Real estate registration verification method and system based on artificial intelligence

Abstract

The invention provides a real estate registration verification method and system based on artificial intelligence, and relates to the technical field of information verification, and the method comprises the steps: determining the information of an owner, the reference face image data of the owner, and the reference signature image data of the owner; obtaining comparison face image data and comparison signature image data of seller personnel; obtaining first facial feature information, second facial feature information, first signature feature information, second signature feature information and certificate feature information; and obtaining the first relation information and the second relation information, determining property right reference feature information and property right verification feature information through a graph neural network model, and determining a verification result. According to the method and the device, whether the relationship among multiple owners changes can be verified through the graph neural network model, and if the authenticity of sale is doubted or the relationship among the multiple owners changes, the risk of the transaction can be prompted, so that the legal risk of the transaction can be reduced, and the security of the transaction can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information verification, and particularly to a method and system for verifying real estate registration based on artificial intelligence. Background Art

[0002] In the related art, before the property rights transaction of a house, the identities of both the buyer and the seller can be verified, and the information of the house itself can also be verified. However, for a house with multiple property owners, the relationship between the owners may be relatively complex, and there may be certain legal risks in the transaction of this house. The related art lacks a verification scheme for the relationship between the owners. Summary of the Invention

[0003] The present invention provides a method and system for verifying real estate registration based on artificial intelligence, which can solve the technical problem that it is difficult to verify the relationship between property owners in the related art, and thus it is difficult to reduce transaction risks.

[0004] According to a first aspect of the present invention, there is provided a method for verifying real estate registration based on artificial intelligence, including: Obtain the property information certificate of the house to be transferred, and determine the owner information of the house according to the property information certificate; Through the background database, obtain the reference facial image data and reference signature image data of the owner corresponding to the owner information; Obtain the comparison facial image data and comparison signature image data of the seller; Through the trained face recognition model, obtain the first facial feature information of the reference facial image data and the second facial feature information of the comparison facial image data; Through the trained signature recognition model, obtain the first signature feature information of the reference signature image data and the second signature feature information of the comparison signature image data; Through the trained document recognition model, obtain the document feature information of the property information certificate; Obtain the first relationship information between multiple owners and the second relationship information between multiple sellers; According to the document feature information, the first relationship information, the first facial feature information, the first signature feature information and the trained graph neural network model, obtain the property reference feature information, and according to the document feature information, the second relationship information, the second facial feature information, the second signature feature information and the trained graph neural network model, obtain the property verification feature information; Obtain the verification result according to the property reference feature information and the property verification feature information.

[0005] According to the present invention, property right reference feature information is obtained based on certificate feature information, first relationship information, first facial feature information, first signature feature information, and a trained graph neural network model, and property right verification feature information is obtained based on certificate feature information, second relationship information, second facial feature information, second signature feature information, and the trained graph neural network model, including: Process the first relationship information through the trained relationship encoding model to obtain a first relationship feature, and input the first relationship feature into a first fully connected layer to obtain a first relationship weight; Process the second relationship information through the trained relationship encoding model to obtain a second relationship feature, and input the second relationship feature into the first fully connected layer to obtain a second relationship weight; Based on the first relationship weight, the property right ratio of the owner, the subordination relationship between the reference signature image data and the owner, the certificate feature information, the first facial feature information, and the first signature feature information, obtain a first graph structure; Based on the second relationship weight, the property right ratio of the seller, the subordination relationship between the comparison signature image data and the seller, the certificate feature information, the second facial feature information, and the second signature feature information, obtain a second graph structure; Process the first graph structure through the trained graph neural network model to obtain property right reference feature information; Process the second graph structure through the trained graph neural network model to obtain property right verification feature information.

[0006] According to the present invention, based on the first relationship weight, the property right ratio of the owner, the subordination relationship between the reference signature image data and the owner, the certificate feature information, the first facial feature information, and the first signature feature information, obtaining a first graph structure includes: Set the certificate feature information, the first facial feature information, and the first signature feature information as the first node information of the nodes of the first graph structure; Set the property right ratio of the owner as the connection weight between the node corresponding to the certificate feature information and the node corresponding to the first facial feature information of the owner in the first graph structure; Set the first relationship weight between different owners as the connection weight between the nodes corresponding to the first facial feature information of different owners; Determine the connection weight between the node corresponding to the first signature feature information and the node corresponding to the first facial feature information of the owner according to the subordination relationship between the reference signature image data and the owner; Obtain a first adjacency matrix according to the connection weight; Obtain a first degree matrix according to the first adjacency matrix; Obtain the first graph structure according to the first node information, the first adjacency matrix, and the first degree matrix.

[0007] According to the present invention, the method further includes: Obtain a sample property information certificate, the first facial image sample of the sample owner corresponding to the sample property information certificate, the first signature image sample, and the first sample relationship information between each sample owner; Through a face recognition model, obtain the first sample facial feature information of the first facial image sample; Through a signature recognition model, obtain the first sample signature feature information of the first signature image sample; Through a certificate recognition model, obtain the sample certificate feature information of the sample property information certificate; Through a relationship coding model, obtain the first sample relationship weight corresponding to the first sample relationship information; According to the sample certificate feature information, the first sample facial feature information, the first sample signature feature information, the property ratio of the sample owner, obtain the first sample graph structure; Perform at least one of the following processes: use other facial image samples of the same sample owner to replace the first facial image sample and obtain the second sample facial feature information, use other signature image samples of the same sample owner to replace the first signature image sample and obtain the second sample signature feature information; Based on at least one of the second sample facial feature information and the second sample signature feature information, obtain the second sample graph structure; Perform at least one of the following processes: use facial image samples of different persons to replace the first facial image sample and obtain the third sample facial feature information, use signature image samples obtained by different persons signing the name of the sample owner to replace the first signature image sample and obtain the third sample signature feature information, randomly modify the first sample relationship information between sample owners and obtain the modified second sample relationship weight; Based on at least one of the third sample facial feature information, the third sample signature feature information, and the second sample relationship weight, obtain the third sample graph structure; Process the first sample graph structure, the second sample graph structure, and the third sample graph structure respectively through a graph neural network model, and obtain the first property sample feature information corresponding to the first sample graph structure, the second property sample feature information corresponding to the second sample graph structure, and the third property sample feature information corresponding to the third sample graph structure; According to the first sample signature feature information, the second sample signature feature information, and the third sample signature feature information, obtain a signature loss function; According to the first sample relationship weight and the second sample relationship weight, determine a relationship weight attention coefficient; Obtain a global loss function based on the first property sample feature information, the second property sample feature information, the third property sample feature information, and the relationship weight attention coefficient; Train a face recognition model, a signature recognition model, a document recognition model, a relationship encoding model, and a graph neural network model based on the global loss function and the signature loss function.

[0008] According to the present invention, obtaining a signature loss function according to the first sample signature feature information, the second sample signature feature information, and the third sample signature feature information includes: Decode the first sample signature feature information to obtain a first word vector of the text corresponding to the first signature feature information; Decode the second sample signature feature information to obtain a second word vector of the text corresponding to the second signature feature information; Decode the third sample signature feature information to obtain a third word vector of the text corresponding to the third signature feature information; Concatenate the first sample signature feature information and the second sample signature feature information to obtain a first concatenated vector; Concatenate the first sample signature feature information and the third sample signature feature information to obtain a second concatenated vector; Input the first concatenated vector into a second fully connected layer to obtain first matching probability information; Input the second concatenated vector into a second fully connected layer to obtain second matching probability information; Obtain a signature loss function according to the first word vector, the second word vector, the third word vector, the first matching probability information, and the second matching probability information.

[0009] According to the present invention, obtaining a signature loss function according to the first word vector, the second word vector, the third word vector, the first matching probability information, and the second matching probability information includes: According to the formula

[0010] Obtain the signature loss function , where is the first word vector, is the second word vector, is the third word vector, sim is a similarity function, is the first matching probability information, is the second matching probability information, is a preset parameter.

[0011] According to the present invention, determining a relationship weight attention coefficient according to the first sample relationship weight and the second sample relationship weight includes: Determine the average value of the first sample relationship weights corresponding to the unmodified first sample relationship information and the minimum value of the relationship weight change corresponding to the modified first sample relationship information according to the first sample relationship weights and the second sample relationship weights. Determine the ratio of the average value of the first sample relationship weights corresponding to the unmodified first sample relationship information to the minimum value of the relationship weight change corresponding to the modified first sample relationship information as the relationship weight attention coefficient.

[0012] According to the present invention, a global loss function is obtained based on the first property sample feature information, the second property sample feature information, the third property sample feature information, and the relationship weight attention coefficient, including: According to the formula

[0013] Obtain the global loss function , where is the first property sample feature information, is the second property sample feature information, is the third property sample feature information, is the relationship weight attention coefficient, sim is the similarity function, is a preset parameter.

[0014] According to the second aspect of the present invention, a real estate registration verification system based on artificial intelligence is provided, including: A first acquisition module, configured to acquire the property information certificate of the property to be transferred, and determine the owner information of the property according to the property information certificate; A second acquisition module, configured to acquire the reference facial image data and the reference signature image data of the owner corresponding to the owner information through the background database; A third acquisition module, configured to acquire the comparison facial image data and the comparison signature image data of the seller; A face recognition module, configured to acquire the first facial feature information of the reference facial image data and the second facial feature information of the comparison facial image data through the trained face recognition model; A signature recognition module, configured to acquire the first signature feature information of the reference signature image data and the second signature feature information of the comparison signature image data through the trained signature recognition model; A certificate recognition module, configured to acquire the certificate feature information of the property information certificate through the trained certificate recognition model; A relationship module, configured to acquire the first relationship information between multiple owners and the second relationship information between multiple sellers; A verification module, configured to obtain property right reference feature information according to certificate feature information, first relationship information, first facial feature information, first signature feature information and a trained graph neural network model, and obtain property right verification feature information according to certificate feature information, second relationship information, second facial feature information, second signature feature information and the trained graph neural network model; An inspection result module, configured to obtain an inspection result according to the property right reference feature information and the property right verification feature information.

[0015] By adopting the above technical solutions, the present invention can achieve the following technical effects: According to the present invention, not only can the facial image data and signature image of the seller be verified to ensure the authenticity of the sale, but also whether the relationships of multiple owners have changed can be verified through the graph neural network model. If the authenticity of the sale is in doubt or the relationships of multiple owners have changed, it can be prompted that there are risks in the transaction, thereby reducing the legal risks of the transaction and enhancing the security of the transaction. Not only can the facial images and signatures of the seller personnel be verified, but also the relationships between the seller personnel can be verified. When verifying, the relationship weights of the relationships between the seller personnel can be calculated, the subordination relationship between the signature and the owner, as well as the property right ratio of the owner can be determined, so as to objectively determine the adjacency matrix, and property right reference feature information that can comprehensively and accurately describe the property and its property right owners and property right verification feature information that can comprehensively describe the property and the total features of its seller personnel can be obtained through the graph neural network model, and the property right reference feature information and the property right verification feature information are compared. Therefore, in the comparison process, various information such as facial images, signatures, and relationship information can be comprehensively compared to improve the accuracy of the inspection result and help reduce transaction risks. When determining the signature loss function, the recognition ability of the model for the text of the signature can be improved through the construction of the signature loss function, and the model can still accurately recognize the signatures of the same person when there are slight errors in the signatures of the same person at different times, improving the robustness of the model, and being able to distinguish the same names signed by different people, reducing the possibility of forged signatures. When determining the global loss function, the recognition ability of the model for facial images taken at different times or signatures signed at different times of the same owner can be improved through the construction of the global loss function, improving the robustness of the model, and the recognition ability of the model for the situation of different facial images or forged signatures, as well as the recognition ability for risks caused by relationship changes, which is beneficial to reducing transaction risks. Description of the Drawings

[0016] Figure 1 Exemplarily shows a schematic flowchart of a property registration inspection method based on artificial intelligence according to an embodiment of the present invention; Figure 2 Exemplarily shows a schematic diagram of obtaining an inspection result according to an embodiment of the present invention; Figure 3 Exemplarily shown is a block diagram of an artificial intelligence-based real estate registration verification system according to an embodiment of the present invention. Detailed implementation manners

[0017] The technical solution of the present invention will be described in detail below with specific embodiments. These several specific embodiments can be combined with each other, and for the same or similar concepts or processes, they may not be repeated in some embodiments.

[0018] Figure 1 Exemplarily shown is a schematic flowchart of an artificial intelligence-based real estate registration verification method according to an embodiment of the present invention, and the method includes: Step S1: Obtain the property right information certificate of the property to be transferred, and determine the owner information of the property according to the property right information certificate; Step S2: Through the background database, obtain the reference facial image data and reference signature image data of the owner corresponding to the owner information; Step S3: Obtain the comparison facial image data and comparison signature image data of the seller; Step S4: Through the trained face recognition model, obtain the first facial feature information of the reference facial image data and the second facial feature information of the comparison facial image data; Step S5: Through the trained signature recognition model, obtain the first signature feature information of the reference signature image data and the second signature feature information of the comparison signature image data; Step S6: Through the trained certificate recognition model, obtain the certificate feature information of the property right information certificate; Step S7: Obtain the first relationship information between multiple owners and the second relationship information between multiple sellers; Step S8: According to the certificate feature information, the first relationship information, the first facial feature information, the first signature feature information, and the trained graph neural network model, obtain the property right reference feature information, and according to the certificate feature information, the second relationship information, the second facial feature information, the second signature feature information, and the trained graph neural network model, obtain the property right verification feature information; Step S9: Obtain the verification result according to the property right reference feature information and the property right verification feature information.

[0019] The artificial intelligence-based real estate registration verification method according to the embodiment of the present invention can not only verify the facial image data and signature image of the seller to ensure the authenticity of the sale, but also verify whether the relationship between multiple owners has changed through the graph neural network model. If the authenticity of the sale is in doubt or the relationship between multiple owners has changed, it can prompt that there is a risk in the transaction, thereby reducing the legal risk of the transaction and improving the transaction security.

[0020] According to an embodiment of the present invention, in step S1, the property right information certificate (for example, the house ownership certificate, the real estate certificate or the real estate property right certificate) may record the information of the property owner, such as the name of the owner, the property right ratio, etc. These information can be read by means of OCR (Optical Character Recognition) and the like for subsequent processing, and the identification code of the property right information certificate can also be obtained.

[0021] According to an embodiment of the present invention, in step S2, the background database can be queried through the information read from the property right information certificate above to obtain the reference facial image data and the reference signature image data of the owner corresponding to each owner information. When registering the property right, a reference facial image data can be taken for each owner, and the signature of each owner can be obtained, and the reference signature image data can be taken for subsequent verification when the property right is changed.

[0022] According to an embodiment of the present invention, in step S3, when the owner intends to sell the property, the owner is the seller. Usually, all the owners recorded in the property right information certificate need to agree to the sale before the sale can be carried out. Therefore, the seller needs to include all the owners. Therefore, the seller can be verified. The comparison facial image data of the seller can be taken, and the signature of the seller can be obtained to obtain the comparison signature image data, so as to verify the seller. It is not only necessary to verify whether the seller is complete (that is, whether all the owners are included), but also necessary to verify whether the comparison facial image data and the comparison signature image data match the reference facial image data and the reference signature image data respectively.

[0023] According to an embodiment of the present invention, in step S4, the face recognition model is a convolutional neural network model, for example, models such as VGGNet, ResNet, etc. After training, it can be used to obtain the first facial feature information of the reference facial image data and the second facial feature information of the comparison facial image data. In the example, both the first facial feature information and the second facial feature information are information in vector form. Similarly, in step S5, the signature recognition model is also a convolutional neural network model, for example, models such as VGGNet, ResNet, etc. After training, it can be used to obtain the first signature feature information of the reference signature image data and the second signature feature information of the comparison signature image data. Similarly, in step S6, the document recognition model is also a convolutional neural network model, for example, models such as VGGNet, ResNet, etc. After training, it can be used to obtain the document feature information of the property information document. For example, an image of the property information document can be taken and the feature information of the image can be obtained as the document feature information. Alternatively, the document recognition model is a natural language processing model, such as an RNN model, an LSTM model, etc., which can be used to determine the semantic information of the text in the property information document as the document feature information. The present invention does not limit this. The first facial feature information, the second facial feature information, the first signature feature information, the second signature feature information, and the document feature information are vectors with the same dimension.

[0024] According to an embodiment of the present invention, in step S7, the first relationship information between multiple owners and the second relationship information between multiple seller personnel can be obtained. For example, the first relationship information between two owners may include father-son relationship, mother-son relationship, husband-wife relationship, partnership relationship, etc. Some of these relationships can change, such as husband-wife relationship, partnership relationship, and some relationships cannot change, such as father-son relationship, mother-son relationship. The second relationship information is similar to the first relationship information and will not be elaborated here.

[0025] According to an embodiment of the present invention, in step S8, the above information can be verified through a graph neural network model, and the graph neural network model can be a graph convolutional neural network model, which is not limited in the present invention. For example, the graph neural network model processes the certificate feature information, the first relationship information, the first facial feature information, and the first signature feature information to obtain property right reference feature information, and processes the certificate feature information, the second relationship information, the second facial feature information, and the second signature feature information through the graph neural network model to obtain property right verification feature information. The property right reference feature information can reflect various information of the registered owner in the background database (including information such as signature, facial image, and first relationship information), and the property right verification feature information can reflect various information obtained for verification when selling the property (including information such as signature, facial image, and second relationship information). If there is a large difference between the property right verification feature information and the property right reference feature information, it may be that there is a difference between the owner and the seller personnel for verification. For example, the seller personnel do not include all the owners, the facial image of the seller personnel does not match the facial image of the owner, the signature of the seller personnel does not match the signature of the owner, and the relationship between the seller personnel is different from the relationship between the owners (for example, the relationship has changed, such as a divorce between husband and wife, or partners no longer cooperate). In this case, there may be risks in the transaction (for example, the property is jointly owned property to be divided during a divorce, and there may be legal risks in trading the property), and a risk reminder can be made in the verification result.

[0026] Figure 2 Exemplarily shows a schematic diagram of obtaining a verification result according to an embodiment of the present invention.

[0027] According to an embodiment of the present invention, in step S8, according to the document feature information, the first relationship information, the first facial feature information, the first signature feature information, and the trained graph neural network model, property right reference feature information is obtained, and according to the document feature information, the second relationship information, the second facial feature information, the second signature feature information, and the trained graph neural network model, property right verification feature information is obtained, including: processing the first relationship information through the trained relationship encoding model to obtain a first relationship feature, and inputting the first relationship feature into the first fully connected layer to obtain a first relationship weight; processing the second relationship information through the trained relationship encoding model to obtain a second relationship feature, and inputting the second relationship feature into the first fully connected layer to obtain a second relationship weight; obtaining a first graph structure according to the first relationship weight, the property right ratio of the owner, the subordination relationship between the reference signature image data and the owner, the document feature information, the first facial feature information, and the first signature feature information; obtaining a second graph structure according to the second relationship weight, the property right ratio of the seller, the subordination relationship between the comparison signature image data and the seller, the document feature information, the second facial feature information, and the second signature feature information; processing the first graph structure through the trained graph neural network model to obtain property right reference feature information; and processing the second graph structure through the trained graph neural network model to obtain property right verification feature information.

[0028] According to an embodiment of the present invention, the first relationship information can be represented by a canonical text. For example, the first relationship information between owner A and owner B is "spouse relationship", and a list of multiple relationships can also be provided for the owner to choose from. Based on the canonical text, through the relationship encoding model, which can be a natural language processing model, such as an RNN model or an LSTM model, the text information can be processed to obtain the semantic vector of the text information, that is, the first relationship feature. The first relationship feature can be input into the first fully connected layer, and after being processed by the activation function, the first relationship weight can be obtained. Based on the same method, the first relationship weights between each owner can be obtained. Similarly, the second relationship weights between each seller can be determined.

[0029] According to an embodiment of the present invention, a first graph structure is obtained based on a first relationship weight, the property right ratio of the owner, the subordination relationship between the reference signature image data and the owner, document feature information, first facial feature information, and first signature feature information, including: setting the document feature information, the first facial feature information, and the first signature feature information as the first node information of the nodes of the first graph structure; setting the property right ratio of the owner as the connection weight between the node corresponding to the document feature information and the node corresponding to the first facial feature information of the owner in the first graph structure; setting the first relationship weight between different owners as the connection weight between the nodes corresponding to the first facial feature information of different owners; determining the connection weight between the node corresponding to the first signature feature information and the node corresponding to the first facial feature information of the owner according to the subordination relationship between the reference signature image data and the owner; obtaining a first adjacency matrix according to the connection weight; obtaining a first degree matrix according to the first adjacency matrix; and obtaining a first graph structure according to the first node information, the first adjacency matrix, and the first degree matrix.

[0030] According to an embodiment of the present invention, in the first graph structure, the property right information document, the reference facial image data of each owner, and the reference signature image data can be used as nodes, the document feature information is the first node information of the node corresponding to the property right information document, the first facial feature information is the first node information of the node corresponding to the reference facial image data of the owner, and the first signature feature information is the first node information of the node corresponding to the reference signature image data of the owner.

[0031] According to an embodiment of the present invention, in the first graph structure, information related to edges can also be determined. If two nodes are connected, there is an edge between the two nodes. Moreover, an edge can not only represent whether there is a relationship between two nodes, but also determine the closeness of the relationship between the two nodes through the connection weight of the edge. Among them, the property right ratio of the owner can be set as the connection weight between the node corresponding to the certificate feature information and the node corresponding to the first facial feature information of the owner in the first graph structure. For example, if owner A owns 50% of the property right of this property, the connection weight between the node corresponding to the first facial feature information of owner A and the node corresponding to the certificate feature information is 0.5. And the first relationship weight between different owners can be set as the connection weight between the nodes corresponding to different owners. For example, if the first relationship weight between owner A and owner B is 0.6, the connection weight between the nodes corresponding to owner A and owner B is also 0.6. Further, based on the membership relationship between the reference signature image data and the owner, the connection weight between the node corresponding to the first signature feature information and the node corresponding to the first facial feature information of the owner can be determined. For example, if the reference signature image data A is an image of the signature of owner A, then the reference signature image data A belongs to owner A, and the connection weight between the node corresponding to the reference signature image data A and the node corresponding to owner A is 1. There is no connection relationship between the nodes corresponding to other owners and the reference signature image data A, and the connection weight is 0. There is also no connection relationship between the node corresponding to the certificate feature information and the nodes corresponding to each reference signature image data, and the connection weight is 0. After the above processing, the connection weights of each edge can objectively represent the relationship between each node.

[0032] According to an embodiment of the present invention, based on the above connection weights, a first adjacency matrix can be determined. In the first adjacency matrix, the data in the i-th row and j-th column is the connection weight between the i-th node and the j-th node. Both the i-th node and the j-th node can be any one of the nodes corresponding to the above certificate feature information, the first facial feature information of the owner, and the first signature feature information. After determining the first adjacency matrix, a first degree matrix can be obtained. For example, the i-th data on the diagonal of the first degree matrix is equal to the sum of all the data in the i-th row of the first adjacency matrix. In the first degree matrix, except for the data on the diagonal, other data are all 0, and both i and j are positive integers.

[0033] According to an embodiment of the present invention, based on the first node information, the first adjacency matrix, and the first degree matrix obtained above, a first graph structure can be obtained, that is, the information included in the first graph structure is the first node information, the first adjacency matrix, and the first degree matrix. Similarly, a second graph structure can be obtained.

[0034] According to an embodiment of the present invention, the trained graph neural network model can process the first graph structure, and based on the first degree matrix, the first adjacency matrix, the first node information, and the parameters of the graph neural network model itself, obtain the feature information of each node, that is, the original first node information of the node and the first node information of its adjacent nodes are processed through the first degree matrix, the first adjacency matrix, and the parameters of the graph neural network model itself, so as to fuse the original first node information of the node with the first node information of its adjacent nodes (nodes with a connection relationship), obtain the feature information of the node, and each node performs the above processing through the graph neural network model, so as to fuse with the first node information of the adjacent nodes, thereby obtaining the feature information of the node. In other words, after fusing the first node information with the first node information of the surrounding neighbor nodes, it is updated to the said feature information. And, the feature information of the node corresponding to the document feature information is the property right reference feature information. Similarly, the second graph structure can be processed to obtain the property right verification feature information of the node corresponding to the document feature information.

[0035] According to an embodiment of the present invention, in step S9, the property right reference feature information and the property right verification feature information can be used to obtain a verification result. For example, considering that even for the same person, the facial images and signatures taken at different times may have some differences, so a preset threshold is set, for example, 0.8 and 0.5. If the similarity (such as cosine similarity) between the property right reference feature information and the property right verification feature information is higher than 0.8, the verification result is that the legal risk is lower. If the similarity between the property right reference feature information and the property right verification feature information is higher than 0.5 and less than or equal to 0.8, the verification result is that the legal risk is medium. If the similarity between the property right reference feature information and the property right verification feature information is less than or equal to 0.5, the verification result is that the legal risk is higher.

[0036] In this way, not only can the facial images and signatures of the seller personnel be verified, but also the relationships between the seller personnel can be verified. And when verifying, the relationship weight of the relationships between the seller personnel can be calculated, the subordination relationship between the signature and the owner, as well as the property right ratio of the owner can be determined, so as to objectively determine the adjacency matrix, and through the graph neural network model, obtain the property right reference feature information that can comprehensively and accurately describe the real estate and its property right owner, and the property right verification feature information of the sum features of the real estate and its seller personnel, and compare the property right reference feature information and the property right verification feature information. Thus, in the comparison process, various information such as facial images, signatures, and relationship information can be comprehensively compared, improving the accuracy of the verification result and helping to reduce the transaction risk.

[0037] According to an embodiment of the present invention, the above face recognition model, signature recognition model, document recognition model, relationship encoding model, and graph neural network model can be trained before use. The method further includes: Step S101, obtaining a sample property information document, a first facial image sample of the sample owner corresponding to the sample property information document, a first signature image sample, and first sample relationship information between each sample owner; Step S102, obtaining first sample facial feature information of the first facial image sample through the face recognition model; Step S103, obtaining first sample signature feature information of the first signature image sample through the signature recognition model; Step S104, obtaining sample document feature information of the sample property information document through the document recognition model; Step S105, obtaining a first sample relationship weight corresponding to the first sample relationship information through the relationship encoding model; Step S106, obtaining a first sample graph structure according to the sample document feature information, the first sample facial feature information, the first sample signature feature information, the first sample relationship weight, and the property ratio of the sample owner; Step S107, performing at least one of the following processes: using other facial image samples of the same sample owner to replace the first facial image sample and obtaining second sample facial feature information, using other signature image samples of the same sample owner to replace the first signature image sample and obtaining second sample signature feature information; Step S108, obtaining a second sample graph structure based on at least one of the second sample facial feature information and the second sample signature feature information; Step S109, performing at least one of the following processes: using facial image samples of different persons to replace the first facial image sample and obtaining third sample facial feature information, using signature image samples obtained by different persons signing the name of the sample owner to replace the first signature image sample and obtaining third sample signature feature information, randomly modifying the first sample relationship information between sample owners and obtaining a modified second sample relationship weight; Step S110, obtaining a third sample graph structure based on at least one of the third sample facial feature information, the third sample signature feature information, and the second sample relationship weight; Step S111, respectively processing the first sample graph structure, the second sample graph structure, and the third sample graph structure through the graph neural network model to obtain first property sample feature information corresponding to the first sample graph structure, second property sample feature information corresponding to the second sample graph structure, and third property sample feature information corresponding to the third sample graph structure; Step S112: Obtain a signature loss function based on the first sample signature feature information, the second sample signature feature information, and the third sample signature feature information; Step S113: Determine a relationship weight attention coefficient based on the first sample relationship weight and the second sample relationship weight; Step S114: Obtain a global loss function based on the first property sample feature information, the second property sample feature information, the third property sample feature information, and the relationship weight attention coefficient; Step S115: Train a face recognition model, a signature recognition model, a document recognition model, a relationship encoding model, and a graph neural network model based on the global loss function and the signature loss function.

[0038] According to an embodiment of the present invention, in steps S101 to S106, the method of obtaining the first sample graph structure is similar to the method of obtaining the first graph structure above, and will not be elaborated here. In step S107, in order to improve the robustness and adaptability of the model, at least one of the following processes may be executed: use other facial image samples of the same sample owner to replace the first facial image sample and obtain the second sample facial feature information, use other signature image samples of the same sample owner to replace the first signature image sample and obtain the second sample signature feature information. In this way, the above models can still accurately judge when facing errors in facial images and signatures of the same person, that is, they can still accurately recognize the facial images and signatures of the same person. Moreover, this method can also obtain more positive samples in a simple way, improve the training intensity, and reduce the training cost. Further, in step S108, the second sample graph structure can be obtained in a manner similar to the method of obtaining the first sample graph structure above.

[0039] According to an embodiment of the present invention, in step S109, in order to improve the model's ability to identify risks when the seller is different from the owner, the signatures are inconsistent, or the relationship between the owners changes, at least one of the following processes may be executed: use facial image samples of different people to replace the first facial image sample and obtain the third sample facial feature information, use signature image samples obtained by signing the names of different sample owners to replace the first signature image sample and obtain the third sample signature feature information, randomly modify the first sample relationship information between the sample owners and obtain the modified second sample relationship weight. The determination method of the second sample relationship weight is similar to the determination method of the first relationship weight above and will not be elaborated here. In this way, more negative samples can be obtained in a simple way, the training intensity can be improved, the training cost can be reduced, and the model's ability to identify risks can be enhanced. Further, in step S110, the third sample graph structure can be obtained in a manner similar to the method of obtaining the first sample graph structure above.

[0040] According to an embodiment of the present invention, in step S111, the first property sample feature information corresponding to the first sample graph structure, the second property sample feature information corresponding to the second sample graph structure, and the third property sample feature information corresponding to the third sample graph structure can be obtained through a graph neural network model. The obtaining method is similar to the above-mentioned property reference feature information and property verification feature information, and will not be elaborated here.

[0041] According to an embodiment of the present invention, in step S112, a signature loss function is obtained according to the first sample signature feature information, the second sample signature feature information, and the third sample signature feature information, including: decoding the first sample signature feature information to obtain a first word vector of the text corresponding to the first signature feature information; decoding the second sample signature feature information to obtain a second word vector of the text corresponding to the second signature feature information; decoding the third sample signature feature information to obtain a third word vector of the text corresponding to the third signature feature information; concatenating the first sample signature feature information and the second sample signature feature information to obtain a first concatenated vector; concatenating the first sample signature feature information and the third sample signature feature information to obtain a second concatenated vector; inputting the first concatenated vector into a second fully connected layer to obtain first matching probability information; inputting the second concatenated vector into a second fully connected layer to obtain second matching probability information; obtaining a signature loss function according to the first word vector, the second word vector, the third word vector, the first matching probability information, and the second matching probability information.

[0042] According to an embodiment of the present invention, the first sample signature feature information, the second sample signature feature information, and the third sample signature feature information can be decoded by a decoder (for example, CTC decoding), and the recognition result of the text corresponding to the first signature feature information can be output. And through a natural language processing model, the recognition result is processed to obtain the word vector of each character in the text corresponding to the first signature feature information, and the word vectors of each character are weighted and summed to obtain the first word vector. Similarly, the second word vector and the third word vector can be obtained.

[0043] According to an embodiment of the present invention, the first word vector and the second word vector should theoretically be the same, and the first sample signature feature information representing the image features of the first signature image sample and the second sample signature feature information of the image features of other signatures of the same owner should theoretically be similar. After splicing the first sample signature feature information and the second sample signature feature information, a first splicing vector can be obtained. The first word vector and the third word vector should theoretically be the same, but the first sample signature feature information representing the image features of the first signature image sample and the third sample signature feature information of the signature image sample obtained by signing the sample owner by different persons should theoretically be quite different, so that the model can identify the same signature signed by different persons. Of course, the cosine similarity between the first sample signature feature information and the second sample signature feature information can also be directly used as the first matching probability information, and the cosine similarity between the first sample signature feature information and the third sample signature feature information can be used as the second matching probability information. The present invention does not limit this.

[0044] According to an embodiment of the present invention, based on the first word vector, the second word vector, the third word vector, the first matching probability information, and the second matching probability information, a signature loss function is obtained, including: obtaining the signature loss function according to formula (1) , (1) Wherein, is the first word vector, is the second word vector, is the third word vector, sim is a function for calculating similarity, is the first matching probability information, is the second matching probability information, is a preset parameter.

[0045] According to an embodiment of the present invention, in formula (1), can be equal to 1, or less than 1. For example, 0.99. The present invention does not limit this. As described above, the first word vector and the second word vector should theoretically be the same, and the first sample signature feature information and the second sample signature feature information should theoretically be similar. Therefore, theoretically should be close to 1, or close to , that is, the similarity (for example, cosine similarity) between the first word vector and the second word vector should theoretically be close to 1, and the first matching probability information indicating that the first sample signature feature information and the second sample signature feature information are similar should theoretically be close to 1, so that the model can accurately identify the signature of the same person when there are small errors in the signatures of the same person at different times. Therefore, is and The error between them can be decreased during the training process to improve the value and enhance the model's signature recognition ability and robustness.

[0046] According to an embodiment of the present invention, as described above, the first word vector and the third word vector should theoretically be the same, but the first sample signature feature information and the third sample signature feature information should theoretically be quite different. Therefore, during the training process, make decrease, so that the second matching probability information indicating the similarity between the first sample signature feature information and the third sample signature feature information decreases, and the indicating the similarity between the first word vector and the third word vector is enhanced (i.e., make decrease). Thus, while improving the model's signature recognition ability, it can distinguish the same name signed by different people and reduce the possibility of forged signatures. Add the above and to obtain the signature loss function.

[0047] In this way, through the construction of the signature loss function, the model's recognition ability for the signature text can be enhanced, and the model can still accurately recognize the signatures of the same person when there are slight errors in the signatures of the same person at different times, improving the model's robustness, and can distinguish the same name signed by different people, reducing the possibility of forged signatures.

[0048] According to an embodiment of the present invention, in step S113, according to the first sample relationship weight and the second sample relationship weight, determining the relationship weight attention coefficient includes: determining the average value of the first sample relationship weights corresponding to the unmodified first sample relationship information and the minimum value of the relationship weight change corresponding to the modified first sample relationship information according to the first sample relationship weight and the second sample relationship weight; and determining the ratio of the average value of the first sample relationship weights corresponding to the unmodified first sample relationship information to the minimum value of the relationship weight change corresponding to the modified first sample relationship information as the relationship weight attention coefficient.

[0049] According to an embodiment of the present invention, when randomly modifying the first sample relationship information among sample owners and obtaining the modified second sample relationship weights, the first relationship information of some owners is modified. For example, owner A and owner B were originally in a spousal relationship, and during the modification training, this relationship is modified to no relationship (it can be considered that owner A and owner B are divorced). There may be a certain error between the relationship encoding model and the first fully connected layer. Therefore, it is possible that the first relationship information of the owner is modified, but the change in the relationship weight between the owners is small, resulting in the possibility that it is difficult for the model to recognize. Therefore, the first sample relationship weight corresponding to the first sample relationship information before modification and the second sample relationship weight corresponding to the first sample relationship information after modification can be determined, that is, and the difference between the two is determined. For example, the absolute value is taken after subtracting the first sample relationship weight from the second sample relationship weight as the change in the relationship weight, and the minimum value of the change in the relationship weight corresponding to multiple modified first sample relationship information is determined. During training, the minimum value of the change in the relationship weight can be amplified, so that before and after the relationship modification, the gap between the first sample relationship weight and the second sample relationship weight is enlarged, making it easier for the model to recognize the change in the relationship and improving the model's discrimination ability before and after the relationship change. The ratio of the average value of the first sample relationship weights corresponding to the unmodified first sample relationship information to the minimum value of the change in the relationship weight corresponding to the modified first sample relationship information can be determined as the relationship weight attention coefficient, that is, the minimum value of the change in the relationship weight is the denominator, and reducing the relationship weight attention coefficient can enlarge the gap between the first sample relationship weight and the second sample relationship weight before and after the relationship modification.

[0050] According to an embodiment of the present invention, in step S114, a global loss function is obtained based on the first property sample feature information, the second property sample feature information, the third property sample feature information, and the relationship weight attention coefficient, including: obtaining the global loss function according to formula (2) , (2) Wherein, is the first property sample feature information, is the second property sample feature information, is the third property sample feature information, is the relationship weight attention coefficient, sim is a function for calculating similarity, is a preset parameter.

[0051] According to an embodiment of the present invention, It can be equal to 1, or less than 1. For example, 0.99. The present invention does not limit this. When obtaining the second property sample feature information, facial images taken at different times or signatures signed at different times of the same owner are used. However, these facial images and signatures still belong to the same person. Therefore, the finally obtained second property sample feature information should theoretically be similar to the first property sample feature information, thereby improving the robustness of the model. is the similarity (e.g., cosine similarity) between the first property sample feature information and the second property sample feature information and the gap of. The gap can be narrowed to improve the similarity between the first property sample feature information and the second property sample feature information and enhance the robustness of the model.

[0052] According to an embodiment of the present invention, when obtaining the third property sample feature information, facial images of different persons are used, or forged signatures are used. Therefore, the finally obtained third property sample feature information is theoretically not similar to the first property sample feature information, and the similarity between the two is lower, and the stronger the model's recognition ability for the situation of different facial images or forged signatures. Further, as described above, narrowing the relationship weight attention coefficient can widen the gap between the first sample relationship weight and the second sample relationship weight before and after the relationship modification, thereby enhancing the model's recognition ability for the risk caused by the relationship change. Therefore, the relationship weight attention coefficient can be multiplied by the similarity between the first property sample feature information and the third property sample feature information. During training, make this product decrease to narrow the relationship weight attention coefficient and the similarity between the third property sample feature information and the first property sample feature information, enhance the model's recognition ability for the situation of different facial images or forged signatures, and the recognition ability for the risk caused by the relationship change. Add the above and to obtain the global loss function.

[0053] In this way, through the construction of the global loss function, the model's recognition ability for facial images taken at different times or signatures signed at different times of the same owner can be improved, the model robustness can be enhanced, and the model's recognition ability for the situation of different facial images or forged signatures, and the recognition ability for the risk caused by the relationship change can be enhanced, which is beneficial to reducing transaction risks.

[0054] According to an embodiment of the present invention, in step S115, the face recognition model, signature recognition model, document recognition model, relationship encoding model, and graph neural network model, as well as the first fully connected layer, can be trained through a global loss function and a signature loss function. For example, in a backpropagation manner, the parameters of the above-mentioned multiple models can be updated through the global loss function and the signature loss function, so as to train the multiple models. After multiple trainings and verifying that the accuracy of the model meets the requirements in the validation set (for example, the accuracy rate of obtaining the verification result reaches a predetermined standard), the training can be completed, and the trained face recognition model, trained signature recognition model, trained document recognition model, trained relationship encoding model, and trained graph neural network model can be obtained and used for the processing of determining the verification result.

[0055] The artificial intelligence-based real estate registration verification method according to the embodiment of the present invention can not only verify the facial image data and signature image of the seller to ensure the authenticity of the sale, but also verify whether the relationship between multiple owners has changed through the graph neural network model. If the authenticity of the sale is in doubt or the relationship between multiple owners has changed, it can prompt that there is a risk in the transaction, thereby reducing the legal risk of the transaction and enhancing the transaction security. It can not only verify the facial images and signatures of the seller personnel, but also verify the relationship between the seller personnel. When verifying, the relationship weight of the relationship between the seller personnel can be calculated, and the subordination relationship between the signature and the owner, as well as the property right ratio of the owner, can be determined, so as to objectively determine the adjacency matrix, and obtain the property right reference feature information that can comprehensively and accurately describe the real estate and its property right owners and the property right verification feature information of the sum features of the real estate and its seller personnel through the graph neural network model, and compare the property right reference feature information and the property right verification feature information. Therefore, in the comparison process, various information such as facial images, signatures, and relationship information can be comprehensively compared to improve the accuracy of the verification result and help reduce the transaction risk. When determining the signature loss function, the recognition ability of the model for the text of the signature can be improved through the construction of the signature loss function, and the model can still accurately recognize the signature of the same person when there are small errors in the signatures of the same person at different times, improving the robustness of the model and being able to distinguish the same name signed by different people, reducing the possibility of forged signatures. When determining the global loss function, the recognition ability of the model for facial images taken at different times or signatures signed at different times of the same owner can be improved through the construction of the global loss function, improving the robustness of the model, and the recognition ability of the model for the case of different facial images or forged signatures, as well as the recognition ability of the risk caused by relationship changes, which is beneficial to reducing the transaction risk.

[0056] Figure 3A block diagram of an artificial intelligence-based real estate registration verification system according to an embodiment of the present invention is exemplarily shown. The system includes: A first acquisition module, configured to acquire the property right information certificate of the property to be transferred, and determine the owner information of the property according to the property right information certificate; A second acquisition module, configured to acquire the reference facial image data and reference signature image data of the owner corresponding to the owner information through the background database; A third acquisition module, configured to acquire the comparison facial image data and comparison signature image data of the seller; A face recognition module, configured to obtain the first facial feature information of the reference facial image data and the second facial feature information of the comparison facial image data through a trained face recognition model; A signature recognition module, configured to obtain the first signature feature information of the reference signature image data and the second signature feature information of the comparison signature image data through a trained signature recognition model; A document recognition module, configured to obtain the document feature information of the property right information certificate through a trained document recognition model; A relationship module, configured to obtain the first relationship information between multiple owners and the second relationship information between multiple sellers; A verification module, configured to obtain the property right reference feature information according to the document feature information, the first relationship information, the first facial feature information, the first signature feature information and a trained graph neural network model, and obtain the property right verification feature information according to the document feature information, the second relationship information, the second facial feature information, the second signature feature information and a trained graph neural network model; A verification result module, configured to obtain a verification result according to the property right reference feature information and the property right verification feature information.

[0057] The present invention may be a method, an apparatus, a system, and / or a computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions thereon for performing various aspects of the present invention.

[0058] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the drawings are only examples and do not limit the present invention. The object of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been shown and described in the embodiments. Without departing from the above principles, the embodiments of the present invention may have any deformation or modification.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An artificial intelligence-based real estate registration verification method, characterized in that, Including: Obtain the property right information certificate of the property to be transferred, and determine the owner information of the property according to the property right information certificate; Through the background database, obtain the reference facial image data and reference signature image data of the owner corresponding to the owner information; Obtain the comparison facial image data and comparison signature image data of the seller; Through the trained face recognition model, obtain the first facial feature information of the reference facial image data and the second facial feature information of the comparison facial image data; Through the trained signature recognition model, obtain the first signature feature information of the reference signature image data and the second signature feature information of the comparison signature image data; Through the trained document recognition model, obtain the document feature information of the property right information certificate; Obtain the first relationship information among multiple owners and the second relationship information among multiple sellers; According to the document feature information, the first relationship information, the first facial feature information, the first signature feature information and the trained graph neural network model, obtain the property right reference feature information, and according to the document feature information, the second relationship information, the second facial feature information, the second signature feature information and the trained graph neural network model, obtain the property right verification feature information; Obtain the verification result according to the property right reference feature information and the property right verification feature information.

2. The method for verifying real estate registration based on artificial intelligence according to claim 1, wherein According to the document feature information, the first relationship information, the first facial feature information, the first signature feature information and the trained graph neural network model, obtain the property right reference feature information, and according to the document feature information, the second relationship information, the second facial feature information, the second signature feature information and the trained graph neural network model, obtain the property right verification feature information, including: Process the first relationship information through the trained relationship encoding model to obtain the first relationship feature, and input the first relationship feature into the first fully connected layer to obtain the first relationship weight; Process the second relationship information through the trained relationship encoding model to obtain the second relationship feature, and input the second relationship feature into the first fully connected layer to obtain the second relationship weight; According to the first relationship weight, the property right ratio of the owner, the subordination relationship between the reference signature image data and the owner, the document feature information, the first facial feature information and the first signature feature information, obtain the first graph structure; According to the second relationship weight, the property right ratio of the seller, the subordination relationship between the comparison signature image data and the seller, the document feature information, the second facial feature information and the second signature feature information, obtain the second graph structure; Process the first graph structure through the trained graph neural network model to obtain the property right reference feature information; Process the second graph structure through the trained graph neural network model to obtain the property right verification feature information.

3. The method for verifying real estate registration based on artificial intelligence according to claim 2, wherein, According to the first relationship weight, the property right ratio of the owner, the subordination relationship between the reference signature image data and the owner, the document feature information, the first facial feature information and the first signature feature information, obtain the first graph structure, including: Set the document feature information, the first facial feature information and the first signature feature information as the first node information of the nodes of the first graph structure; Set the property right ratio of the owner to the connection weight between the node corresponding to the document feature information in the first graph structure and the node corresponding to the first facial feature information of the owner; Set the first relationship weight between different owners to the connection weight between the nodes corresponding to the first facial feature information of different owners; Determine the connection weight between the node corresponding to the first signature feature information and the node corresponding to the first facial feature information of the owner according to the membership relationship between the reference signature image data and the owner; Obtain the first adjacency matrix according to the connection weight; Obtain the first degree matrix according to the first adjacency matrix; Obtain the first graph structure according to the first node information, the first adjacency matrix and the first degree matrix.

4. The method for verifying real estate registration based on artificial intelligence according to claim 1, wherein The method further includes: Obtain a sample property right information document, the first facial image sample of the sample owner corresponding to the sample property right information document, the first signature image sample, and the first sample relationship information between each sample owner; Obtain the first sample facial feature information of the first facial image sample through a face recognition model; Obtain the first sample signature feature information of the first signature image sample through a signature recognition model; Obtain the sample document feature information of the sample property right information document through a document recognition model; Obtain the first sample relationship weight corresponding to the first sample relationship information through a relationship coding model; Obtain the first sample graph structure according to the sample document feature information, the first sample facial feature information, the first sample signature feature information, the first sample relationship weight, and the property right ratio of the sample owner; Perform at least one of the following processes: use other facial image samples of the same sample owner to replace the first facial image sample and obtain the second sample facial feature information, use other signature image samples of the same sample owner to replace the first signature image sample and obtain the second sample signature feature information; Obtain the second sample graph structure based on at least one of the second sample facial feature information and the second sample signature feature information; Perform at least one of the following processes: use facial image samples of different persons to replace the first facial image sample and obtain the third sample facial feature information, use signature image samples obtained by different persons signing the name of the sample owner to replace the first signature image sample and obtain the third sample signature feature information, randomly modify the first sample relationship information between sample owners and obtain the modified second sample relationship weight; Obtain the third sample graph structure based on at least one of the third sample facial feature information, the third sample signature feature information, and the second sample relationship weight; Process the first sample graph structure, the second sample graph structure, and the third sample graph structure respectively through a graph neural network model to obtain the first property sample feature information corresponding to the first sample graph structure, the second property sample feature information corresponding to the second sample graph structure, and the third property sample feature information corresponding to the third sample graph structure; Obtain a signature loss function according to the first sample signature feature information, the second sample signature feature information, and the third sample signature feature information; Determine a relationship weight attention coefficient according to the first sample relationship weight and the second sample relationship weight; Obtain a global loss function based on the first property sample feature information, the second property sample feature information, the third property sample feature information, and the relationship weight attention coefficient; Train a face recognition model, a signature recognition model, a document recognition model, a relationship encoding model, and a graph neural network model based on the global loss function and the signature loss function.

5. The method for verifying real estate registration based on artificial intelligence according to claim 4, wherein Obtain a signature loss function according to the first sample signature feature information, the second sample signature feature information, and the third sample signature feature information, including: Decode the first sample signature feature information to obtain the first word vector of the text corresponding to the first signature feature information; Decode the second sample signature feature information to obtain the second word vector of the text corresponding to the second signature feature information; Decode the third sample signature feature information to obtain the third word vector of the text corresponding to the third signature feature information; Concatenate the first sample signature feature information and the second sample signature feature information to obtain a first concatenated vector; Concatenate the first sample signature feature information and the third sample signature feature information to obtain a second concatenated vector; Input the first concatenated vector into the second fully connected layer to obtain first matching probability information; Input the second concatenated vector into the second fully connected layer to obtain second matching probability information; Obtain a signature loss function according to the first word vector, the second word vector, the third word vector, the first matching probability information, and the second matching probability information.

6. The method for verifying real estate registration based on artificial intelligence according to claim 5, wherein Obtain a signature loss function according to the first word vector, the second word vector, the third word vector, the first matching probability information, and the second matching probability information, including: According to the formula ； Obtain a signature loss function , where is the first word vector, is the second word vector, is the third word vector, sim is a function for calculating similarity, is the first matching probability information, is the second matching probability information, is a preset parameter.

7. The method for verifying real estate registration based on artificial intelligence according to claim 4, wherein Determine the relationship weight attention coefficient according to the first sample relationship weight and the second sample relationship weight, including: Determine the average value of the first sample relationship weights corresponding to the unmodified first sample relationship information and the minimum value of the relationship weight change corresponding to the modified first sample relationship information according to the first sample relationship weight and the second sample relationship weight; Determine the ratio of the average value of the first sample relationship weights corresponding to the unmodified first sample relationship information to the minimum value of the relationship weight change corresponding to the modified first sample relationship information as the relationship weight attention coefficient.

8. The method for verifying real estate registration based on artificial intelligence according to claim 4, wherein Obtain a global loss function according to the first property sample feature information, the second property sample feature information, the third property sample feature information, and the relationship weight attention coefficient, including: According to the formula ； Obtain the global loss function , where is the first property sample feature information is the second property sample feature information is the third property sample feature information is the relationship weight attention coefficient, and sim is the similarity function is the preset parameter 9. An artificial intelligence-based real estate registration verification system for performing the method according to any one of claims 1-8, characterized in that, Including: A first acquisition module for acquiring the property information document of the property to be transferred and determining the owner information of the property according to the property information document; A second acquisition module for acquiring the reference facial image data and the reference signature image data of the owner corresponding to the owner information through the background database; A third acquisition module for acquiring the comparison facial image data and the comparison signature image data of the seller; A face recognition module for obtaining the first facial feature information of the reference facial image data and the second facial feature information of the comparison facial image data through the trained face recognition model; A signature recognition module for obtaining the first signature feature information of the reference signature image data and the second signature feature information of the comparison signature image data through the trained signature recognition model; The document recognition module is used to obtain the document feature information of the property right information document through the trained document recognition model; The relationship module is used to obtain the first relationship information among multiple owners and the second relationship information among multiple seller personnel; The verification module is used to obtain the property right reference feature information according to the document feature information, the first relationship information, the first facial feature information, the first signature feature information and the trained graph neural network model, and obtain the property right verification feature information according to the document feature information, the second relationship information, the second facial feature information, the second signature feature information and the trained graph neural network model; The verification result module is used to obtain the verification result according to the property right reference feature information and the property right verification feature information.

Images