtain the text content and spatial coordinate information of each text region in the target micro-drama image. Specifically, the process includes the following steps: First, text detection processing is performed on the target micro-drama image. Using a text detection model or character region localization algorithm, various text regions in the image that may contain character appearance information, character names, or subtitle descriptions are identified and located, resulting in multiple text regions. Second, each located text region is further processed using OCR. The text content is extracted using a character segmentation and recognition network to generate corresponding text string information. Simultaneously, to support subsequent spatial relationship-based clustering analysis, the spatial coordinate information of each text region in the target micro-drama image also needs to be obtained. The spatial coordinate information may include the bounding box coordinates of the text region (such as the coordinates of the upper left and lower right corners) or the center point coordinates of the text region, used to describe the specific location distribution of the text in the image. Finally, the identified text content and corresponding spatial coordinate information are associated and stored to form a structured text recognition result set.

[0030] As an optional example, clustering all text regions based on spatial distance relationships generates multiple first-person names, yielding the first-person identification results: Based on the spatial coordinate information of each text region, calculate the spatial distance between any two text regions; Text regions with a spatial distance less than a preset distance threshold are divided into the same text cluster to generate multiple text clusters; Aggregate the text content of all text regions in each text cluster to obtain the corresponding first person name and spatial location, and output the first person identification result composed of all first person names.

[0031] Optionally, in this embodiment, all text regions are clustered based on spatial distance relationships to obtain the first person identification result. Specifically, firstly, using the spatial coordinate information of each text region, the spatial distance between any two text regions is calculated. The spatial distance can be based on the Euclidean distance between the center points of the text regions or on the minimum interval distance between the bounding boxes, thereby quantifying the relative positional relationship of different text regions in the image. Secondly, text regions with a spatial distance less than a preset distance threshold are divided into the same text cluster to generate multiple text clustering results. This preset distance threshold can be set according to factors such as the text layout density and character appearance information layout characteristics in the micro-drama image, used to distinguish the spatial distribution differences between text related to the same character and text of different characters. Through this clustering method, text regions with similar spatial positions and closely related content can be merged into the same cluster to form several candidate character information blocks. Subsequently, the text regions in each text cluster are aggregated to obtain the corresponding first person name and spatial position. For example, the text content within the cluster is spliced ​​or merged, and the cluster spatial position is uniformly represented to obtain the corresponding first person name and cluster spatial position. Finally, output the first person identification result consisting of all the first person names.

[0032] As an optional example, the target micro-drama image is input into a large visual model, enabling the large visual model to perform visual reasoning on the target micro-drama image, and the resulting second person recognition result includes: The target micro-drama image is input into the visual big model so that the visual big model can identify the face region of each character in the target micro-drama image and assign a corresponding face region number to each face region. The identified face regions are used to infer the character name to generate a second character name corresponding to each character. The facial region of each person, the corresponding name of the second person, and the assigned facial region number are associated and stored to form the second person recognition result.

[0033] Optionally, in this embodiment, the target micro-drama image is input into a visual big data model, enabling the model to perform visual reasoning to obtain a second character recognition result. Specifically, firstly, the target micro-drama image is input into the visual big data model, which can be a visual language model with multimodal understanding capabilities, capable of detecting and semantically reasoning about characters in the image. The model analyzes the target micro-drama image, identifying the facial region corresponding to each character, for example, by locating the character's avatar frame or facial region using a face detection network, and assigning a corresponding number to each identified facial region. The number can be assigned based on the positional order of the facial regions in the image, detection confidence, or character importance, and is used for subsequent character merging, deduplication, and sorting selection. Subsequently, the model further performs character name reasoning on the identified facial regions, combining the appearance text information in the image, character relationship descriptions, and contextual semantics to generate a second character name for each character, thus obtaining preliminary recognition information containing the second character name and the facial region. Finally, the facial region of each person, the corresponding name of the second person, and the assigned facial region number are associated and stored to form the second person recognition result.

[0034] As an optional example, the first person recognition result is compared with the second person recognition result, and based on the comparison result, the second person recognition result is verified to obtain the verified person recognition result, which includes: Obtain the name of the second person corresponding to each person in the second person identification results; The target person's name is matched against all the first person's names in the first person's identification results, where the target person's name is any second person's name in the second person's identification results. If the target person's name does not appear in any of the first person's names in the first person's recognition results, the target person's name and the corresponding face region are deleted from the second person's recognition results.

[0035] Optionally, in this embodiment, the second person recognition result is compared with the first person recognition result, and the second person recognition result is verified based on the comparison result to obtain a verified person recognition result. Specifically, firstly, the second person name corresponding to each person is extracted from the second person recognition result. The second person name is generated by the visual big data model based on facial region reasoning, and there may be cases of illusion recognition or missegmentation. Subsequently, for each second person name in the second person recognition result, it is used as the target person name and matched against all the first person names in the first person recognition result. The first person names come from the text region aggregation result of OCR recognition and can reflect the text information of the real person appearing in the target micro-drama image, so they can be used as the basis for the validity of the second person name. During the matching process, methods such as string complete matching, keyword inclusion matching, or similarity matching can be used to determine whether the target person name appears in any of the first person names. If the target person name does not appear in any of the first person names in the first person recognition result, it indicates that the person name lacks OCR text evidence support and belongs to invalid person information or illusion results generated by the visual big data model reasoning. In this case, the target person's name and its corresponding facial region information are removed from the second person recognition result, thereby filtering out unreliable recognition outputs. Through the above verification process, the person information recognized by the large visual model can be effectively constrained and filtered based on the OCR text clustering results.

[0036] As an optional example, after matching the target person's name with all the first person names in the first person identification results, the above method further includes: If there is a partial match between the target person's name and any of the first person's names in the first person's identification results, the target person's name is expanded and completed based on the corresponding first person's name.

[0037] Optionally, in this embodiment, in actual micro-drama image scenarios, the second character name obtained by the visual big data model inference may only be a part of the OCR text. For example, the visual big data model recognizes "Zhang San," while the corresponding complete character name in the OCR clustered text content is "Zhang Sanfeng." In this case, if only the part of the name output by the visual big data model is retained, it may lead to missing character identity information, affecting subsequent character relationship analysis and role consistency maintenance. Therefore, when there is a partial match between the target character name and any of the first character names in the first character recognition results, the target character name is expanded and completed based on the complete character name in the corresponding first character name. Partial matching can be manifested as the target character name being a substring of the corresponding first character name, or the target character name and the corresponding first character name having an inclusion relationship in the character sequence. During the completion process, the target character name is replaced with the complete character name in the corresponding first character name, or missing characters are appended after the target character name to generate complete name information. Through this completion mechanism, the character name can be kept consistent with the OCR recognized text, avoiding the problem of incomplete names caused by truncation due to big data model inference.

[0038] As an optional example, based on the first person recognition result, the verification person recognition results are merged and deduplicated to obtain the final person recognition results, including: Map each second person name in the verified person recognition result to the first person name corresponding to the first person recognition result to obtain the mapping result; Based on the mapping results, multiple second-person names mapped to the same first-person name are grouped into the same person group; For each group of characters, the character name, the smallest corresponding face region number, or the first and second character name is taken as the representative character name of the corresponding character group. The names of representative figures in all groups are counted. If the total number of representative figures is greater than the number of first-person figures, all representative figures are sorted in ascending order according to the face region number. The top N representative figures are selected as the final person recognition result, where N is the number of first-person figures.

[0039] Optionally, in this embodiment, when the number of second person names is greater than the number of first person names, to avoid redundant person outputs generated by the large visual model affecting the final recognition accuracy, the verification person recognition results are merged and deduplicated based on the first person recognition results to obtain the final person recognition results. Specifically, firstly, each second person name in the verification person recognition results is mapped and matched with each first person name in the first person recognition results. The mapping basis may include the matching relationship between the second person name and the first person name, the spatial proximity relationship between the person's face region and the text region, etc., to obtain the mapping result from the second person name to the first person name. Secondly, based on the mapping result, multiple second person names mapped to the same first person name are divided into the same person group to indicate that these second person names may belong to the same role in duplicate recognition or missegmentation results. Then, for each person group, a representative person name is selected as the final output person name for that group. The selection strategy for the representative person name can be: prioritizing the selection of person names with clear role names; if there are no complete names in the group, selecting the person name with the smallest face region number; or defaulting to selecting the first person name in the group as the representative person name. This method ensures that each character group retains at most one representative character name, achieving cluster-level merging and deduplication. Furthermore, the representative character names corresponding to all character groups are counted, and it is determined whether the total number of representative character names is still greater than the number of first-person character names. If the number of characters after merging still exceeds the number of first-person character names, all representative character names are sorted in ascending order according to the face region index, and the top N representative character names with the smallest index are selected as the final character recognition result, where N is the number of first-person character names. This ensures that the final output number of character names does not exceed the number of first-person character names and guarantees that each character group corresponds to at most one character role, effectively eliminating duplicate or incorrect recognition caused by the illusion of a large model.

[0040] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0041] According to another aspect of the embodiments of this application, a correction device for recognizing person information is also provided, such as... Figure 3 As shown, it includes: The recognition module 302 is used to perform optical character recognition processing on the target micro-drama image to obtain the text content and spatial coordinate information of each text region in the target micro-drama image, and to cluster all text regions based on spatial distance relationship to generate multiple first character names and obtain the first character recognition result. The reasoning module 304 is used to input the target micro-drama image into the visual big model so that the visual big model can perform visual reasoning on the target micro-drama image to obtain the second character recognition result. The second character recognition result includes multiple second character names and their corresponding face regions. The comparison module 306 is used to compare the first person recognition result with the second person recognition result, and to perform verification processing on the second person recognition result based on the comparison result to obtain the verified person recognition result. The verification processing includes deletion and / or completion processing. The merging module 308 is used to compare the number of second person names in the verification person recognition result with the number of first person names in the first person recognition result. If the number of second person names is greater than the number of first person names, the verification person recognition result is merged and deduplicated according to the first person recognition result to obtain the final person recognition result.

[0042] It should be noted that the identification module 302 in this embodiment can be used to execute step S102 in this application embodiment, the reasoning module 304 in this embodiment can be used to execute step S104 in this application embodiment, the comparison module 306 in this embodiment can be used to execute step S106 in this application embodiment, and the merging module 308 in this embodiment can be used to execute step S108 in this application embodiment.

[0043] As an optional example, the recognition module includes: The detection unit is used to perform text detection processing on the target micro-drama image in order to locate each text region in the target micro-drama image; The extraction unit is used to perform optical character recognition processing on each text region located to extract the corresponding text content; The first acquisition unit is used to acquire the spatial coordinate information of each text region in the target micro-drama image, wherein the spatial coordinate information includes the bounding box coordinates and / or center point coordinates of the text region; The first storage unit is used to associate and store the recognized text content with the corresponding spatial coordinate information to generate text recognition results.

[0044] As an optional example, the recognition module includes: The calculation unit is used to calculate the spatial distance between any two text regions based on the spatial coordinate information of each text region; The first partitioning unit is used to divide text regions with a spatial distance less than a preset distance threshold into the same text cluster, so as to generate multiple text clusters; The aggregation unit is used to aggregate the text content of all text regions in each text cluster to obtain the corresponding first person name and spatial location, and output the first person identification result composed of all first person names.

[0045] As an optional example, the inference module includes: The reasoning unit is used to input the target micro-drama image into the visual big model, so that the visual big model can identify the face region of each character in the target micro-drama image, assign a corresponding face region number to each face region, and perform character name reasoning on the identified face region to generate a second character name corresponding to each character. The second storage unit is used to associate and store the facial region of each person, the corresponding name of the second person, and the assigned facial region number to form the second person recognition result.

[0046] As an optional example, the alignment module includes: The second acquisition unit is used to acquire the name of the second person corresponding to each person in the second person recognition result; The judgment unit is used to match the target person name with all the first person names in the first person recognition result, wherein the target person name is any second person name in the second person recognition result; The deletion unit is used to delete the target person's name and the corresponding face region from the second person recognition result if the target person's name does not appear in any of the first person's names in the first person recognition result.

[0047] As an optional example, the alignment module also includes: The completion unit is used to perform extended completion processing on the target person name based on the corresponding first person name after matching the target person name with all the first person names in the first person recognition result.

[0048] As an optional example, the merge module includes: The mapping unit is used to map each second person name in the verification person recognition result to the first person name corresponding to the first person recognition result, so as to obtain the mapping result. The second division unit is used to divide multiple second person names mapped to the same first person name into the same person group according to the mapping result; The unit is used to group each character, and the character name, the smallest corresponding face region number, or the first and second character name is used as the representative character name of the corresponding character group. The selection unit is used to count the representative names of all groups of people, and when the total number of representative names is greater than the number of first person names, sort all representative names in ascending order according to the face region number, and select the top N representative names as the final person recognition result, where N is the number of first person names.

[0049] For other examples of this embodiment, please refer to the examples above, which will not be repeated here.

[0050] Figure 4 This is a schematic diagram of an optional electronic device according to an embodiment of this application, such as... Figure 4 As shown, it includes a processor 402, a communication interface 404, a memory 406, and a communication bus 408. The processor 402, communication interface 404, and memory 406 communicate with each other via the communication bus 408. Memory 406 is used to store computer programs; When processor 402 executes a computer program stored in memory 406, it performs the following steps: Optical character recognition (OCR) is performed on the target micro-drama image to obtain the text content and spatial coordinate information of each text region in the target micro-drama image. Based on the spatial distance relationship, all text regions are clustered to generate multiple first character names and obtain the first character recognition results. The target micro-drama image is input into the visual big model so that the visual big model can perform visual reasoning on the target micro-drama image to obtain the second character recognition result. The second character recognition result includes multiple second character names and their corresponding face regions. The first person recognition result is compared with the second person recognition result, and the second person recognition result is verified based on the comparison result to obtain the verified person recognition result. The verification process includes deletion and / or completion. The number of second person names in the verification person recognition results is compared with the number of first person names in the first person recognition results. If the number of second person names is greater than the number of first person names, the verification person recognition results are merged and deduplicated based on the first person recognition results to obtain the final person recognition results.

[0051] Optionally, in this embodiment, the communication bus can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. This communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 The symbol is represented by a single thick line, but this does not indicate that there is only one bus or one type of bus. The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0052] The memory may include RAM, or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0053] As an example, the memory 406 described above may include, but is not limited to, the recognition module 302, reasoning module 304, comparison module 306, and merging module 308 of the aforementioned correction device for recognizing person information. Furthermore, it may include, but is not limited to, other module units of the aforementioned correction device for recognizing person information, which will not be elaborated upon in this example.

[0054] The processor mentioned above can be a general-purpose processor, including but not limited to: CPU (Central Processing Unit), NP (Network Processor), etc.; it can also be DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0055] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments, and will not be repeated here.

[0056] Those skilled in the art will understand that Figure 4The structure shown is for illustrative purposes only. The device that implements the above-mentioned method for correcting personal information can be a terminal device, such as a smartphone (e.g., Android phone, iOS phone), tablet computer, handheld computer, mobile internet device (MID), PAD, etc. Figure 4 This does not limit the structure of the aforementioned electronic devices. For example, the electronic device may also include components that are more... Figure 4 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 4 The different configurations shown.

[0057] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, ROM, RAM, disk or optical disk, etc.

[0058] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer program, which, when executed by a processor, performs the steps in the above-described method for correcting the identification of personal information.

[0059] Optionally, in this embodiment, those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0060] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0061] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause one or more computer devices (which may be personal computers, servers, or network devices, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0062] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0063] In the several embodiments provided in this application, it should be understood that the disclosed client can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection between units or modules, and may be electrical or other forms.

[0064] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0065] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0066] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for correcting personal information identification, characterized in that, include: Optical character recognition (OCR) is performed on the target micro-drama image to obtain the text content and spatial coordinate information of each text region in the target micro-drama image. Based on the spatial distance relationship, all text regions are clustered to generate multiple first character names and obtain the first character recognition result. The target micro-drama image is input into a large visual model so that the large visual model performs visual reasoning on the target micro-drama image to obtain a second character recognition result, wherein the second character recognition result includes multiple second character names and their corresponding face regions; The first person recognition result is compared with the second person recognition result, and the second person recognition result is verified based on the comparison result to obtain the verified person recognition result. The verification process includes deletion and / or completion processing. The number of second person names in the verified person recognition result is compared with the number of first person names in the first person recognition result. If the number of second person names is greater than the number of first person names, the verified person recognition result is merged and deduplicated according to the first person recognition result to obtain the final person recognition result.

2. The method according to claim 1, characterized in that, Optical character recognition (OCR) processing is performed on the target micro-drama image to obtain the text content and spatial coordinate information of each text region in the target micro-drama image, including: Text detection processing is performed on the target micro-drama image to locate each text region in the target micro-drama image; Optical character recognition (OCR) is performed on each text region located to extract the corresponding text content; Obtain the spatial coordinate information of each text region in the target micro-drama image, wherein the spatial coordinate information includes the bounding box coordinates and / or center point coordinates of the text region; The recognized text content is associated with and stored with the corresponding spatial coordinate information to generate text recognition results.

3. The method according to claim 1, characterized in that, Clustering of all text regions based on spatial distance relationships generates multiple first-person names, resulting in the following first-person identification results: Based on the spatial coordinate information of each text region, calculate the spatial distance between any two text regions; Text regions with a spatial distance less than a preset distance threshold are divided into the same text cluster to generate multiple text clusters; Aggregate the text content of all text regions in each text cluster to obtain the corresponding first person name and spatial location, and output the first person identification result composed of all first person names.

4. The method according to claim 1, characterized in that, The target micro-drama image is input into a large visual model, which then performs visual reasoning on the target micro-drama image to obtain the second person recognition result, including: The target micro-drama image is input into the visual big model so that the visual big model can identify the facial region of each character in the target micro-drama image and assign a corresponding facial region number to each facial region. The identified facial regions are used to infer the character name to generate a second character name corresponding to each character. The facial region of each person, the corresponding name of the second person, and the assigned facial region number are associated and stored to form the recognition result of the second person.

5. The method according to claim 1, characterized in that, The first person recognition result is compared with the second person recognition result, and the second person recognition result is verified based on the comparison result to obtain the verified person recognition result, which includes: Obtain the name of the second person corresponding to each person in the second person identification result; The target person name is matched with all the first person names in the first person identification result, wherein the target person name is any second person name in the second person identification result; If the target person's name does not appear in any of the first person's names in the first person recognition result, the target person's name and the corresponding face region are deleted from the second person recognition result.

6. The method according to claim 5, characterized in that, After matching the target person's name with all the first person's names in the first person identification result, the method further includes: If the target person's name partially matches any of the first person's names in the first person identification results, the target person's name is expanded and completed based on the corresponding first person's name.

7. The method according to any one of claims 1 to 6, characterized in that, Based on the first person recognition result, the verification person recognition results are merged and deduplicated to obtain the final person recognition result, including: Map each second person name in the verified person recognition result to the first person name corresponding to the first person recognition result to obtain the mapping result; Based on the mapping results, multiple second-person names mapped to the same first-person name are grouped into the same person group; For each group of characters, the character name, the smallest corresponding face region number, or the first and second character name is taken as the representative character name of the corresponding character group. The names of representative figures in all groups are counted. If the total number of representative figures is greater than the number of the first person names, all representative figures are sorted in ascending order according to the face region number. The top N representative figures are selected as the final person recognition result, where N is the number of the first person names.

8. A correction device for recognizing personal information, characterized in that, include: The recognition module is used to perform optical character recognition processing on the target micro-drama image to obtain the text content and spatial coordinate information of each text region in the target micro-drama image, and to cluster all text regions based on spatial distance relationship to generate multiple first character names and obtain the first character recognition result. The reasoning module is used to input the target micro-drama image into the visual big model, so that the visual big model performs visual reasoning on the target micro-drama image to obtain a second character recognition result, wherein the second character recognition result includes multiple second character names and their corresponding face regions; The comparison module is used to compare the first person recognition result with the second person recognition result, and to perform verification processing on the second person recognition result based on the comparison result to obtain a verified person recognition result. The verification processing includes deletion and / or completion processing. The merging module is used to compare the number of second person names in the verified person recognition result with the number of first person names in the first person recognition result, and if the number of second person names is greater than the number of first person names, the verified person recognition result is merged and deduplicated according to the first person recognition result to obtain the final person recognition result.

9. A computer-readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor to perform the method described in any one of claims 1 to 7.

10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the method described in any one of claims 1 to 7 through the computer program.

Images