A house property information verification method and system based on image technology

Abstract

The application discloses a kind of based on image technology's house property information verification method and system, it is related to house property information verification technical field, the present application includes the following steps: image acquisition: using the acquisition equipment equipped with multiple functions, by adjustable illumination, automatic paper leveling, multi-mode scanning and real-time quality monitoring etc., obtain high-quality and comprehensive paper house property certificate file image;Paper house property certificate file image is collected using the acquisition equipment equipped with adjustable illumination system and automatic paper leveling function;The present application, adopt multi-mode scanning to obtain comprehensive information, effectively reduce information omission risk, image preprocessing accurately removes noise, enhances contrast and repairs information loss, provides clear image for subsequent processing, a variety of accurate feature analysis is added in feature extraction link, forms rich and accurate feature vector, greatly enhances the judgment ability to file authenticity and integrity, to significantly improve the accuracy and reliability of verification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of real estate information verification, in particular to a real estate information verification method and system based on image technology. BACKGROUND

[0002] Real estate plays an important role in people's life and economic activities, and the accuracy of real estate information is crucial for real estate transactions, mortgage, inheritance and many other aspects. In real estate related businesses, paper real estate certificate documents are the key basis for confirming real estate ownership and related information. Traditional real estate information verification mainly relies on manual methods, and staff need to manually check paper documents and compare database information. This method has many drawbacks, such as low efficiency of manual verification, time-consuming and laborious when dealing with a large number of real estate certificate documents. Moreover, manual operation is easily influenced by subjective factors, and the judgment standards of different verification personnel may differ, making it difficult to ensure the consistency of the verification results. Long-term manual viewing can also cause visual fatigue, increasing the probability of errors and affecting the accuracy of verification. In addition, paper documents may be damaged or stained during storage and transmission, further increasing the difficulty of manual verification.

[0003] Paper real estate certificate documents themselves have certain fragility and risk of being tampered with. During long-term storage, paper may be affected by natural environmental factors (such as humidity, light, etc.) and deteriorate, with problems such as blurred writing and damaged paper, making it difficult to accurately identify the information in the document and causing great difficulty in verification work. At the same time, with the development of technology, the means of forging and tampering with paper documents are becoming increasingly complex, and some unscrupulous individuals may create false real estate certificate documents by forging seals and modifying text content. Traditional manual verification methods often fail to effectively identify these forgeries and tampering, allowing false documents to pass, causing serious safety hazards in the real estate transaction market, damaging the legal rights and interests of all parties, and disrupting the market order.

[0004] To improve verification efficiency and accuracy, image technology has been gradually applied in the field of real estate information verification. However, existing image technology-based verification methods still have many shortcomings. In the image acquisition stage, due to the diversity of paper documents (such as different paper quality, printing clarity, etc.) and possible wrinkles, stains, etc., the quality of the acquired images is uneven, affecting subsequent identification and analysis. Moreover, traditional image acquisition devices and methods may not be able to fully obtain hidden information (such as anti-forgery marks, watermarks, etc.) in the document, resulting in information omission. In the image analysis process, the comparison with standard templates is relatively single, and is not sensitive enough to some low-level errors (such as similar character confusion, partial information missing or duplication, etc.) generated by scanning recognition, making it difficult to discover and correct in a timely manner. SUMMARY

[0005] The purpose of this invention is to solve the problems of low efficiency, poor accuracy, fragility and forgery of paper-based property certificate verification, as well as the problems of poor image quality, easy omission of basic errors in analysis and lack of effective comparison with similar information in existing image technology verification. Therefore, this invention proposes a property information verification method and system based on image technology.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A method for verifying real estate information based on image technology includes the following steps:

[0008] S1. Image Acquisition: Using acquisition equipment equipped with multiple functions, high-quality and comprehensive images of paper property ownership documents are obtained through adjustable lighting, automatic paper leveling, multi-mode scanning, and real-time quality monitoring.

[0009] S2. Image preprocessing: To address the issues of noise, contrast, and missing information in the acquired images, corresponding algorithms are used to perform noise reduction, contrast enhancement, and information restoration operations to optimize image quality.

[0010] S3. Feature Extraction: Based on the extraction of conventional features, logical structure analysis is added to refine text features and extract anti-counterfeiting element features, forming a rich and accurate file feature vector.

[0011] S4. Information Matching and Verification: The features of the document to be verified are compared with the same type and standard templates. The distribution pattern of similar information is statistically analyzed and the error type is identified by machine learning model. At the same time, the similarity threshold is dynamically calculated for accurate judgment.

[0012] S5. Results Output and Report Generation: Generate a verification report containing comprehensive information, output it in multiple formats, mark and record suspicious information areas, send the report to human reviewers and receive feedback for processing, thus completing the verification process.

[0013] Furthermore, the specific steps of S1 are as follows:

[0014] Images of paper property ownership certificates are captured using a data acquisition device equipped with an adjustable lighting system and automatic paper leveling function. Sensors detect the light exposure on the paper, and the lighting intensity and angle are automatically adjusted using the following formula: Among them, I adj This is the lighting intensity adjustment value, where k is the adjustment coefficient, set according to equipment characteristics and experience. t It is the target light intensity, I o It is the currently detected light intensity on the paper surface, I maxIt is the maximum lighting intensity provided by the lighting equipment. Based on this formula, the lighting intensity is adjusted proportionally according to the difference between the current lighting and the target lighting to ensure that the paper receives the correct lighting.

[0015] At the same time, negative pressure adsorption or flexible flattening device is used to reduce paper wrinkles; three modes are used: optical scanning, infrared scanning and ultraviolet scanning. Optical scanning is used to acquire the basic content of the document, infrared scanning is used to detect the anti-counterfeiting marks or watermarks hidden inside the paper, and ultraviolet scanning is used for documents containing fluorescent anti-counterfeiting materials. The image data of the three modes are then fused together.

[0016] During the acquisition process, image quality is monitored in real time using an image sharpness assessment algorithm and noise level detection methods; the image sharpness assessment algorithm calculates the image sharpness index C using the variance method. Where M and N are the number of rows and columns of the image, respectively, and f(i,j) is the pixel value at coordinates (i,j) in the image. It is the average pixel value of the image; the larger the sharpness index C, the sharper the image; a sharpness threshold C is set. min When C < C min If the image is deemed too blurry, it needs to be recaptured; if the image is too blurry, the image for that page of the file will be automatically recaptured.

[0017] Noise level detection first calculates the difference between the original image and the smoothed image, e(i,j) = f(i,j) - g(i,j), where f(i,j) are the pixel values ​​of the original image and g(i,j) are the pixel values ​​of the smoothed image; then, it calculates the noise level N. Set noise threshold N max When N>N max If the noise level is too high, the image needs to be reacquired.

[0018] Furthermore, the specific operation steps of S2 are as follows:

[0019] For salt-and-pepper noise caused by paper texture and stains in the acquired images, a median filtering algorithm is used; for Gaussian noise caused by uneven lighting or interference from electronic devices, a Gaussian filtering algorithm is used, and the filtering parameters are automatically adjusted according to the spatial distribution characteristics of the noise.

[0020] An adaptive histogram equalization algorithm is used to enhance image contrast, making text and patterns clearer. For information loss caused by paper damage or stains, interpolation algorithms based on neighboring pixel information and texture synthesis algorithms are used for repair, and the corresponding algorithm parameters are selected according to the size of the missing area and the features of surrounding pixels.

[0021] Furthermore, the specific operation steps of S3 are as follows:

[0022] In addition to extracting text, seals, and anti-counterfeiting features, the analysis of the logical structure of the document content is added, including the relative positional relationship and arrangement order between different information blocks; the extraction of text features is refined, including the slant angle of the text and the variation in the thickness of the strokes;

[0023] For anti-counterfeiting elements such as microtext and rainbow printing patterns in paper property certificates, their features are extracted using image and image analysis algorithms;

[0024] The size, spacing, and arrangement of microtext, as well as the color transition characteristics of the rainbow printing pattern, are analyzed. The size and spacing characteristics of the microtext are quantified by calculating the pixel statistics of the microtext region. Let the pixel value of the microtext region be m(i, j), i = 1, 2, ..., M. w j = 1, 2, ..., N w M w and N w These are the number of rows and columns of the microtext area; the average size S of the microtext is calculated as follows: Where n w It is the number of effective pixels in the microtext area, which is determined by setting a pixel value threshold;

[0025] The microtext spacing D is obtained by calculating the average distance between adjacent microtext regions. First, the boundaries of the microtext regions are detected, then the distance between adjacent boundaries is calculated, and finally the average value is calculated to obtain the spacing D. These quantified features describe microtext more accurately and are used for anti-counterfeiting verification.

[0026] Furthermore, the specific steps for calculating the character tilt angle in S3 are as follows:

[0027] The text tilt angle is calculated by determining the tilt angle of the smallest bounding rectangle of the text region. Specifically, the pixel coordinates of the text region are set as (x...). i y i ), i = 1, 2, ..., n, where n is the number of pixels in the text region; first, calculate the centroid coordinates (x, y) of the text region. c y c ): Then calculate the covariance matrix C:

[0028] Finally, by calculating the eigenvalues ​​and eigenvectors of the covariance matrix, the tilt angle θ of the minimum bounding rectangle is obtained: Where C 11 C 12 C 22 It is an element of the covariance matrix C; this angle serves as a text tilt feature, used for subsequent matching and analysis.

[0029] Furthermore, the specific operation steps of S4 are as follows:

[0030] The characteristics of the documents to be verified are compared with those of similar property ownership certificates and standard templates. Statistical analysis is used to identify patterns in the distribution of similar information within these documents, including the area range of properties in the same region and the format of the property owner's name, to initially screen for suspicious information. Specifically:

[0031] Collect existing and verified property ownership documents of the same type, including property area, owner's name, property location, and building structure information, and store them in the database according to a preset format and structure to ensure the accuracy and completeness of the data. At the same time, classify and organize the data according to the dimensions of region and property type.

[0032] The feature extraction step extracts the features of the document to be verified and forms a complete feature vector of the document to be verified.

[0033] Regarding property area: Calculate the mean and standard deviation of the area of ​​similar properties to determine a reasonable range and screen for suspicious information;

[0034] For property owner names: relevant records are filtered from a database of similar properties to extract property owner name information; statistics are compiled on all property owner surnames to calculate the frequency of each surname; based on the statistical results, a surname frequency threshold is set; property owner surnames are extracted from the documents to be verified, and it is determined whether their frequency is lower than the threshold; if it is lower than the threshold, it is determined that there is an anomaly, the document is marked as having a suspicious property owner name format, and relevant information is recorded, including the document number, property owner name, and the frequency of the surname in similar documents;

[0035] A model is built using machine learning algorithms trained on verified document data to identify common error types in scanning, including similar character confusion, information duplication or missing error patterns, to help determine whether there are class errors.

[0036] When calculating the similarity between the document to be verified and the standard template and similar property information, the similarity thresholds of different features are dynamically adjusted according to document type, age, and region factors.

[0037] Furthermore, the specific steps for adjusting the similarity thresholds of different features in S4 are as follows:

[0038] After obtaining the document type, date, and region factors, the document type, date, and region factors are evaluated respectively to obtain type score, date score, and region score. After normalization, a base circle is established with the sum of the type score and date score as the radius, and a cone model is established with the region score as the height. The volume of the cone model is calculated and recorded as the threshold value. The obtained threshold value is compared with several preset threshold value intervals. The preset threshold value intervals correspond to different similarity thresholds. When the threshold value interval to which the threshold value belongs is determined, the similarity threshold is determined and substituted into the similarity calculation process between the document to be verified and the standard template and similar real estate information.

[0039] Furthermore, the specific operation steps of S5 are as follows:

[0040] Based on the information matching and verification results, a paper property certificate information verification report is generated. The report includes basic document information, detailed image acquisition information, specifically the acquisition mode, quality assessment results, number of re-acquisitions, description of the preprocessing process, feature extraction results, comparison with standard templates and similar property information, matching details, and the final verification conclusion. The report is presented in various formats, including charts to show changes in feature similarity and images to mark suspicious information areas. It supports multiple output formats, including PDF, HTML, and XML, for archiving and retrieval.

[0041] Furthermore, the S5 process for handling suspicious information areas is as follows:

[0042] Establish standardized guidelines for labeling suspicious information areas, including specifying the symbols, colors, shapes, and format requirements for the labeled content; categorize suspicious information into types such as stamps, text information, and overall images; and develop corresponding labeling methods and prompts for different types of suspicious information.

[0043] Use image processing software or annotation tools to draw rectangles and add text annotations to images;

[0044] When generating the report, the suspicious information areas identified during the information matching and verification process are annotated using the annotation tool according to the established specifications.

[0045] At the same time, a suspicious area information record table is established to record the location, suspicious type, relevant feature values, and corresponding file page number of each suspicious area; this record information, together with the labeled images, will provide detailed data support for manual review.

[0046] The generated report will be sent to human reviewers via the system's internal task allocation mechanism or email. When sending the email, the importance of the report and the key points of the review will be highlighted in the subject line and explained in the body of the email, reminding reviewers to pay special attention to areas of suspicious information. At the same time, a link to view the report or an attachment download method will be provided to ensure that reviewers can easily obtain the report.

[0047] After receiving the report, the auditors reviewed the report content according to the guidelines, paying attention to the marked suspicious information areas; they then conducted in-depth analysis of the suspicious areas based on their professional knowledge and experience; and during the review process, they used auditing tools to record their audit comments and the problems they found.

[0048] After completing the audit, the auditors will provide feedback on the audit results, which will include the final judgment on suspicious areas, handling suggestions, and suggestions for improving the verification system.

[0049] Update the verification status of the documents based on the audit results, and take appropriate follow-up actions based on the processing suggestions.

[0050] A real estate information verification system based on image technology includes:

[0051] The image acquisition module consists of a device control unit, a multi-mode scanning unit, and an image quality monitoring unit.

[0052] The equipment control unit operates the acquisition device, adjusts the lighting according to the light conditions, reduces paper wrinkles, and ensures high-quality image acquisition; the multi-mode scanning unit executes three scanning modes and merges the data to obtain comprehensive document information; the image quality monitoring unit monitors the image quality in real time, judges based on thresholds and triggers re-sampling to ensure image quality.

[0053] The image preprocessing module consists of a noise removal unit, an image enhancement unit, and an information restoration unit.

[0054] The noise removal unit uses appropriate filtering algorithms and adjusts parameters to remove noise for different types of noise; the image enhancement unit uses algorithms to enhance image contrast for easier subsequent processing; and the information restoration unit repairs information loss caused by damage or stains.

[0055] The feature extraction module consists of a conventional feature extraction unit, a logical structure analysis unit, a text refinement feature unit, and an anti-counterfeiting element feature unit.

[0056] The standard feature extraction unit is used to extract basic features of text and seals; the logical structure analysis unit is used to analyze the logical structure of document content to assist in judgment; the text refinement feature unit is used to refine text features and accurately depict text; the anti-counterfeiting element feature unit extracts anti-counterfeiting element features for anti-counterfeiting verification.

[0057] The information matching and verification module consists of a data comparison unit of the same type, a machine learning model unit, and a similarity calculation unit.

[0058] The data comparison unit is used to compare similar data and screen for suspicious information such as property area and owner's name; the machine learning model unit uses the model to identify common scanning error types; the similarity calculation unit dynamically calculates the similarity threshold to accurately determine the file matching degree.

[0059] The results output and report generation module includes a report generation unit, a suspicious information annotation unit, and a report sending and feedback unit.

[0060] The report generation unit generates verification reports containing comprehensive information and outputs them in multiple formats; the suspicious information annotation unit is used to annotate suspicious areas and record relevant information to assist manual review; the report sending and feedback unit sends reports to manual reviewers, receives and processes feedback, and improves the verification process.

[0061] Compared with the prior art, the beneficial effects of the present invention are:

[0062] (1) In this invention, multi-mode scanning is used to obtain comprehensive information, which effectively reduces the risk of information omission. Image preprocessing accurately removes noise, enhances contrast and repairs missing information, providing clear images for subsequent processing. The feature extraction stage adds a variety of precise feature analysis to form rich and accurate feature vectors, which greatly enhances the ability to judge the authenticity and integrity of documents, thereby significantly improving the accuracy and reliability of verification.

[0063] (2) In this invention, the optimization measures during image acquisition ensure the rapid acquisition of high-quality images, the real-time monitoring and automatic re-acquisition mechanism avoids ineffective work, the machine learning model helps to quickly identify common errors and shorten the error screening time, the dynamic adjustment of the similarity threshold improves the accuracy and efficiency of matching judgment, and at the same time enhances the system's adaptability to different types of files, effectively improving the overall verification efficiency.

[0064] (3) The present invention generates detailed verification reports that are presented intuitively in multiple ways and support multi-format output, optimizes information transmission and recording, improves the labeling standards and processing procedures for suspicious information areas, facilitates manual reviewers to quickly locate and understand problems, improves review efficiency, and achieves complementary advantages through close collaboration between the system and manual review, improves the real estate information verification work, effectively protects the security of real estate transactions, and comprehensively optimizes the verification process. Attached Figure Description

[0065] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0066] Figure 1 This is the overall system block diagram of the present invention. Detailed Implementation

[0067] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] It should be understood that the terms “comprising” and “including” used in this disclosure and claims indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0069] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure. As used in this disclosure and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this disclosure and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0070] like Figure 1 As shown, a method for verifying real estate information based on image technology includes the following steps:

[0071] Image acquisition: Utilizing acquisition equipment equipped with multiple functions, high-quality and comprehensive images of paper property ownership documents are obtained through adjustable lighting, automatic paper leveling, multi-mode scanning, and real-time quality monitoring.

[0072] Images of paper property ownership certificates are captured using a data acquisition device equipped with an adjustable lighting system and automatic paper leveling function. Sensors detect the light exposure on the paper, and the lighting intensity and angle are automatically adjusted using the following formula: Among them, I adj This is the lighting intensity adjustment value, where k is the adjustment coefficient, set according to equipment characteristics and experience. t It is the target light intensity (the ideal light intensity on the paper surface, set according to paper type and scanning requirements), I o It is the currently detected light intensity on the paper surface, I maxThis is the maximum illumination intensity that the lighting equipment can provide. Based on this formula, the illumination intensity is adjusted proportionally according to the difference between the current illumination and the target illumination to ensure that the paper receives appropriate lighting. At the same time, negative pressure adsorption or flexible flattening devices are used to reduce paper wrinkles. Three modes are employed: high-resolution optical scanning, infrared scanning, and ultraviolet scanning. High-resolution optical scanning acquires the basic content of the document, infrared scanning detects hidden anti-counterfeiting marks or watermarks inside the paper, and ultraviolet scanning is used for documents containing fluorescent anti-counterfeiting materials. The image data from the three modes are then fused together.

[0073] During the acquisition process, image quality is monitored in real time using an image sharpness assessment algorithm and noise level detection methods; the image sharpness assessment algorithm calculates the image sharpness index C using the variance method. Where M and N are the number of rows and columns of the image, respectively, and f(i,j) is the pixel value at coordinates (i,j) in the image. It is the average pixel value of the image; the larger the sharpness index C, the sharper the image; a sharpness threshold C is set. min (Specific values ​​are determined based on experiments and experience), when C < C min If the image is deemed too blurry, it needs to be recaptured. If the image is too blurry, the file will be automatically recaptured, with a maximum of three attempts.

[0074] Noise level detection first calculates the difference between the original image and the smoothed image, e(i,j) = f(i,j) - g(i,j), where f(i,j) are the pixel values ​​of the original image and g(i,j) are the pixel values ​​of the smoothed image; then, it calculates the noise level N. Set noise threshold N max (The specific determination depends on the characteristics and requirements of the image sensor), when N > N max At that time, it was determined that there was too much noise and that the image needed to be reacquired.

[0075] Image preprocessing: To address the issues of noise, contrast, and missing information in the acquired images, corresponding algorithms are employed to perform noise reduction, contrast enhancement, and information restoration operations, thereby optimizing image quality;

[0076] For salt-and-pepper noise caused by paper texture and stains in the acquired images, a median filtering algorithm is used; for Gaussian noise caused by uneven lighting or interference from electronic devices, a Gaussian filtering algorithm is used, and the filtering parameters are automatically adjusted according to the spatial distribution characteristics of the noise; an adaptive histogram equalization algorithm is used to enhance image contrast, making text and patterns clearer; for information loss caused by paper damage or stains, interpolation algorithms based on neighborhood pixel information and texture synthesis algorithms are used for repair, and appropriate algorithm parameters are selected according to the size of the missing area and the characteristics of surrounding pixels.

[0077] Feature extraction: Based on the extraction of conventional features, logical structure analysis is added to refine text features and accurately extract anti-counterfeiting element features, forming a rich and accurate document feature vector;

[0078] In addition to extracting traditional features such as text, seals, and anti-counterfeiting marks, the analysis of the document's logical structure is enhanced, including the relative positional relationships and arrangement order between different information blocks. Text feature extraction is refined, including text tilt angle and stroke thickness variations. The text tilt angle is calculated by determining the tilt angle of the smallest bounding rectangle of the text region. Specifically, the pixel coordinates of the text region are set as (x...). i y i ), i = 1, 2, ..., n, n (where n is the number of pixels in the text region); first calculate the centroid coordinates (x, y) of the text region. c y c ): Then calculate the covariance matrix C:

[0079] Finally, by calculating the eigenvalues ​​and eigenvectors of the covariance matrix, the tilt angle θ of the minimum bounding rectangle is obtained: Where C 11 C 12 C 22 It is an element of the covariance matrix C; this angle serves as a text tilt feature, used for subsequent matching and analysis.

[0080] This study targets the microtext and rainbow-printed patterns, special anti-counterfeiting elements in paper property ownership certificates, and utilizes high-resolution images and specialized image analysis algorithms to accurately extract their features. It analyzes the size, spacing, and arrangement of the microtext, as well as the color transition characteristics of the rainbow-printed patterns. For the size and spacing features of the microtext, quantification is achieved by calculating the pixel statistics of the microtext region. Let the pixel value of the microtext region be m(i, j), i = 1, 2, ..., M. w j = 1, 2, ..., N w M w and N w These are the number of rows and columns of the microtext area; the average size S of the microtext (in pixels) is calculated as follows: Where n w It is the number of effective pixels (i.e., pixels belonging to the text portion) in the miniature text area, which is determined by setting a pixel value threshold;

[0081] The microtext spacing D (in pixels) is obtained by calculating the average distance between adjacent microtext regions. First, the boundaries of the microtext regions are detected, then the distance between adjacent boundaries is calculated, and finally the average value is taken to obtain the spacing D. These quantified features describe microtext more accurately and are used for anti-counterfeiting verification.

[0082] Information matching and verification: The features of the document to be verified are compared with those of the same type and standard templates. The distribution patterns of similar information are statistically analyzed and machine learning models are used to identify error types. At the same time, the similarity threshold is dynamically calculated for accurate judgment.

[0083] The characteristics of the document to be verified are compared with those of similar property ownership certificates and standard templates. Statistical analysis is used to identify patterns in the distribution of similar information within the same type of document, including the range of property area in the same region and the format of the property owner's name, to initially screen for suspicious information; specifically:

[0084] A large amount of existing, verified property ownership documents of the same type are collected, including information such as property area, owner's name, property location, and building structure. This information is stored in a database according to a specific format and structure to ensure data accuracy and completeness. The data is also categorized, for example, by region and property type (residential, commercial, etc.). Feature extraction steps are used to extract features from the documents to be verified, forming a complete feature vector. For property area: the mean and standard deviation of similar property areas are calculated to determine a reasonable range and screen for suspicious information. For owner's name: relevant records are selected from the database of similar property information, and owner's name information is extracted. All owner surnames are statistically analyzed, and the frequency of each surname is calculated. Based on the statistical results, a surname frequency threshold is set. The owner's surname is extracted from the documents to be verified, and its frequency is checked against this threshold. If it is below the threshold, an anomaly may exist; the document is marked as having a suspicious owner's name format, and relevant information (such as document number, owner's name, and the frequency of the surname in similar documents) is recorded. Further analysis can be conducted using other information to determine if there are any problems.

[0085] A model is built using machine learning algorithms trained on a large amount of verified document data to identify common error types in scanning, including error patterns such as similar character confusion ("0" and "O", "1" and "l"), duplicate or missing information, etc., to help determine whether such low-level errors exist.

[0086] When calculating the similarity between the document to be verified and the standard template and similar property information, the similarity thresholds of different features are dynamically adjusted according to document type, year, and region factors. Specifically, after obtaining the document type, year, and region factors, the document type, year, and region factors are evaluated respectively to obtain type score, year score, and region score. After normalization, a base circle is established with the sum of the type score and the year score as the radius, and a cone model is established with the region score as the height. The volume of the cone model is calculated and recorded as the threshold value. The obtained threshold value is compared with several preset threshold value intervals. The preset threshold value intervals correspond to different similarity thresholds. When the threshold value interval to which the threshold value belongs is determined, the similarity threshold is determined and substituted into the similarity calculation process between the document to be verified and the standard template and similar property information.

[0087] Results Output and Report Generation: Generate a verification report containing multiple aspects of information, output it in various formats, mark and record suspicious information areas, send the report to human reviewers and receive feedback for processing, and complete the verification process;

[0088] A paper-based property ownership certificate information verification report is generated based on the information matching and verification results. The report includes basic document information, detailed image acquisition information (acquisition mode, quality assessment results, number of re-acquisitions, etc.), a description of the preprocessing process, feature extraction results, comparison with standard templates and similar property information, matching details, and the final verification conclusion. The report is presented in an intuitive and easy-to-understand manner, such as using charts to display changes in feature similarity and annotating suspicious information areas with images. It supports multiple output formats (such as PDF, HTML, XML, etc.) for easy archiving and retrieval. The processing procedure for suspicious information areas is as follows:

[0089] Establish standardized guidelines for labeling suspicious information areas, including specifying the symbols, colors, shapes, and format requirements for the labeled content. Classify potentially suspicious information into categories such as: stamp-related (including blurry stamps, stamp shapes that do not conform to standards, abnormal stamp colors, etc.), textual information-related (such as blurry text, confusion of similar characters, duplicate or missing information, significant differences between key information and standard templates, etc.), and overall image-related (such as low image clarity, shadows, or obstructions that affect information recognition). For different types of suspicious information, develop corresponding labeling methods and prompts to enable human reviewers to quickly understand the suspicious points. Utilize image processing software or develop specialized labeling tools with functions such as drawing rectangles on images and adding text annotations. Ensure that the labeling tools can be integrated with the real estate information verification system for easy direct use during report generation. When generating reports, label the suspicious information areas identified during information matching and verification using the labeling tools according to the established standards. For areas with suspicious seals, draw accurate rectangles based on the seal's location and extent, and label them "Suspicious Seal" along with the specific reason; for areas with suspicious text information, if similar characters are found to be confused, select the confused characters and label them with details such as "Similar characters, possibly confused (e.g., '0' and 'O')"; for overall image problems, label them "Image Quality Issue" along with relevant descriptions at the image edges or in blank areas.

[0090] Simultaneously, a suspicious area information record table is established, recording the location of each suspicious area (e.g., the coordinates of the top left and bottom right corners of the rectangle), the type of suspiciousness, relevant feature values ​​(e.g., text feature vector values, similarity values ​​with the standard template, etc.), and the corresponding document page number (if it is a multi-page document). This recorded information, along with the labeled images, provides detailed data support for manual review. The generated report is sent to manual reviewers via the system's internal task allocation mechanism or email. When sending the email, the subject line should clearly state "Property Certificate Verification Report - Suspicious Document Requires Review," and the body of the email should briefly explain the importance of the report and the key points of the review, reminding reviewers to focus on suspicious information areas. A link to view the report (if it is in online format) or an attachment download method (if it is in PDF format, etc.) should be provided to ensure that reviewers can easily access the report. After receiving the report, reviewers should carefully review the report content according to the specifications and training requirements, focusing on the labeled suspicious information areas. They should then conduct in-depth analysis of the suspicious areas based on their professional knowledge and experience, such as comparing with real samples of seals and verifying textual information by consulting relevant property archives. During the review process, reviewers use review tools to record their opinions and identified issues, such as confirming whether suspicious areas do indeed have problems, the severity of the problems, and possible causes. After completing the review, reviewers feed the results back to the property information verification system. The review results should include the final judgment on suspicious areas (confirmation of problems, exclusion of suspicion, or need for further investigation), processing suggestions (such as requesting supplementary materials, rescanning documents, rejecting documents, etc.), and suggestions for improving the verification system (such as improving labeling accuracy, adding feature extraction dimensions, etc.). The system updates the document verification status based on the review results and performs corresponding follow-up operations according to the processing suggestions, such as notifying the applicant to supplement materials or marking and archiving documents with serious problems for later querying and statistical analysis.

[0091] A real estate information verification system based on image technology includes:

[0092] The image acquisition module consists of a device control unit, a multi-mode scanning unit, and an image quality monitoring unit;

[0093] The equipment control unit operates the acquisition device, adjusts the lighting according to the lighting conditions, reduces paper wrinkles, and ensures the acquisition of high-quality images; the multi-mode scanning unit executes three scanning modes and merges the data to obtain comprehensive document information; the image quality monitoring unit monitors the image quality in real time, judges based on thresholds and triggers re-sampling to ensure that the image is qualified.

[0094] The image preprocessing module consists of a noise removal unit, an image enhancement unit, and an information restoration unit;

[0095] The noise removal unit uses appropriate filtering algorithms and adjusts parameters to remove noise for different types of noise; the image enhancement unit uses algorithms to enhance image contrast for easier subsequent processing; and the information restoration unit repairs information loss caused by damage or stains.

[0096] The feature extraction module consists of a conventional feature extraction unit, a logical structure analysis unit, a text refinement feature unit, and an anti-counterfeiting element feature unit;

[0097] The standard feature extraction unit is used to extract basic features such as text and seals; the logical structure analysis unit is used to analyze the logical structure of the document content to assist in judgment; the text refinement feature unit is used to refine text features and accurately depict text; the anti-counterfeiting element feature unit extracts special anti-counterfeiting element features for anti-counterfeiting verification.

[0098] The information matching and verification module includes a data comparison unit for similar types, a machine learning model unit, and a similarity calculation unit;

[0099] The data comparison unit compares similar data to screen for suspicious information such as property area and owner's name; the machine learning model unit uses models to identify common scanning error types; and the similarity calculation unit dynamically calculates similarity thresholds to accurately determine file matching degree.

[0100] The results output and report generation module includes a report generation unit, a suspicious information annotation unit, and a report sending and feedback unit.

[0101] The report generation unit generates verification reports containing comprehensive information and outputs them in multiple formats; the suspicious information marking unit marks suspicious areas and records relevant information to assist manual review; the report sending and feedback unit sends reports to manual reviewers, receives and processes feedback, and improves the verification process.

[0102] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for verifying real estate information based on image technology, characterized in that, include; S1. Image Acquisition: Using acquisition equipment equipped with multiple functions, high-quality and comprehensive images of paper property ownership documents are obtained through adjustable lighting, automatic paper leveling, multi-mode scanning, and real-time quality monitoring. S2. Image preprocessing: To address the issues of noise, contrast, and missing information in the acquired images, corresponding algorithms are used to perform noise reduction, contrast enhancement, and information restoration operations to optimize image quality. S3. Feature Extraction: Based on the extraction of conventional features, logical structure analysis is added to refine text features and extract anti-counterfeiting element features, forming a rich and accurate file feature vector. The specific operation steps of S3 are as follows: In addition to extracting text, seals, and anti-counterfeiting features, the analysis of the logical structure of the document content is added, including the relative positional relationship and arrangement order between different information blocks; the extraction of text features is refined, including the slant angle of the text and the variation in the thickness of the strokes; For anti-counterfeiting elements such as microtext and rainbow printing patterns in paper property certificates, image analysis algorithms are used to extract their features. This study analyzes the size, spacing, and arrangement of microtext, as well as the color transition characteristics of rainbow printing patterns. For the size and spacing characteristics of the microtext, quantification is achieved by calculating the pixel statistics of the microtext region, where the pixel value of the microtext region is set as... ,in and These are the number of rows and columns of the microtext area; the average size S of the microtext is calculated as follows: ;in It is the number of effective pixels in the microtext area, which is determined by setting a pixel value threshold; The microtext spacing D is obtained by calculating the average distance between adjacent microtext regions; first, the boundaries of the microtext regions are detected, then the distance between adjacent boundaries is calculated, and finally the average value is calculated to obtain the spacing D; these quantified features describe microtext more accurately and are used for anti-counterfeiting verification. S4. Information Matching and Verification: Compare the features of the document to be verified with the features of similar real estate certificate documents and standard templates, statistically analyze the distribution patterns of similar information, use machine learning models to identify error types, and dynamically calculate similarity thresholds for accurate judgment. S5. Results Output and Report Generation: Generate a verification report containing comprehensive information, output it in multiple formats, mark and record suspicious information areas, send the report to human reviewers and receive feedback for processing, thus completing the verification process.

2. The method for verifying real estate information based on image technology according to claim 1, characterized in that, The specific steps of S1 are as follows: Images of paper property ownership certificates are captured using a data acquisition device equipped with an adjustable lighting system and automatic paper leveling function. Sensors detect the light exposure on the paper, and the lighting intensity and angle are automatically adjusted using the following formula: in, This is the lighting intensity adjustment value. It is an adjustment factor, set according to equipment characteristics and experience. It is the target light intensity. This is the currently detected light intensity on the paper surface. It is the maximum lighting intensity provided by the lighting equipment. Based on this formula, the lighting intensity is adjusted proportionally according to the difference between the current lighting and the target lighting to ensure that the paper receives the correct lighting. At the same time, negative pressure adsorption or flexible flattening device is used to reduce paper wrinkles; three modes are used: optical scanning, infrared scanning and ultraviolet scanning. Optical scanning is used to acquire the basic content of the document, infrared scanning is used to detect the anti-counterfeiting marks or watermarks hidden inside the paper, and ultraviolet scanning is used for documents containing fluorescent anti-counterfeiting materials. The image data of the three modes are then fused together. During the acquisition process, image quality is monitored in real time using an image sharpness assessment algorithm and noise level detection methods; the image sharpness assessment algorithm calculates the image sharpness index C using the variance method. Where M and N are the number of rows and columns of the image, respectively. The image is in coordinates Pixel value at that location, It is the average pixel value of the image; the larger the sharpness index C, the sharper the image; a sharpness threshold is set. ,when If the image is deemed too blurry, it needs to be re-captured; if the image is too blurry, the file image will be automatically re-captured. Noise level detection first calculates the difference between the original image and the smoothed image. in These are the original image pixel values. These are the smoothed image pixel values; then the noise level N is calculated. Set noise threshold ,when If the noise level is too high, the image needs to be reacquired.

3. The method for verifying real estate information based on image technology according to claim 1, characterized in that, The specific operation steps of S2 are as follows: For salt-and-pepper noise caused by paper texture and stains in the acquired images, a median filtering algorithm is used; for Gaussian noise caused by uneven lighting or interference from electronic devices, a Gaussian filtering algorithm is used, and the filtering parameters are automatically adjusted according to the spatial distribution characteristics of the noise. An adaptive histogram equalization algorithm is used to enhance image contrast, making text and patterns clearer. For information loss caused by paper damage or stains, interpolation algorithms based on neighboring pixel information and texture synthesis algorithms are used for repair, and the corresponding algorithm parameters are selected according to the size of the missing area and the features of surrounding pixels.

4. The method for verifying real estate information based on image technology according to claim 1, characterized in that, The specific steps for calculating the character tilt angle in S3 are as follows: The text tilt angle is calculated by determining the tilt angle of the smallest bounding rectangle of the text region. Specifically, the pixel coordinates of the text region are set as follows: n is the number of pixels in the text region; first, calculate the centroid coordinates of the text region. : Then calculate the covariance matrix C: ; Finally, by calculating the eigenvalues ​​and eigenvectors of the covariance matrix, the tilt angle of the minimum bounding rectangle is obtained. : ;in It is an element of the covariance matrix C; this angle serves as a text tilt feature, used for subsequent matching and analysis.

5. The method for verifying real estate information based on image technology according to claim 1, characterized in that, The specific operation steps of S4 are as follows: The characteristics of the documents to be verified are compared with those of similar property ownership certificates and standard templates. Statistical analysis is used to identify patterns in the distribution of similar information within these documents, including the area range of properties in the same region and the format of the property owner's name, to initially screen for suspicious information. Specifically: Collect existing and verified property ownership documents of the same type, including property area, owner's name, property location, and building structure information, and store them in the database according to a preset format and structure to ensure the accuracy and completeness of the data. At the same time, classify and organize the data according to the dimensions of region and property type. The feature extraction step extracts the features of the document to be verified and forms a complete feature vector of the document to be verified. Regarding property area: Calculate the mean and standard deviation of the area of ​​similar properties to determine a reasonable range and screen for suspicious information; Regarding the property owner's name: relevant records are filtered from the database of similar properties, and the property owner's name information is extracted; statistics are performed on all property owner surnames, and the frequency of each surname is calculated; based on the statistical results, a surname frequency threshold is set; the property owner's surname is extracted from the documents to be verified, and it is determined whether its frequency is lower than the threshold. If the value is below the threshold, an anomaly is identified, and the file is marked as having a suspicious property owner name format. Relevant information is recorded, including the file number, the property owner's name, and the frequency of that surname in similar files. A model is built using machine learning algorithms trained on verified document data to identify common error types in scanning, including similar character confusion, information duplication or missing error patterns, to help determine whether there are class errors. When calculating the similarity between the document to be verified and the standard template and similar property information, the similarity thresholds of different features are dynamically adjusted according to document type, year, and region factors.

6. The method for verifying real estate information based on image technology according to claim 5, characterized in that, The specific steps for adjusting the similarity thresholds of different features in S4 are as follows: After obtaining the document type, date, and region factors, the document type, date, and region factors are evaluated respectively to obtain type score, date score, and region score. After normalization, a base circle is established with the sum of the type score and date score as the radius, and a cone model is established with the region score as the height. The volume of the cone model is calculated and recorded as the threshold value. The obtained threshold value is compared with several preset threshold value intervals. The preset threshold value intervals correspond to different similarity thresholds. When the threshold value interval to which the threshold value belongs is determined, the similarity threshold is determined and substituted into the similarity calculation process between the document to be verified and the standard template and similar real estate information.

7. The method for verifying real estate information based on image technology according to claim 1, characterized in that, The specific operation steps of S5 are as follows: Based on the information matching and verification results, a paper property certificate information verification report is generated. The report includes basic document information, detailed image acquisition information, specifically the acquisition mode, quality assessment results, number of re-acquisitions, description of the preprocessing process, feature extraction results, comparison with standard templates and similar property information, matching details, and the final verification conclusion. The report is presented in various formats, including charts to show changes in feature similarity and images to mark suspicious information areas. It supports multiple output formats, including PDF, HTML, and XML, for archiving and retrieval.

8. The method for verifying real estate information based on image technology according to claim 7, characterized in that, The S5 process for handling suspicious information areas is as follows: Establish standardized guidelines for labeling suspicious information areas, including specifying the symbols, colors, shapes, and format requirements for the labeled content; categorize suspicious information into types such as stamps, text information, and overall images; and develop corresponding labeling methods and prompts for different types of suspicious information. Use image processing software or annotation tools to draw rectangles and add text annotations to images; When generating the report, the suspicious information areas identified during the information matching and verification process are annotated using the annotation tool according to the established specifications. At the same time, a suspicious area information record table is established to record the location, suspicious type, relevant feature values, and corresponding file page number of each suspicious area; this record information, together with the labeled images, will provide detailed data support for manual review. The generated report will be sent to human reviewers via the system's internal task allocation mechanism or email. When sending the email, the importance of the report and the key points of the review will be highlighted in the subject line and explained in the body of the email, reminding reviewers to pay special attention to areas of suspicious information. At the same time, a link to view the report or an attachment download method will be provided to ensure that reviewers can easily obtain the report. After receiving the report, the auditors reviewed the report content according to the guidelines, paying attention to the marked suspicious information areas; they then conducted in-depth analysis of the suspicious areas based on their professional knowledge and experience; and during the review process, they used auditing tools to record their audit comments and the problems they found. After completing the audit, the auditors will provide feedback on the audit results, which will include the final judgment on suspicious areas, handling suggestions, and suggestions for improving the verification system. Update the verification status of the documents based on the audit results, and take appropriate follow-up actions based on the processing suggestions.

9. A system applied to the image-based real estate information verification method according to any one of claims 1-8, characterized in that, include: The image acquisition module consists of a device control unit, a multi-mode scanning unit, and an image quality monitoring unit. The equipment control unit operates the acquisition device, adjusts the lighting according to the light conditions, reduces paper wrinkles, and ensures high-quality image acquisition; the multi-mode scanning unit executes three scanning modes and merges the data to obtain comprehensive document information; the image quality monitoring unit monitors the image quality in real time, judges based on thresholds and triggers re-sampling to ensure image quality. The image preprocessing module consists of a noise removal unit, an image enhancement unit, and an information restoration unit. The noise removal unit employs appropriate filtering algorithms and adjusts parameters to remove different types of noise. The image enhancement unit uses algorithms to enhance image contrast, facilitating subsequent processing; the information repair unit repairs information loss caused by damage or stains. The feature extraction module consists of a conventional feature extraction unit, a logical structure analysis unit, a text refinement feature unit, and an anti-counterfeiting element feature unit. The standard feature extraction unit is used to extract basic features of text and seals; The logical structure analysis unit is used to analyze the logical structure of the file content to assist in judgment; the text refinement feature unit is used to refine text features and accurately depict text; the anti-counterfeiting element feature unit extracts anti-counterfeiting element features for anti-counterfeiting verification. The information matching and verification module consists of a data comparison unit of the same type, a machine learning model unit, and a similarity calculation unit. The data comparison unit is used to compare similar data and screen for suspicious information such as property area and owner's name; The machine learning model unit uses the model to identify common scanning error types; the similarity calculation unit dynamically calculates the similarity threshold to accurately determine the file matching degree. The results output and report generation module includes a report generation unit, a suspicious information annotation unit, and a report sending and feedback unit. The report generation unit generates verification reports containing comprehensive information and outputs them in multiple formats; the suspicious information annotation unit is used to annotate suspicious areas and record relevant information to assist manual review; the report sending and feedback unit sends reports to manual reviewers, receives and processes feedback, and improves the verification process.

Images