Intelligent Quality Inspection and Error Correction Methods and Systems for Real Estate Registration Archives

Abstract

This invention relates to the field of real estate management technology and discloses an intelligent quality inspection and error correction method and system for real estate registration and confirmation archives. The method involves scanning real estate archive image data and performing image enhancement processing. The enhanced image data undergoes layout analysis and format standardization. The standardized archive data is then subjected to parallel quality inspection across five dimensions. A priority value YX is calculated based on the results, and error correction is performed according to the order of priority values ​​YX. The corrected results are then reviewed and stored. This invention achieves multi-dimensional synchronous detection of archive data by setting up five parallel quality inspection units, employing format standardization template verification, semantic analysis using a large language model in the real estate field, comparative learning using a knowledge graph graph attention network, a spatiotemporal constraint inference engine, and a cross-modal comparative learning framework. This improves the recall and precision of anomaly detection and solves the problem of missed detections in single-dimensional quality inspection.

Description

Technical Field

[0001] This invention relates to the field of real estate management technology, and more specifically to a method and system for intelligent quality inspection and error correction of real estate registration and ownership confirmation archives. Background Technology

[0002] Real estate registration archives are important historical records and legal documents formed during the implementation of the national unified real estate registration system. They have multiple core functions: they can clarify the ownership of property rights, clearly define the right holder, scope and content of real estate property rights through registers and original materials, avoid ownership disputes, and can be used as legal evidence. In litigation, arbitration and administrative proceedings, they can be directly used as legally valid evidence to protect the legitimate rights and interests of the parties involved. Existing real estate archive quality inspection methods are too simplistic and cannot detect anomalies in different dimensions simultaneously, resulting in insufficient comprehensiveness of quality inspection. Existing quality inspection systems lack a quantitative priority assessment mechanism after detecting multiple anomalies, making it impossible to determine a reasonable order of error correction and affecting the overall efficiency and accuracy of error correction. Existing automatic error correction results lack credible quantifiable evaluation and rely too heavily on manual review, leading to inefficiency and failing to achieve intelligent integration between automatic error correction and manual review. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, the implementation regulations of the present invention provide a method and system for intelligent quality inspection and error correction of real estate registration archives, so as to solve the technical problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method and system for intelligent quality inspection and error correction of real estate registration archives. The intelligent quality inspection and error correction method for real estate registration archives includes the following steps: Step S1: Scan the real estate archive image data and perform image enhancement processing; Step S2: Perform layout analysis and format standardization on the enhanced image data; Step S3: Perform parallel quality checks on the standardized archive data across five dimensions. Step S4: Calculate the priority value YX based on the previous results, and perform error correction according to the order of priority values ​​YX; Step S5: Review and store the corrected results.

[0005] The intelligent quality inspection and error correction system for real estate registration archives includes an image unit, a layout unit, a standardization unit, a quality inspection unit, an evaluation unit, an error correction unit, an audit unit, a manual unit, and a storage unit. The image unit enhances the archive image data; the layout unit analyzes the layout of the enhanced image data; the standardization unit standardizes the text format; the quality inspection unit performs parallel quality inspection of the archive data across five dimensions; the evaluation unit prioritizes and sorts the data based on priority values ​​Y and X; the error correction unit automatically repairs the data; the audit unit scores the error correction results; the manual unit performs manual error correction; and the storage unit stores the corrected archive data. The image unit is used to scan real estate archive image data, and the image unit performs image enhancement processing on the collected image data. The image unit calculates the pixel value of the scanned image using a weighted average method and converts it into a grayscale image. The calculated pixel value is linearly normalized, and after normalization, bilateral filtering is used for noise reduction. Adaptive histogram equalization is used for contrast enhancement, and then Gamma correction is used to complete the image enhancement processing.

[0006] In a preferred embodiment, the layout unit analysis performs connected component analysis on the enhanced binary image to distinguish between foreground and background. A YOLOv8-seg model trained on a real estate archive layout dataset is used to detect and classify the title, table, rights holder information area, signature and official seal area, and attached drawing area. The natural reading order is reconstructed based on the geometric position of the area and the prior knowledge of the archive template. The table area is further subdivided into cells through horizontal / vertical projection. The red seal interference area is filtered based on the HSV color space to complete the layout analysis.

[0007] In a preferred embodiment, the standardization unit uses OCR to extract editable text content from the image area after the layout unit is analyzed and performs text standardization processing, converting all text into UTF-8 encoding, performing standardization processing, and converting all text into archive data with a uniform format.

[0008] In a preferred embodiment, the quality inspection unit performs parallel quality inspections on the standardized archive data in five dimensions: format standardization detection, semantic rationality detection, ownership consistency detection, spatiotemporal logical coherence detection, and image-text consistency detection, generating five corresponding abnormal records: format abnormal records, semantic abnormal records, ownership abnormal records, spatiotemporal abnormal records, and image-text abnormal records.

[0009] In a preferred embodiment, the evaluation unit receives anomaly records from the quality inspection unit, calculates a priority value YX, sorts the data by priority, and determines the processing order. The calculation formula is as follows: In the formula, QS is the number of abnormal ownership information entries, QZ is the total number of ownership information entries, YX is the number of fields affected by the abnormal record, GL is the number of associated files affected by the abnormal record, k1 and k2 are weight coefficients. When the abnormal record involves abnormal ownership relationships and spatiotemporal abnormalities, the weight coefficient k2 is automatically increased by 0.1-0.2. The evaluation unit arranges the calculated priority values ​​YX in reverse order.

[0010] In a preferred embodiment, the error correction unit retrieves error correction schemes from the error correction strategy knowledge base and automatically corrects the data. For records with abnormal format, it completes and converts the format; for records with abnormal semantics, it adjusts logical contradictions and unifies consistency; for records with abnormal ownership, it updates the ownership status; for records with abnormal spatiotemporal relationships, it adjusts the time order and coordinate address; and for records with abnormal image and text, it corrects them based on the high-confidence modality.

[0011] In a preferred embodiment, the review unit scores the error correction results from the error correction unit, and the review unit calculates an error correction score value DF. The formula for calculating the error correction score value DF is as follows: In the formula, ZX is the inherent confidence of the policy, CF is the context sufficiency score, YZ is the consistency check score, and LS is the historical score rate.

[0012] In a preferred embodiment, the review unit compares the calculated error correction score DF with its internal error correction threshold Y. When the error correction score DF > the error correction threshold Y, the review unit sends a pass instruction to the storage unit. The storage unit receives the storage instruction and stores the file data with the error correction score DF > the error correction threshold Y. When the error correction score DF ≤ the error correction threshold Y, the review unit sends a manual instruction to the manual unit. The manual unit receives the file data with the error correction score DF ≤ the error correction threshold Y, performs manual review and error correction, and then sends it to the storage unit for storage.

[0013] In a preferred embodiment, the error correction unit is equipped with an error correction strategy knowledge base. The error correction strategy knowledge base is organized in the form of a knowledge graph, with error type as the head entity and specific error correction strategy as the tail entity. The edge attributes record the strategy priority, default confidence level, applicable scenario tags, and historical success rate weights. The error correction strategy knowledge base supports strategy registration, retrieval, dynamic weight adjustment, confidence factor update, and strategy execution log correlation analysis.

[0014] The technical effects and advantages of this invention are as follows: 1. This invention sets up five parallel quality inspection units with different dimensions, which respectively adopt format specification template verification, semantic analysis of large language model in the real estate field, knowledge graph graph attention network comparative learning, spatiotemporal constraint inference engine and cross-modal comparative learning framework to achieve multi-dimensional synchronous detection of archival data, improve the recall rate and precision of anomaly detection, and solve the problem of easy omission in single-dimensional quality inspection. 2. This invention calculates a priority value YX through an evaluation unit. The priority value YX comprehensively considers the proportion of ownership anomalies, the number of affected fields, and the number of related files, and determines the error correction order by arranging them in reverse order. The weight coefficient of anomalies involving ownership or time and space is dynamically increased to ensure that high-importance anomalies are processed first, thereby optimizing the error correction process and reducing the risk of chain errors caused by improper order. 3. This invention calculates the error correction score DF through an audit unit. The error correction score DF integrates the inherent confidence of the strategy, contextual sufficiency, consistency verification, and historical success rate, and compares it with a preset threshold Y. It automatically performs data entry and manual audit operations according to different situations, thereby realizing intelligent diversion of error correction results. While ensuring data accuracy, it minimizes manual intervention and improves processing efficiency. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0016] Figure 2 This is a schematic diagram of the overall system composition of the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The intelligent quality inspection and error correction method and system for real estate registration archives involved in the present invention are not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Reference Figure 1 This invention provides an intelligent quality inspection and error correction method and system for real estate registration archives, including the following steps: Step S1: Scan the real estate archive image data and perform image enhancement processing; Step S2: Perform layout analysis and format standardization on the enhanced image data; Step S3: Perform parallel quality checks on the standardized archive data across five dimensions. Step S4: Calculate the priority value YX based on the previous results, and perform error correction according to the order of priority values ​​YX; Step S5: Review and store the corrected results.

[0019] In this embodiment, during quality inspection and error correction, the application improves image quality and text readability through scanning enhancement, layout analysis, and standardization processing. It employs parallel quality inspection across five dimensions to comprehensively cover anomalies related to format, semantics, ownership, spatiotemporal consistency, and text-image consistency. Combined with priority value sorting for error correction, it ensures that highly important anomalies are processed first. The final review ensures the quality of error correction. This application automates, multi-dimensionalizes, and intelligently performs quality inspection and error correction of real estate archives, significantly improving the efficiency and accuracy of quality inspection.

[0020] Reference Figure 2 The intelligent quality inspection and error correction system for real estate registration archives includes an image unit, a layout unit, a standardization unit, a quality inspection unit, an evaluation unit, an error correction unit, an audit unit, a manual unit, and a storage unit. The image unit enhances the archive image data; the layout unit analyzes the layout of the enhanced image data; the standardization unit standardizes the text format; the quality inspection unit performs parallel quality inspection of the archive data across five dimensions; the evaluation unit prioritizes and sorts the data based on priority values ​​Y and X; the error correction unit automatically repairs the data; the audit unit scores the error correction results; the manual unit performs manual error correction; and the storage unit stores the corrected archive data. The image unit is used to scan real estate archive image data, and the image unit performs image enhancement processing on the acquired image data. The image unit calculates the pixel values ​​of the scanned image using a weighted average method and then converts it into a grayscale image. The weighted average calculation formula is G = 0.299×R + 0.587×G + 0.114×B, where G is the calculated pixel value, R is the pixel value of red pixels, G is the pixel value of green pixels, and B is the pixel value of blue pixels. The calculated pixel values ​​are linearly normalized, and then bilateral filtering is used for noise reduction. Adaptive histogram equalization is used for contrast enhancement, and then Gamma correction is used to complete the image enhancement processing.

[0021] In this embodiment, the application effectively improves the quality of archival images by weighted average grayscale conversion of image units, bilateral filtering for noise reduction, adaptive histogram equalization, and Gamma correction, thereby improving the accuracy of subsequent processing.

[0022] Reference Figure 2The layout unit analysis performs connected component analysis on the enhanced binary image to distinguish between foreground and background. It uses a YOLOv8-seg model trained on a real estate archive layout dataset to detect and classify the title, table, rights holder information area, official seal area, and attached image area. It reconstructs the natural reading order based on the geometric position of the area and the prior knowledge of the archive template. The table area is further divided into cells through horizontal / vertical projection. The red seal interference area is filtered based on the HSV color space to complete the layout analysis. When reconstructing the natural reading order, a basic sort is first performed based on the bounding box coordinates of each region. Rows are formed by grouping according to the upper boundary y-value, and within each row, they are arranged from left to right according to the left boundary x-value. For complex regions such as tables, row-major order is further adopted based on row and column indices. When pure coordinate sorting may cause misalignment, a priori real estate registration file template is introduced. The actual detected region type sequence is matched with the preset template using the longest common subsequence, and a forced rearrangement is performed based on the matched template. For multi-column or nested layouts, a depth-first traversal of the document tree is constructed or a projection histogram is used to divide the columns before sorting. The final output is a list containing region ID, category, coordinates, and reading sequence number. The table is organized in a two-dimensional order. If the template matching fails, it falls back to geometric sorting and marks low confidence for manual review.

[0023] In this embodiment, the application employs connected component analysis and the YOLOv8-seg model to accurately detect and classify the title, table, rights holder information area, official seal area, and attached image area. By combining prior reconstruction of the natural reading order with the archival template, the table area is split into cells, and red seal interference is filtered based on the HSV color space. This can solve the problems of false detection of areas, disordered reading order, and seal occlusion in complex layouts, providing high-quality structured input for subsequent OCR recognition and content analysis.

[0024] Reference Figure 2The standardization unit uses OCR to extract editable text content from the image area after page layout unit analysis and performs text standardization processing. All text is converted to UTF-8 encoding and standardized. All text is converted to a unified format. During standardization processing, invalid control characters are removed, traditional Chinese characters in the archives are converted to simplified Chinese using OpenCC, variant characters and colloquial characters are converted to standard characters, date expressions are standardized, numbers are standardized to half-width characters, addresses are standardized by administrative level, isolated punctuation and duplicate characters caused by image noise are deleted, erroneous word segmentation caused by recognition breaks is merged, the recognized text is mapped to specific data fields according to the area type obtained from page layout analysis, full-width punctuation is converted to half-width characters, and redundant spaces and blank lines are removed to complete the standardization processing. For text covered by seals, generative repair technology is used to first predict the text under the seal, then recognize the seal content and text content separately, and finally merge them. For old archives, deep learning super-resolution is used to improve text readability, and morphological reconstruction is used to repair damaged strokes.

[0025] In this embodiment, the text is extracted by OCR and then processed by UTF-8 encoding, simplified / traditional character conversion, variant character standardization, date and number unification, and address level standardization to eliminate isolated punctuation and duplicate characters caused by image noise. Generative repair technology is used to predict missing content for text covered by seals, and super-resolution is used to improve readability of old archives, thereby improving the degree of text standardization and recognition accuracy, and ensuring the consistency and integrity of archive data.

[0026] Reference Figure 2 The quality inspection unit performs parallel quality inspections on the standardized archive data in five dimensions: format standardization detection, semantic rationality detection, ownership consistency detection, spatiotemporal logical coherence detection, and image-text consistency detection. It generates five corresponding abnormal records: format abnormal records, semantic abnormal records, ownership abnormal records, spatiotemporal abnormal records, and image-text abnormal records. When working on the format standardization detection dimension, a preset template library of real estate registration archive format standards is loaded. The detection items include: table structure integrity, data format verification, and field value range verification. The table structure integrity detection checks whether there are any missing mandatory fields in the real estate registration application form and approval form. The data format verification checks whether the location address conforms to the format standard, whether the land parcel code conforms to the 19-digit coding rule, and whether the right holder's identity information conforms to the 18-digit ID card number format. The field value range verification checks whether the registration type field is within the preset type set. If a format abnormality is detected, a format abnormality record is generated. The semantic reasonableness detection dimension employs a large language model from the real estate registration domain for semantic understanding and reasonableness analysis. The construction method of this large language model is as follows: based on a general large language model, incremental pre-training is performed using specialized corpora from the real estate registration domain. The corpus includes: the Property Law section of the Civil Code, the Provisional Regulations on Real Estate Registration and its implementing rules, local real estate registration operation procedures, and approximately 500,000 manually reviewed and confirmed real estate registration files. During training, domain-specific loss functions are introduced, including semantic alignment loss for ownership relationships, reasonableness loss for the temporal sequence of registration events, and coherence loss for the property rights chain. The specific content of the semantic reasonableness detection includes: Logical contradiction detection: Detects whether there are logical contradictions within the same file; Semantic anomaly detection: Detects whether there are semantically incoherent or factually incorrect expressions in the text; Context consistency check: Checks whether the name of the same rights holder is consistent in different fields; When a semantic anomaly is detected, a semantic anomaly record is generated.

[0027] When performing consistency checks on ownership relationships, a real estate registration knowledge graph is constructed. This knowledge graph uses four types of entities as nodes: real estate unit nodes, containing attributes such as parcel code, location, land use, and building area; rights holder nodes, containing attributes such as rights holder name / title, document type, and document number; ownership relationship nodes, representing the ownership relationship between the rights holder and the real estate unit, containing attributes such as rights type, rights ratio, and rights term; and registration event nodes, representing specific registration events, containing attributes such as registration time, registration type, and registration agency. A graph attention network is used to encode the knowledge graph, mapping the ownership information in the files to be checked into graph structure data. Through comparative learning, the topological differences between the subgraph to be checked and the standard subgraph in the knowledge graph are calculated. If an inconsistency in ownership relationships is detected, an abnormal ownership relationship record is generated.

[0028] The spatiotemporal logical coherence detection dimension extracts time series information and spatial coordinate information from the archives to construct a set of spatiotemporal constraint rules, including: Registration time sequence constraints: the registration time for change of registration must be after the initial registration time; Consequences of ownership change time: The time interval between two consecutive ownership changes should be greater than a reasonable threshold. The spatial topological relationship of the land parcel boundaries must be constrained; the boundary lines of adjacent land parcels must coincide; and there should be no overlap or gaps between land parcels. Consistency constraint between location address and coordinates: The location address described in the text should match the address corresponding to the GIS coordinates. The spatiotemporal constraint inference engine verifies each of the above constraints, detects spatiotemporal coherence anomalies, and generates spatiotemporal anomaly records when spatiotemporal coherence anomalies occur.

[0029] When performing image-text consistency detection, a cross-modal contrastive learning framework is used to align and verify image layout information with text semantic information. A pre-trained visual encoder is used to extract visual feature vectors from the archive image, and a pre-trained text encoder is used to extract semantic feature vectors from the OCR-recognized text. A contrastive loss function is used to narrow the distance between positive sample pairs of images and text and widen the distance between negative sample pairs. For the archive to be detected, the similarity score between visual features and text features is calculated. If the similarity is lower than a preset threshold, it is judged as an image-text inconsistency anomaly, and an image-text anomaly record is generated.

[0030] In this embodiment, the application performs parallel detection of five dimensions: format standardization, semantic rationality, ownership consistency, spatiotemporal logical coherence, and text-image consistency, generating corresponding anomaly records respectively. This enables the application to capture multiple types of defects at once, avoiding omissions and delays in serial detection, significantly improving the comprehensiveness and efficiency of quality inspection, and providing complete anomaly information for subsequent priority assessment and error correction.

[0031] Reference Figure 2 The evaluation unit receives the abnormal records from the quality inspection unit, calculates the priority value YX, sorts them by priority, and determines the processing order. The calculation formula is as follows: In the formula, QS represents the number of ownership information entries with abnormalities, QZ represents the total number of ownership information entries, YX represents the number of fields affected by the abnormal record, GL represents the number of associated files affected by the abnormal record, k1 and k2 are weight coefficients, and the weight coefficients k1 and k2 are the proportion of file problems caused by the number of fields affected by the abnormal record and the number of associated files affected by the abnormal record. When the abnormal record involves abnormal ownership relationships and spatiotemporal abnormalities, the weight coefficient k2 is automatically increased by 0.1-0.2. The evaluation unit arranges the calculated priority values ​​YX in reverse order.

[0032] In this embodiment, the application calculates a priority value YX. In the priority value calculation, QS / QZ reflects the proportion of ownership anomalies, YC is the number of affected fields, GL is the number of affected related files, and k1 and k2 are weight coefficients. The processing order is determined by arranging them in reverse order. High-impact and high-risk anomaly records are automatically identified and corrected first through quantitative evaluation, optimizing resource allocation, avoiding the delay of critical errors, and improving the overall timeliness and effectiveness of error correction. The weight coefficient k2 is dynamically adjusted upward for anomalies involving ownership or time and space to ensure that high-importance anomalies are processed first.

[0033] Reference Figure 2 The error correction unit retrieves error correction schemes from the error correction strategy knowledge base and automatically corrects the data. For records with abnormal format, it completes and converts the format; for records with abnormal semantics, it adjusts logical contradictions and unifies consistency; for records with abnormal ownership, it updates the ownership status; for records with abnormal spatiotemporal relationships, it adjusts the time order and coordinate address; and for records with abnormal image and text, it corrects them based on the high-confidence modality.

[0034] In this embodiment, the application retrieves matching schemes from the error correction strategy knowledge base, completes and converts format anomalies, adjusts logical contradictions for semantic anomalies, updates ownership status for ownership relationship anomalies, adjusts time order and coordinate address for spatiotemporal anomalies, and corrects image and text anomalies based on high-confidence modalities. The classification error correction mechanism of this application achieves accurate repair of anomaly types, improves the success rate of automatic error correction, and reduces the need for manual intervention.

[0035] Reference Figure 2 The review unit scores the error correction results from the error correction unit, and the review unit calculates the error correction score value DF. The formula for calculating the error correction score value DF is as follows: In the formula, ZX is the inherent confidence of the policy, CF is the context sufficiency score, YZ is the consistency check score, and LS is the historical score rate. The inherent confidence score ZX is predefined in the error correction strategy knowledge base and is directly assigned a value based on the strategy type, with a value range of 0-1. The context sufficiency score CF is calculated by weighted summation of the reliability coefficient and relevance coefficient of the reference data source, divided by a normalization threshold, and limited to an upper limit of 1. The reference sources include historical registration databases, standard address databases, OCR-recognized text, etc. The consistency verification score YZ adopts a deduction system, subtracting the sum of the penalty weights of all violated preset check items from 1, and the sum of all penalty items is normalized to 1. The historical success rate LS is calculated based on the proportion of recent correction results of the same error type and the same error correction strategy that have been manually confirmed. The new combination uses the global average success rate as the initial value.

[0036] In this embodiment, the error correction score DF is calculated by formula, and the inherent confidence of the strategy, contextual sufficiency, consistency verification and historical success rate are combined to realize a multi-dimensional quantitative evaluation of the error correction result. This application provides a reliable credibility index for automatic error correction results, which facilitates subsequent intelligent diversion decision-making and ensures that only high-quality error correction results enter the archive.

[0037] Reference Figure 2The review unit compares the calculated error correction score DF with its internal error correction threshold Y. When the error correction score DF > the error correction threshold Y, the review unit sends a pass instruction to the storage unit. The storage unit receives the storage instruction and stores the file data with the error correction score DF > the error correction threshold Y. When the error correction score DF ≤ the error correction threshold Y, the review unit sends a manual instruction to the manual unit. The manual unit receives the file data with the error correction score DF ≤ the error correction threshold Y, performs manual review and error correction, and then sends it to the storage unit for storage.

[0038] In this embodiment, the review unit compares the error correction score DF with a preset threshold Y. When the error correction score DF > the error correction threshold Y, it is automatically stored. When the error correction score DF ≤ the error correction threshold Y, it is transferred to manual review, thus realizing dynamic intelligent diversion. This dynamic diversion mechanism reduces the workload of manual review while ensuring data accuracy, and achieves an intelligent balance between automatic error correction and manual review, thereby improving processing efficiency and reducing the risk of incorrect data entry.

[0039] Reference Figure 2 The error correction unit is equipped with an error correction strategy knowledge base, which is organized in the form of a knowledge graph. The error type is the head entity and the specific error correction strategy is the tail entity. The edge attributes record the strategy priority, default confidence, applicable scenario tags and historical success rate weights. The error correction strategy knowledge base supports strategy registration, retrieval, dynamic weight adjustment, confidence factor update and strategy execution log correlation analysis.

[0040] In this embodiment, the error correction strategy is organized in the form of a knowledge graph, with error type as the head entity and error correction strategy as the tail entity. Edge attributes record strategy priority, default confidence, applicable scenario tags, and historical success rate weights. It supports strategy registration, retrieval, dynamic weight adjustment, confidence factor update, and execution log correlation analysis. This knowledge base architecture makes the error correction strategy scalable, optimizable, and traceable. As the number of uses increases, the strategy weights can be adaptively adjusted to ensure that the application has high accuracy even after long-term use.

[0041] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. The units and algorithm steps of the various examples described in the embodiments can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0042] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, 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; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0043] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0044] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent quality inspection and error correction of real estate registration archives, characterized by: Includes the following steps: Step S1: Scan the real estate archive image data and perform image enhancement processing; Step S2: Perform layout analysis and format standardization on the enhanced image data; Step S3: Perform parallel quality checks on five dimensions on the standardized archive data; Step S4: Calculate the priority value YX based on the previous results, and perform error correction according to the order of priority values ​​YX; Step S5: Review and store the corrected results.

2. An intelligent quality inspection and error correction system for real estate registration and confirmation archives, characterized in that: It includes an image unit, a layout unit, a standardization unit, a quality inspection unit, an evaluation unit, an error correction unit, an auditing unit, a manual unit, and a storage unit. The image unit enhances the archival image data; the layout unit performs layout analysis on the enhanced image data; the standardization unit standardizes and unifies the text format; the quality inspection unit performs parallel quality inspection of the archival data across five dimensions; the evaluation unit uses priority values ​​YX and sorts the data; the error correction unit automatically repairs the data; the auditing unit scores the error correction results; the manual unit performs manual error correction; and the storage unit stores the corrected archival data. The image unit is used to scan real estate archive image data, and the image unit performs image enhancement processing on the collected image data. The image unit calculates the pixel value of the scanned image using a weighted average method and converts it into a grayscale image. The calculated pixel value is linearly normalized, and after normalization, bilateral filtering is used for noise reduction. Adaptive histogram equalization is used for contrast enhancement, and then Gamma correction is used to complete the image enhancement processing.

3. The intelligent quality inspection and error correction system for real estate registration archives according to claim 2, characterized in that: The layout unit analysis performs connected component analysis on the enhanced binary image to distinguish between foreground and background. It uses a YOLOv8-seg model trained on a real estate archive layout dataset to detect and classify the title, table, rights holder information area, signature and official seal area, and attached drawing area. The natural reading order is reconstructed based on the geometric position of the area and prior knowledge of the archive template. The table area is further subdivided into cells through horizontal / vertical projection. The red seal interference area is filtered based on the HSV color space to complete the layout analysis.

4. The intelligent quality inspection and error correction system for real estate registration archives according to claim 2, characterized in that: The standardization unit uses OCR to extract editable text content from the image area after the layout unit is analyzed and performs text standardization processing, converting all text into UTF-8 encoding, and then converting all text into archive data with a unified format.

5. The intelligent quality inspection and error correction system for real estate registration archives according to claim 2, characterized in that: The quality inspection unit performs parallel quality inspections on the standardized archive data in five dimensions: format standardization, semantic rationality, ownership consistency, spatiotemporal logical coherence, and image-text consistency. This generates five corresponding abnormal records: format abnormal records, semantic abnormal records, ownership abnormal records, spatiotemporal abnormal records, and image-text abnormal records.

6. The intelligent quality inspection and error correction system for real estate registration archives according to claim 2, characterized in that: The evaluation unit receives anomaly records from the quality inspection unit, calculates the priority value YX, sorts them by priority, and determines the processing order. The calculation formula is as follows: In the formula, QS is the number of abnormal ownership information entries, QZ is the total number of ownership information entries, YX is the number of fields affected by the abnormal record, GL is the number of associated files affected by the abnormal record, k1 and k2 are weight coefficients. When the abnormal record involves abnormal ownership relationships and spatiotemporal abnormalities, the weight coefficient k2 is automatically increased by 0.1-0.

2. The evaluation unit arranges the calculated priority values ​​YX in reverse order.

7. The intelligent quality inspection and error correction system for real estate registration archives according to claim 2, characterized in that: The error correction unit retrieves error correction schemes from the error correction strategy knowledge base and automatically corrects the data. For records with abnormal format, it completes and converts the format; for records with abnormal semantics, it adjusts logical contradictions and unifies consistency; for records with abnormal ownership, it updates the ownership status; for records with abnormal spatiotemporal relationships, it adjusts the time order and coordinate address; and for records with abnormal image and text, it corrects them based on the high-confidence modality.

8. The intelligent quality inspection and error correction system for real estate registration archives according to claim 2, characterized in that: The review unit scores the error correction results from the error correction unit, and the review unit calculates the error correction score value DF. The formula for calculating the error correction score value DF is as follows: In the formula, ZX is the inherent confidence of the policy, CF is the context sufficiency score, YZ is the consistency check score, and LS is the historical score rate.

9. The intelligent quality inspection and error correction system for real estate registration archives according to claim 2, characterized in that: The review unit compares the calculated error correction score DF with its internal error correction threshold Y. When the error correction score DF > the error correction threshold Y, the review unit sends a pass instruction to the storage unit. The storage unit receives the storage instruction and stores the file data with the error correction score DF > the error correction threshold Y. When the error correction score DF ≤ the error correction threshold Y, the review unit sends a manual instruction to the manual unit. The manual unit receives the file data with the error correction score DF ≤ the error correction threshold Y, performs manual review and error correction, and then sends it to the storage unit for storage.

10. The intelligent quality inspection and error correction system for real estate registration archives according to claim 2, characterized in that: The error correction unit is equipped with an error correction strategy knowledge base, which is organized in the form of a knowledge graph. The error type is the head entity and the specific error correction strategy is the tail entity. The edge attributes record the strategy priority, default confidence, applicable scenario tags and historical success rate weights. The error correction strategy knowledge base supports strategy registration, retrieval, dynamic weight adjustment, confidence factor update and strategy execution log correlation analysis.

Images