A three-dimensional cadastral data anomaly detection method based on deep learning

Abstract

The application discloses a kind of three-dimensional cadastral data anomaly detection method based on deep learning, specifically related to data processing field, including S1: cadastral unit coding, S2: cadastral survey results generation, S3: ownership analysis, S4: abnormal perception, S5: risk analysis, S6: man-machine interaction.The application realizes the unique identification of each cadastral data by cadastral unit coding, and then carries out abnormal perception on cadastral data, and carries out resource perception and abnormal state monitoring based on cadastral survey results, thereby carrying out risk analysis of cadastral data, forming a closed loop of monitoring, analysis and early warning, thereby optimizing the mode of traditional passive processing problem, avoiding the existence of unable to meet the demand of large-scale three-dimensional cadastral data rapid processing due to the limitation of data processing efficiency, which is beneficial to improving the abnormal detection efficiency while ensuring normal cadastral data collection, and provides a safe and reliable tool for unified management of cadastral data.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a method for detecting anomalies in three-dimensional cadastral data based on deep learning. Background Technology

[0002] With rapid economic development and accelerated urbanization, urban land resources are becoming increasingly scarce, land use is showing a trend of high density and three-dimensionality, and property rights management is becoming increasingly complex. Cadastral data is real estate and natural resource property rights data with spatiotemporal characteristics, and it requires high-precision spatial and attribute data. Its data sources are diverse and complex, and it often faces great difficulties in the process of data acquisition and integration, and data anomalies are prone to occur.

[0003] For example, the existing invention patent publication number CN109376158A discloses a cadastral survey data processing method, the technical solution of which includes: difference analysis, fuzzy analysis, result judgment, attribute association, and office operation; the cadastral survey data processing method of the present invention, through comparison and simulation analysis of observation data in cadastral survey, ensures the reliability of cadastral survey data in office work, and can effectively reduce or eliminate the occurrence of obvious errors in cadastral survey statistics.

[0004] However, in practical use, it still has some shortcomings. First, the existing technology is difficult to meet the high-precision requirements when dealing with complex three-dimensional cadastral scenarios. In particular, when converting from simple two-dimensional cadastral to complex three-dimensional cadastral, it will inevitably expose technical shortcomings and pose a greater risk to the accurate management of three-dimensional property rights. Based on this situation, more professional three-dimensional data acquisition and processing technology is needed. However, most existing processes are limited to the processing of two-dimensional topological relationships, which limits the ability to manage three-dimensional spatial relationships and increases the possibility of property rights disputes.

[0005] Secondly, the current data entry and processing rely too heavily on manual operations. When the data volume is large and the land parcel types are complex, this will inevitably increase the probability of operational errors and pose a significant risk to data accuracy. In this case, a large amount of manpower is needed for verification, but the current operations are mostly limited to manual checks, which limits the efficiency of data processing and poses a risk of not being able to meet the needs of rapid processing of large-scale three-dimensional cadastral data. In addition, over-reliance on manual operations will also prolong the overall processing cycle to some extent, making it impossible to complete data update tasks on time. At the same time, frequent manual operations mean more repetitive work, which will increase labor costs and reduce work efficiency. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide a method for anomaly detection of three-dimensional cadastral data based on deep learning. By encoding cadastral units during the cadastral data collection process to generate cadastral survey results, and then performing resource perception and anomaly monitoring on the cadastral survey results after ownership analysis, a refined and flexible risk analysis is conducted, maximizing the effectiveness of both cadastral data collection and anomaly perception. At the same time, the results are visualized after risk warning, effectively solving the problems raised in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] S1: Cadastral Unit Code: A preliminary survey of cadastral data is conducted to determine the cadastral unit code;

[0009] S2: Generation of cadastral survey results: After the cadastral unit is coded, the cadastral data is subjected to quality inspection and analysis, and a cadastral survey result data package is generated based on the quality-inspected cadastral data, thereby obtaining the cadastral survey results;

[0010] S3: Ownership Analysis: Construct a cadastral database based on the results of the cadastral survey, analyze the ownership status of the target spatial area, and obtain the specified code.

[0011] S4: Anomaly Detection: Extract the cadastral survey results corresponding to the specified code, and use the resource sensing terminal and resource storage terminal to perform resource sensing and anomaly monitoring of the cadastral survey results corresponding to the specified code. The resource sensing objects of the cadastral survey results are location integrity and attribute integrity, and the anomaly monitoring objects are location deviation and data anomaly.

[0012] S5: Risk Analysis: Based on the resource perception and abnormal status monitoring of the cadastral survey results corresponding to the specified codes, conduct cadastral data risk analysis and thus provide risk warnings;

[0013] S6: Human-computer interaction: Visualize cadastral data according to preset display methods and perform statistics based on anomaly detection and risk analysis.

[0014] The technical effects and advantages of this invention are as follows:

[0015] 1. This invention achieves a unique identifier for each cadastral data point through cadastral unit coding, providing a unified benchmark framework for all subsequent stages. This standardized coding avoids data confusion and generates survey result data packages based on the quality verification of data using cadastral unit coding. This enables flexible detection of data anomalies, allowing for timely removal of erroneous data and reducing rework caused by data quality issues in subsequent stages, thus ensuring the accuracy and effectiveness of cadastral survey results from the source. On the other hand, it ensures that cadastral data from different stages and sources are correlated and traceable, improving the standardization of overall data management.

[0016] 2. This invention uses resource sensing terminals and resource storage terminals to detect anomalies and monitors resource perception and abnormal states based on cadastral survey results. This enables risk analysis of cadastral data, which is not limited to manual inspection. On the one hand, it can maximize the satisfaction of anomaly detection needs, and on the other hand, it can avoid the inability to meet the needs of rapid processing of large-scale three-dimensional cadastral data due to limited data processing efficiency. This is beneficial to improve anomaly detection efficiency while ensuring normal cadastral data collection.

[0017] 3. This invention performs risk analysis and early warning based on anomaly perception results, forming a closed loop of monitoring, analysis, and early warning. This optimizes the traditional passive problem-solving model, can identify high-risk points in advance, and reduce losses such as ownership disputes and decision-making errors caused by cadastral data errors. Furthermore, by combining anomaly perception and risk analysis results for statistical analysis, complex cadastral data is transformed into intuitive charts, indicators, and other forms, providing a safe and reliable tool for unified management of cadastral data. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the implementation steps of the method of the present invention.

[0019] Figure 2 This is a flowchart for analyzing the abnormal ownership status of the present invention.

[0020] Figure 3 This is a flowchart of the cadastral data risk analysis process of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0022] As attached Figure 1-3 The present invention discloses a method for detecting anomalies in three-dimensional cadastral data based on deep learning. The specific implementation of this invention includes the following steps:

[0023] S1: Cadastral Unit Code: A preliminary survey of cadastral data is conducted to determine the cadastral unit code.

[0024] In this embodiment, it is necessary to specifically explain that the cadastral unit coding is based on cadastral area, cadastral sub-area, property rights nature, and rights characteristics, and formulates a nationally unique real estate unit code compilation rule. Among them, the cadastral area and cadastral sub-area adopt administrative region division codes to ensure the accuracy of spatial positioning; the property rights nature field clarifies the ownership type of real estate such as land, sea area, and buildings, such as state-owned or collective; the rights characteristics field distinguishes the rights status of real estate, such as transfer, allocation, and lease; by combining these elements according to cadastral area, cadastral sub-area, property rights nature code, rights characteristics code, and sequence code, a real estate unit code with fixed length and uniform format is formed, thereby ensuring that each real estate unit has a unique identifier nationwide.

[0025] It should be explained that cadastral data refers to real estate and natural resource property rights data with spatiotemporal characteristics, including registration of rights, cadastral surveys, and other data related to real estate rights. Among them, registration of rights data includes real estate registration, natural resource registration, and results of dispute resolution; cadastral survey data includes general cadastral surveys, routine cadastral surveys, and natural resource cadastral surveys; other data related to real estate rights include results related to agricultural land conversion approvals, collective land expropriation, land supply, approvals for sea and island use, planning permits, mineral resource management, transfer of collective construction land, approval of homesteads, management of land contract management contracts, management of forest rights contracts, and real estate transactions.

[0026] It should be added that by using, pre-compiling, or compiling real estate unit codes in different business processes, a single code can be used to link each business process. For example, the function of pre-compiling real estate unit codes is provided in the land use pre-approval and planning site selection stage, the land use approval stage, and the sea use pre-approval stage; real estate unit codes can be compiled in the land supply stage and the land use approval stage, and the real estate unit codes can be pushed back to the pre-approval site selection, land use approval, and other business processes; the function of pre-compiling integrated real estate unit codes is provided in the construction project planning permit, construction project construction stage, and rural construction planning permit stage; and the function of compiling integrated real estate unit codes is provided in the completion and acceptance stage.

[0027] It should be further explained that the natural resource corresponding cadastral unit code is compiled according to the natural resource registration unit code compilation rules, and the natural resource registration unit code compilation rules are associated and mapped with the real estate unit code compilation rules.

[0028] S2: Generation of cadastral survey results: After the cadastral unit is coded, the cadastral data is subjected to quality inspection and analysis, and a cadastral survey result data package is generated based on the quality-inspected cadastral data, thereby obtaining the cadastral survey results.

[0029] In this embodiment, it should be specifically explained that the quality inspection analysis of cadastral data includes result integrity detection, spatial topology detection, and attribute correctness detection. Result integrity is achieved by detecting whether there are any missing necessary elements in each cadastral unit. If so, the number of missing necessary elements is counted and compared with the total number of necessary elements, and the percentage is taken to calculate the result missing rate.

[0030] It should be added that the detection of missing essential elements can be done by iterating through the field values ​​of cadastral data. If a field is empty or contains only placeholders, it is considered missing. By calculating the missing rate of the results through the detection of missing essential elements, the integrity of the results can be checked. By detecting the missing essential elements, the breaks in the property ownership certificate chain can be identified in advance, prompting supplementation and improvement, reducing the risk of legal disputes from the data level, and ensuring the legality of real estate transactions, mortgages, inheritance and other behaviors.

[0031] Spatial topology includes two-dimensional topology and three-dimensional topology. Two-dimensional topology detects the overlapping area of ​​adjacent land parcel boundaries of various cadastral units. If adjacent land parcel boundaries overlap, the overlapping area is obtained, and the percentage is calculated by comparing the overlapping area with the benchmark area. The area overlap rate is calculated by this percentage. Three-dimensional topology detects the overlapping volume of spatial entities with different ownership. If overlapping volume exists, the overlapping volume is obtained, and the percentage is calculated by comparing the overlapping volume with the benchmark volume. The area overlap rate and the volume overlap rate are then weighted and averaged to obtain the spatial overlap rate.

[0032] It should be explained that the benchmark area is the area of ​​the smaller of the two adjacent land parcels. For example, if parcel A has an area of ​​1000㎡ and parcel B has an area of ​​800㎡, the overlap area between the two parcels is 40㎡. Then the area overlap rate = 40 / 800 × 100% = 5%. Choosing the smaller parcel area as the benchmark can avoid the overlap rate being underestimated due to the large difference in parcel area. When a large parcel overlaps with a small parcel, using the area of ​​the large parcel as the benchmark will make the area overlap rate value smaller, thus masking the actual risk of ownership conflict. Similarly, the benchmark volume is the volume of the smaller entity among the spatial entities with different ownership.

[0033] It should be added that, specifically, the two-dimensional topology can be based on the R-tree spatial index to retrieve the neighboring parcels of the target parcel, perform intersection operations on the boundary polygons of the adjacent parcels, and extract the area of ​​the intersection region as the overlapping area; specifically, the three-dimensional topology can be based on the three-dimensional mesh model transformed from the spatial entities of different ownership, use the octree spatial index to retrieve whether the entities of different ownership intersect, extract features through the 3D-CNN model, identify the overlapping voxels of multiple ownership entities, and accumulate the identified overlapping voxels to obtain the overlapping volume.

[0034] Attribute correctness refers to all non-spatial attribute information in cadastral data, including but not limited to right holder information, right nature, right type, and right term. Similarity is obtained by crawling text information and comparing it with the corresponding attribute fields of each cadastral unit. The attribute error rate is obtained by subtracting the value 1 from the similarity.

[0035] It should be further explained that the cadastral survey results are obtained as follows: Based on the results integrity detection, spatial topology detection, and attribute correctness detection, the results missing rate, spatial overlap rate, and attribute error rate corresponding to each cadastral unit are extracted. The cadastral data quality coefficient is then evaluated, specifically expressed as: Qo=1-0.4×Vo+0.3×So+0.3×Ro, where Vo, So, and Ro represent the results missing rate, spatial overlap rate, and attribute error rate corresponding to each cadastral unit, respectively, and Qo represents the cadastral data quality coefficient;

[0036] A quality threshold is set, and the cadastral data quality coefficient is compared with the quality threshold. If the cadastral data quality coefficient is greater than or equal to the quality threshold, it means that the cadastral data quality inspection is qualified. Then, the cadastral data is edited according to the local cadastral data and the quality inspection results to obtain the cadastral survey results. If the cadastral data quality coefficient is less than the quality threshold, it means that the cadastral data quality inspection is unqualified. Then, the code of the cadastral unit corresponding to the unqualified cadastral data is output, and the cadastral data is re-investigated and inspected based on the cadastral unit code until the cadastral data quality coefficient meets the quality threshold. The quality threshold is set according to the relevant requirements of natural resource management.

[0037] It should be added that editing cadastral data includes providing cadastral data format conversion, projection transformation, data editing, and map attribute linking.

[0038] S3: Ownership Analysis: Construct a cadastral database based on the results of the cadastral survey, analyze the ownership status of the target spatial area, and obtain the specified code.

[0039] It should be added that a cadastral survey result data package is generated before obtaining the cadastral survey results. This cadastral survey result data package is used for registration. After the registration is completed, information such as the type of right, right holder, right restriction information, registration agency, registration time, certificate number, and registerer is packaged and imported into the cadastral database, with the real estate unit code as the association. Information such as the registration status, registration agency, registration time, and registerer is packaged and imported into the cadastral database, with the natural resource registration unit number as the association.

[0040] In this embodiment, the specific process of analyzing abnormal ownership status needs to be explained as follows:

[0041] A1: Extract all cadastral data within the target spatial range from the cadastral database;

[0042] A2: Overlay the land parcel vector map layer within the target space with the real estate unit code spatial distribution layer, which includes the land use approval layer, the planned use layer, and the three-dimensional space layer, thereby identifying whether the ownership is abnormal.

[0043] A3: If the ownership is identified as abnormal, extract its corresponding code and set it as the specified code.

[0044] It should be further explained that the land use approval layer is used to compare the spatial overlap between the approval scope and the target spatial scope. If the approval boundary exceeds the registered land parcel scope, it is marked as an ownership anomaly. The planning use layer overlay is used to overlay the land parcel ownership scope with the permitted construction area and restricted construction area layers in the national land space plan. If the ownership scope is located in the prohibited construction area, it is marked as an ownership anomaly. The three-dimensional space layer overlay is used to overlay the three-dimensional ownership space such as underground parking garages and sky bridges with the above-ground buildings, underground pipelines, and other layers. If vertical overlap or intersection of three-dimensional space ownership is identified, it is marked as an ownership anomaly.

[0045] S4: Anomaly Detection: Extract the cadastral survey results corresponding to the specified codes, and use resource sensing terminals and resource storage terminals to perform resource sensing and anomaly monitoring of the cadastral survey results corresponding to the specified codes. The resource sensing objects of the cadastral survey results are location integrity and attribute integrity, and the anomaly monitoring objects are location deviation and data anomaly.

[0046] It should be added that the resource sensing terminal is used to perform resource sensing of cadastral survey results corresponding to a specified code. Specifically, it can be a combination of edge computing devices and GPS positioning devices. The edge computing devices are industrial computers and smart terminals deployed at the cadastral survey site, which can process the real-time collected location and attribute data nearby and reduce data transmission delay. The GPS positioning devices can locate the sensing points and perform location sensing.

[0047] Resource storage terminals are used to store cadastral data. Specifically, they can be distributed storage systems capable of storing massive amounts of unstructured or semi-structured cadastral survey results, ensuring data integrity and durability. They also optimize data query efficiency through indexing technology, enabling computing terminals to quickly extract cadastral data corresponding to specified codes, providing data support for resource perception and anomaly monitoring.

[0048] In this embodiment, it is necessary to specifically explain the location integrity process as follows:

[0049] Based on the expropriation approval time of the cadastral survey results corresponding to the designated codes, an approval time series is constructed, and sensing points are deployed in the target area according to the cadastral data type. The approval time series is set as T = {t1, t2, ..., t...} m}

[0050] It is important to understand that the target area is the region corresponding to the specified code. Deploying sensing points can provide accurate geographic coordinates, ensuring that all cadastral data are based on a unified benchmark. At the same time, the deployment of sensing points should follow these principles: sensing points should be evenly distributed within the target area to ensure global measurement accuracy; sensing points should be deployed in locations with strong physical stability, not easily disturbed or damaged, to ensure long-term effective provision of coordinate benchmarks, facilitating later verification, updates, and maintenance; sensing points should establish spatial associations with key cadastral elements within the target area, such as boundary points, ownership boundary lines, and plot center points, to facilitate coordinate calibration and spatial verification of cadastral data.

[0051] The specific deployment process is as follows: When the cadastral data type of the target area is determined to be real estate data, a regular grid layout is used to set up sensing points. For example, assuming that the area corresponding to the specified code is a residence, a sensing point can be set at the outer corner and the center of the residence, thus forming a grid layout. When the cadastral data type of the target area is determined to be natural resource property rights data, the resource boundary line of the target area is extracted, and then sensing points are set up on the boundary line. For resources with easily changeable boundaries, additional sensing points need to be added to monitor the boundary offset caused by natural factors.

[0052] Based on the approval time series, the location coordinates of each sensing point are extracted using resource sensing terminals, and the coordinate deviation value of the target area is calculated using Euclidean distance, specifically expressed as follows:

[0053]

[0054] Where Lc represents the coordinate deviation value, (x i，k y i，k (x) represents the coordinates of the i-th sensing point in the k-th approval time series. i，1 y i，1 ) represents the coordinates of the sensing point corresponding to the earliest approval time in the approval time series, k = 1, 2, ..., m, i = 1, 2, ..., n. At this time, the coordinates of the sensing point of the earliest approval time t1 in the time series are selected as the benchmark.

[0055] It should be added that the coordinates of each sensing point are based on the nationally unified coordinate system to eliminate deviations caused by inconsistent standards.

[0056] The time weighting factor is calculated based on the approval time series, and is specifically expressed as follows:

[0057]

[0058] Where w represents the time weighting factor, t j t1 and t1 represent the j-th approval time series and the earliest approval time series, respectively.

[0059] It should be explained that, since newer approval times are closer to the current state, they have a greater impact on the deviation; therefore, a time weighting factor is introduced. This indicates that j iterates from 1 to m and is used to calculate the sum of the differences between all approval times and the earliest time. By dividing the time difference between the j-th approval and the earliest approval by the sum of the time differences between all approvals and the earliest approval, the weight of the current approval time in the deviation calculation is obtained. This makes the approvals closer to the current time have a higher weight in the deviation analysis, which is more in line with the actual impact of data changes.

[0060] It should be further noted that the details regarding attribute integrity are as follows:

[0061] Data missing information for various approvals is extracted using resource storage terminals, including approvals for conversion of agricultural land, land supply, approvals for use of sea and islands, planning permits, approvals for homesteads, and real estate transactions, and the anomaly rate is calculated accordingly.

[0062] It should be added that the anomaly rate is obtained by comparing the data of each approval with the total data volume. For example, the number of agricultural land conversion approval records exceeding the approval authority of the current level is extracted and compared with the total number of conversion approvals to calculate the authority over-limit rate; the number of land supply records with illegal land supply methods is extracted and compared with the total number of land supply records to obtain the land supply method error rate; the number of sea and island use approval records with approved areas exceeding the functional zoning limit is extracted and compared with the total number of sea and island use approvals to calculate the approved area over-limit rate; the number of planning permit records with conflicts between permitted uses and land ownership is extracted and compared with the total number of planning permits to calculate the ownership conflict rate; the number of homestead approval records with approved areas exceeding the local standard is extracted and compared with the total number of approved homesteads to obtain the area over-standard rate; and the number of real estate transactions is extracted, the single transaction price and the average price of the same type in the region are obtained, and the difference between the single transaction price and the average price of the same type in the region is compared with the average price of the same type in the region to obtain the price deviation.

[0063] An anomaly threshold is set based on the anomaly rate. The anomaly rate is compared with the anomaly threshold. If the anomaly rate is greater than the anomaly threshold, it is treated as an anomaly record. In this way, the total number of anomaly records of all types is counted.

[0064] S5: Risk Analysis: Based on the resource perception and abnormal status monitoring of the cadastral survey results corresponding to the specified codes, conduct cadastral data risk analysis and issue risk warnings.

[0065] In this embodiment, the cadastral data risk analysis needs to be explained in detail as follows:

[0066] The position deviation coefficient is obtained by multiplying the coordinate deviation value by the time weighting factor and then comparing it with the number of sensing points.

[0067] The data error coefficient is obtained by comparing the total number of abnormal records with the total number of records in the cadastral survey results. The more abnormal records there are, the greater the data error coefficient will be.

[0068] The cadastral anomaly index is calculated by extracting the location deviation coefficient and the data error coefficient from the cadastral survey results corresponding to each specified code. Specifically, it is expressed as follows:

[0069]

[0070] Where DF represents the cadastral anomaly index, and Cm and Tc represent the location deviation coefficient and data error coefficient, respectively. The more severe the location deviation and the higher the degree of data error, the higher the degree of anomaly in the cadastral data, and the larger the cadastral anomaly index.

[0071] It should be further explained that a risk warning judgment needs to be made before the risk warning is implemented. The specific operation is as follows: compare the cadastral anomaly index with the set risk threshold. If the cadastral anomaly index is greater than the set risk threshold, it is determined that a risk warning needs to be made; otherwise, it is determined that a risk warning does not need to be made.

[0072] Furthermore, the specific risk warnings are as follows:

[0073] Extract the location deviation coefficient and data error coefficient corresponding to the risk threshold where the cadastral anomaly index is greater than the set risk threshold, and set the corresponding anomaly preset values ​​for each;

[0074] When the location deviation coefficient does not reach the corresponding abnormal preset value but the data error coefficient does, it means that there is a high risk in the data entry process. This indicates that there are many errors in the attribute data during the entry, conversion, storage or transmission process, such as land ownership information, area value, land use code, registration date, etc. In this case, add automatic data verification rules and manual review to improve the verification accuracy.

[0075] When the location deviation coefficient reaches the corresponding preset value for anomalies but the data error coefficient does not reach the corresponding preset value for anomalies, it means that there is a high risk in the location, indicating that there is chaos in spatial management. At this time, a pop-up prompt will be made through the data management platform, and the land management personnel will be notified to clarify the risk and the number of the plot involved and its location. At the same time, the plot with location deviation will be highlighted with a special color on the electronic map and the deviation amount will be marked to intuitively show the scope of the problem.

[0076] S6: Human-computer interaction: Visualize cadastral data according to preset display methods and perform statistics based on anomaly detection and risk analysis.

[0077] In this embodiment, it should be specifically explained that the preset display methods include data display, chart display, and image display. Depending on the display method, users can query graphic and attribute information such as land parcels and houses in a map scene, and intuitively display it in chart form, achieving seamless and efficient browsing of cadastral data. The statistics are as follows: A leadership dashboard is established, and based on region, real estate type, time scale, etc., spatiotemporal big data analysis and mining technology is used to statistically analyze the distribution, type, transaction, price, historical status, and activity of various real estate units in real time, forming statistical charts.

[0078] It should be added that spatiotemporal big data analysis and mining technology refers to a technical system for in-depth processing, pattern extraction, and value mining of data that simultaneously contains time and space dimensions. By integrating spatiotemporal relationships, it obtains the dynamic patterns, potential correlations, and development trends behind the data.

[0079] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0080] 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 anomaly detection in three-dimensional cadastral data based on deep learning, characterized in that, include: S1: Cadastral Unit Code: A preliminary survey of cadastral data is conducted to determine the cadastral unit code; S2: Generation of cadastral survey results: After the cadastral unit is coded, the cadastral data is subjected to quality inspection and analysis, and a cadastral survey result data package is generated based on the quality-inspected cadastral data, thereby obtaining the cadastral survey results; The quality inspection analysis includes result integrity detection, spatial topology detection, and attribute correctness detection. Result integrity is determined by detecting whether there are any missing necessary elements in each cadastral unit. If so, the number of missing necessary elements is counted and compared with the total number of necessary elements to obtain a percentage, thereby calculating the result missing rate. Attribute correctness refers to all non-spatial attribute information in cadastral data. The similarity is obtained by crawling text information and comparing it with the corresponding attribute fields of each cadastral unit. The attribute error rate is obtained by subtracting the value 1 from the similarity. The cadastral survey results are obtained as follows: Based on the results integrity detection, spatial topology detection, and attribute correctness detection, the results missing rate, spatial overlap rate, and attribute error rate corresponding to each cadastral unit are extracted. The cadastral data quality coefficient is then evaluated, specifically expressed as: Qo=1-0.4×Vo+0.3×So+0.3×Ro, where Vo, So, and Ro represent the results missing rate, spatial overlap rate, and attribute error rate corresponding to each cadastral unit, respectively, and Qo represents the cadastral data quality coefficient; Set a quality threshold and compare the cadastral data quality coefficient with the quality threshold. If the cadastral data quality coefficient is greater than or equal to the quality threshold, the cadastral data is edited according to the local cadastral data and quality inspection results to obtain the cadastral survey results. If the cadastral data quality coefficient is less than the quality threshold, the code of the cadastral unit corresponding to the unqualified cadastral data is output, and the cadastral data is re-investigated and quality inspected based on the cadastral unit code until the cadastral data quality coefficient meets the quality threshold. S3: Ownership Analysis: Construct a cadastral database based on the results of the cadastral survey, analyze the ownership status of the target spatial area, and obtain the specified code. S4: Anomaly Detection: Extract the cadastral survey results corresponding to the specified code, and use the resource sensing terminal and resource storage terminal to perform resource sensing and anomaly monitoring of the cadastral survey results corresponding to the specified code. The resource sensing objects of the cadastral survey results are location integrity and attribute integrity, and the anomaly monitoring objects are location deviation and data anomaly. The location integrity is described in the following process: Based on the expropriation approval time of the cadastral survey results corresponding to the designated codes, an approval time series is constructed, and sensing points are deployed in the target area according to the cadastral data type. The approval time series is set as T={t1, t2, ..., t m }; Based on the approval time series, the location coordinates of each sensing point are extracted using resource sensing terminals, and the coordinate deviation value of the target area is calculated using Euclidean distance. The time weighting factor is calculated based on the approval time series, and is specifically expressed as follows: ， Where w represents the time weighting factor, t j t1 and t1 represent the j-th approval time series and the earliest approval time series, respectively; S5: Risk Analysis: Based on the resource perception and abnormal status monitoring of the cadastral survey results corresponding to the specified codes, conduct cadastral data risk analysis and thus provide risk warnings; The risk analysis of the cadastral data is as follows: The position deviation coefficient is obtained by multiplying the coordinate deviation value by the time weighting factor and then comparing it with the number of sensing points. The data error coefficient is obtained by comparing the total number of abnormal records with the total number of records in the cadastral survey results. The cadastral anomaly index is calculated by extracting the location deviation coefficient and the data error coefficient from the cadastral survey results corresponding to each specified code. Specifically, it is expressed as follows: ， Where DF represents the cadastral anomaly index, and Cm and Tc represent the location deviation coefficient and data error coefficient, respectively; The cadastral anomaly index is compared with the set risk threshold. If the cadastral anomaly index is greater than the set risk threshold, it is determined that a risk warning needs to be issued; otherwise, it is determined that a risk warning does not need to be issued. S6: Human-computer interaction: Visualize cadastral data according to preset display methods and perform statistics based on anomaly detection and risk analysis.

2. The method for anomaly detection of three-dimensional cadastral data based on deep learning according to claim 1, characterized in that: The spatial topology includes two-dimensional topology and three-dimensional topology. The two-dimensional topology detects the overlapping area of ​​adjacent land parcel boundaries of each cadastral unit. If there is an overlap, the overlapping area is obtained and the percentage is calculated by comparing the overlapping area with the benchmark area. The three-dimensional topology detects the overlapping volume of spatial entities with different ownership. If there is an overlap, the overlapping volume is obtained and the percentage is calculated by comparing the overlapping volume with the benchmark volume. The spatial overlap rate is then obtained by weighted averaging the area overlap rate and the volume overlap rate.

3. The method for anomaly detection of three-dimensional cadastral data based on deep learning according to claim 1, characterized in that: The specific process for analyzing the anomalies in ownership status is as follows: A1: Extract all cadastral data within the target spatial range from the cadastral database; A2: Overlay the land parcel vector map layer within the target space with the real estate unit code spatial distribution layer, which includes the land use approval layer, the planned use layer, and the three-dimensional space layer, thereby identifying whether the ownership is abnormal. A3: If the ownership is identified as abnormal, extract its corresponding code and set it as the specified code.

4. The method for anomaly detection of three-dimensional cadastral data based on deep learning according to claim 1, characterized in that: The specific deployment process of the sensing points is as follows: when the cadastral data type of the target area is determined to be real estate data, a regular grid layout is used to set the sensing points; when the cadastral data type of the target area is determined to be natural resource property rights data, the resource boundary line of the target area is extracted, and then sensing points are deployed on the boundary line. For resources with easily changing boundaries, additional sensing points need to be added to monitor the boundary offset caused by natural factors.

5. The method for anomaly detection of three-dimensional cadastral data based on deep learning according to claim 1, characterized in that: For details regarding the integrity of the attributes, please refer to the following: The system utilizes resource storage terminals to extract data missing information for various approvals, including approvals for conversion of agricultural land, land supply, approvals for use of sea and islands, planning permits, approvals for homesteads, and real estate transactions, thereby calculating the anomaly rate. An anomaly threshold is set based on the anomaly rate. The anomaly rate is compared with the anomaly threshold. If the anomaly rate is greater than the anomaly threshold, it is treated as an anomaly record. In this way, the total number of anomaly records of all types is counted.

Images