Intelligent lamp pole census method based on fusion of visual recognition and blockchain storage

Abstract

The application discloses an intelligent lamp pole general investigation method based on fusion of visual identification and blockchain storage, and comprises the following steps: adopting a neural network architecture to perform visual identification on image data of a target lamp pole to obtain structured data; forming collection data by combining the image data of the target lamp pole, correction coordinates and the structured data; comparing and verifying the structured data with the image data of the target lamp pole; when the comparison and verification is passed, performing logical verification on the collection data; when the logical verification is passed, forming a standardized file by combining the collection data with a hash value corresponding to the collection data, updating the standardized file to a cloud database, automatically identifying the characteristics of the target lamp pole and forming a standardized file uploaded to a cloud server, which can replace traditional manual feature acquisition and manual feature input, optimizes the general investigation process of the target lamp pole, improves the general investigation efficiency of the target lamp pole, and effectively reduces errors in the general investigation of the target lamp pole.

Description

Technical Field

[0001] This invention relates to the field of light pole survey technology, and more specifically, to an intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage. Background Technology

[0002] Against the backdrop of the rapid advancement of smart city construction, light poles, as core infrastructure carriers supporting multiple functions such as lighting, communication, and monitoring, directly determine the scientific nature of urban resource allocation and the efficiency of management decisions based on the accuracy, reliability, and timeliness of their survey data. The light pole survey and database construction is the foundation for the operation of smart lighting monitoring systems and a key prerequisite for promoting urban informatization and achieving refined urban management. It is of great significance for optimizing the utilization of urban public resources and facilitating the asset-based management of smart light poles.

[0003] However, existing light pole survey technologies have several significant shortcomings in practical applications, making it difficult to meet the high standards required for smart city construction: First, they rely excessively on manual data entry and information verification, resulting in highly repetitive processes and time-consuming on-site operations. This is especially true in large-scale urban light pole surveys, which require substantial manpower and lengthy survey cycles. Second, manual operations are prone to errors such as data entry and omissions, and lack effective data authenticity control mechanisms. Furthermore, judgments on light pole status and hazard levels rely on human experience, leading to inconsistent data and compromising accuracy and completeness. Third, data storage often employs a centralized model, posing risks of tampering and loss. The lack of a traceability mechanism throughout the data lifecycle makes it difficult to meet the data reliability requirements of cross-departmental collaborative scheduling.

[0004] A prior art census system and method (application number: 201810999675.X, application date: 2018.08.30) discloses a census system and method, including: acquiring street view images of the census area, the shooting location of the street view images, and the shooting direction; performing image recognition on the street view images to identify census objects, known objects, and the morphological characteristics of the census objects in the street view images; calculating the length characteristics and environmental characteristics of the census objects based on the known dimensions of the known objects; and calculating the location of the census objects based on the known dimensions of the known objects, the shooting location, and the shooting direction. In other words, the prior art can automatically measure and calculate the parameters of census objects within the street view, but this method cannot automatically identify the characteristics of census objects within the street view and form standardized files for uploading to a cloud server, thus failing to optimize the census process.

[0005] Therefore, how to automatically identify the characteristics of target light poles and create standardized files for uploading to a cloud server is an urgent problem to be solved. Summary of the Invention

[0006] In view of this, the present invention provides a smart light pole survey method based on the fusion of visual recognition and blockchain evidence storage, comprising:

[0007] The task of surveying light poles is distributed to the application terminal, and the task of surveying light poles includes target light poles;

[0008] Receive image data and positioning data of the target light pole from the application terminal;

[0009] The positioning data of the target light pole is corrected to obtain the corrected coordinates;

[0010] A neural network architecture is used to perform visual recognition on the image data of the target light pole to obtain structured data. This includes: preprocessing the image data of the target light pole to obtain a preprocessed image; identifying feature information based on the preprocessed image; mapping the feature information to standard field representations to obtain recognition data; preprocessing the recognition data to obtain preprocessed data, which includes empty and non-empty fields; marking the empty fields of the preprocessed data as unrecognized; and converting the marked preprocessed data into the structured data according to a preset format.

[0011] The image data of the target light pole, the correction coordinates, and the structured data constitute the acquired data. The acquired data is then encrypted to obtain encrypted data, which includes a hash value.

[0012] The encrypted data is uploaded to a blockchain node, which includes a census execution node, a management and supervision node, and a data application node.

[0013] Perform a consistency check on the hash values ​​of the census execution node, the management and supervision node, and the data application node.

[0014] If the consistency check fails, an anomaly is marked and pushed to the application, which then waits for the application to re-feed back the image data and positioning data of the target light pole.

[0015] If the consistency check passes, the management and supervision node is controlled to decrypt the encrypted data to obtain the collected data. The structured data is then compared and verified with the image data of the target light pole, including:

[0016] The feature vector of the structured data is extracted using a convolutional neural network, and the feature vector of the image data of the target lamp post is extracted using the same convolutional neural network. The cosine similarity between the feature vector of the structured data and the feature vector of the image data of the target lamp post is then calculated.

[0017] If the cosine similarity is less than 90%, the comparison verification fails, an anomaly is marked, and the error is pushed to the application. The application then waits for the application to re-feed back the image data and location data of the target light pole.

[0018] If the cosine similarity is greater than or equal to 90%, the comparison verification passes, and logical verification is performed on the collected data, including:

[0019] The preset rule base includes required fields, geographical range, and associated verification groups. The associated groups include projects and the information range corresponding to the projects. The structured data includes data groups, and the data groups include names and the information corresponding to the names.

[0020] Determine whether the structured data includes all the required fields, determine whether the corrected coordinates are within the geographic range, and determine whether, for any set of identical names and items, the information corresponding to the name is within the information range corresponding to the item;

[0021] If the structured data includes all the required fields, the correction coordinates are within the geographical range, and for any set of identical names and items, the information corresponding to the name is within the information range corresponding to the item, then the logical verification passes, the collected data and the hash value corresponding to the collected data form a standardized archive, and the standardized archive is updated to the cloud database;

[0022] If the structured data does not include all the required fields, the corrected coordinates are not within the geographical range, or there is at least one set of the same name and item, and the information corresponding to the name is not within the information range corresponding to the item, then the logical verification fails, an anomaly is marked, and the error is pushed to the application, waiting for the application to re-feed back the image data and location data of the target light pole.

[0023] Optionally, encrypting the collected data to obtain encrypted data includes:

[0024] The encryption key is generated using a random number generator;

[0025] The collected data is encrypted using the aforementioned encryption key and symmetric encryption algorithm to obtain ciphertext;

[0026] Generate the hash value based on the ciphertext;

[0027] The management and supervision node includes a public key, and the encryption key is encrypted using the public key and an asymmetric encryption algorithm to obtain an encrypted packet;

[0028] The ciphertext, the hash value, and the encryption packet constitute the encrypted data.

[0029] Optionally, the management and supervision node decrypts the encrypted data to obtain the collected data, including:

[0030] The management and supervision node includes a private key;

[0031] The management and supervision node uses the private key to decrypt the encrypted packet to obtain the encryption key;

[0032] The management and supervision node uses the encryption key to decrypt the ciphertext to obtain the collected data.

[0033] Optionally, the calculation of the cosine similarity between the feature vector of the structured data and the feature vector of the image data of the target lamppost is performed in the following manner:

[0034] ;

[0035] in, The cosine similarity is... The feature vector of the structured data, The feature vector of the structured data, For vector dot product operation, This is a multiplication operation.

[0036] Optionally, the step of correcting the positioning data of the target light pole to obtain corrected coordinates includes:

[0037] The positioning data of the target light pole includes the coordinates of the ultra-wideband positioning point and the initial coordinates of the wide-area positioning.

[0038] The three-dimensional map is obtained, and the three-dimensional map includes feature points and the coordinates corresponding to the feature points;

[0039] Based on the coordinates of the ultra-wideband positioning point, a preset area is generated in the three-dimensional map, and three feature points are selected in the preset area.

[0040] Calculate the relative distance between the coordinates of the ultra-wideband positioning point and the coordinates corresponding to the selected feature point;

[0041] Calculate the theoretical distance between the initial coordinates of the wide-area positioning and the coordinates corresponding to the selected feature point;

[0042] The corrected horizontal coordinate, corrected vertical coordinate, and corrected vertical coordinate are calculated based on the initial coordinates of the wide-area positioning, the coordinates corresponding to the selected feature point, the relative distance between the coordinates of the ultra-wideband positioning point and the coordinates corresponding to the selected feature point, and the theoretical distance between the initial coordinates of the wide-area positioning and the coordinates corresponding to the selected feature point.

[0043] Based on the corrected abscissa and the corrected ordinate, a defined position is determined on the 3D map, and it is determined whether the deviation between the terrain elevation corresponding to the defined position on the 3D map and the corrected ordinate exceeds 0.5 meters;

[0044] If the distance is less than 0.5 meters, the corrected horizontal coordinate, the corrected vertical coordinate, and the corrected vertical coordinate shall constitute the corrected coordinate.

[0045] If the distance exceeds 0.5 meters, the corrected vertical coordinate is updated using the terrain elevation corresponding to the defined location on the 3D map, and the corrected horizontal coordinate, the corrected vertical coordinate, and the updated corrected vertical coordinate constitute the corrected coordinates.

[0046] Optionally, the calculation of the relative distance between the coordinates of the ultra-wideband positioning point and the coordinates corresponding to the selected feature point is performed in the following manner:

[0047]

[0048] in, The coordinates of the ultra-wideband positioning point and the selected first The relative distance between the coordinates of the feature points. Let x be the x-coordinate of the ultra-wideband positioning point. For the selected number The x-coordinate of the coordinates corresponding to each of the aforementioned feature points. Let be the ordinate of the ultra-wideband positioning point. For the selected number The ordinate of the coordinates corresponding to each of the aforementioned feature points. Let be the vertical coordinate of the ultra-wideband positioning point. For the selected number The vertical coordinates of the coordinates corresponding to each of the feature points.

[0049] Optionally, the step of calculating the corrected abscissa based on the initial coordinates of the wide-area positioning, the coordinates corresponding to the selected feature point, the relative distance between the coordinates of the ultra-wideband positioning point and the coordinates corresponding to the selected feature point, and the theoretical distance between the initial coordinates of the wide-area positioning and the coordinates corresponding to the selected feature point is performed in the following manner:

[0050]

[0051] in, The corrected abscissa, The x-coordinate of the initial coordinates for the wide-area positioning. The total number of selected feature points. For the selected number The x-coordinate of the coordinates corresponding to each of the aforementioned feature points. The coordinates of the ultra-wideband positioning point and the selected first The relative distance between the coordinates of the feature points. The initial coordinates for the wide-area positioning and the selected first The theoretical distance between the coordinates of the feature points.

[0052] Optionally, the preprocessing of the image data of the target light pole includes:

[0053] The image data of the target lamp post is subjected to at least one of the following processing methods: noise reduction, enhancement, cropping, and normalization.

[0054] Compared with existing technologies, the intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by this invention achieves at least the following beneficial effects:

[0055] 1. The intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by this invention includes: using a neural network architecture to perform visual recognition on the image data of the target light pole to obtain structured data; combining the image data of the target light pole, correction coordinates, and structured data to form collected data; comparing and verifying the structured data with the image data of the target light pole; when the comparison and verification pass, performing logical verification on the collected data; when the logical verification passes, combining the collected data with the hash value corresponding to the collected data to form a standardized file, updating the standardized file to the cloud database, realizing automatic identification of the characteristics of the target light pole and forming a standardized file uploaded to the cloud server, which can replace the traditional manual acquisition of features and manual feature entry, optimize the survey process of the target light pole, improve the survey efficiency of the target light pole, and effectively reduce the errors in the survey of the target light pole.

[0056] 2. The intelligent light pole survey method based on the fusion of visual recognition and blockchain notarization provided by this invention encrypts the collected data after it is obtained. After the encrypted collected data is uploaded, the hash value consistency of the blockchain nodes is checked. After the consistency check is passed, the collected data is decrypted and then used for subsequent data verification. By verifying the hash value consistency across nodes, the integrity and tamper-proofness of the data in the distributed storage can be verified from the bottom layer. The risk of tampering is eliminated from the data notarization level, ensuring that the data source for subsequent image feature comparison and logical verification is the original real data. This effectively solves the core problem of untrustworthy data sources in traditional verification methods.

[0057] Of course, any product implementing this invention does not necessarily need to achieve all of the technical effects described above at the same time.

[0058] Other features and advantages of the invention will become clear from the following detailed description of exemplary embodiments of the invention with reference to the accompanying drawings. Attached Figure Description

[0059] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments of the invention and, together with their description, serve to explain the principles of the invention.

[0060] Figure 1 This is a flowchart illustrating a smart light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by the present invention.

[0061] Figure 2 This is a schematic diagram of a system for implementing the intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by this invention.

[0062] Figure 3 This is another flowchart illustrating the intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by the present invention.

[0063] Figure 4 This is another flowchart illustrating the intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by this invention.

[0064] Figure 5 This is a flowchart illustrating a process for visually recognizing image data of a target light pole using a neural network architecture.

[0065] Figure 6 This is a flowchart illustrating a process for encrypting and verifying collected data.

[0066] In the diagram: 1. Mobile application client; 2. Multi-mode positioning and calibration module; 3. Visual recognition module; 4. Dynamic verification module; 5. Census execution node; 6. Management and supervision node; 7. Data application node; 8. Cloud database. Detailed Implementation

[0067] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention.

[0068] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.

[0069] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0070] In all the examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0071] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0072] Example 1

[0073] Reference Figure 1 and Figure 2 , Figure 1 This is a flowchart illustrating a smart light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by the present invention. Figure 2 This is a schematic diagram of a system for implementing the intelligent street light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by the present invention. It illustrates a specific embodiment of the intelligent street light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by the present invention, including:

[0074] S101: Distribute the light pole survey task to the application terminal. The light pole survey task includes the target light poles.

[0075] S102: Receive image data and positioning data of the target light pole from the application terminal;

[0076] S103: Correct the positioning data of the target light pole to obtain the corrected coordinates;

[0077] S104: A neural network architecture is used to perform visual recognition on the image data of the target light pole to obtain structured data, including:

[0078] S1041: Preprocess the image data of the target light pole to obtain a preprocessed image;

[0079] S1042: Identify feature information based on the preprocessed image;

[0080] S1043: Map the feature information to standard field representations to obtain recognition data;

[0081] S1044: Preprocess the identification data to obtain preprocessed data, which includes empty fields and non-empty fields;

[0082] S1045: Mark empty fields in the preprocessed data as unrecognized;

[0083] S1046: Convert the marked preprocessed data into structured data according to the preset format;

[0084] S105: The image data, correction coordinates and structured data of the target light pole are combined to form the acquired data. The acquired data is then encrypted to obtain encrypted data, which includes a hash value.

[0085] S106: Upload encrypted data to the blockchain node, which includes census execution node 5, management and supervision node 6, and data application node 7;

[0086] S107: Perform a consistency check on the hash values ​​of the census execution node 5, the management and supervision node 6, and the data application node 7;

[0087] S1071: If the consistency check fails, mark it as abnormal and push it to the application, and wait for the application to re-feed back the image data and positioning data of the target light pole;

[0088] S1072: If the consistency check passes, control management and supervision node 6 decrypts the encrypted data to obtain the collected data, and compares and verifies the structured data with the image data of the target light pole, including:

[0089] S10721: Use a convolutional neural network to extract feature vectors from structured data, use a convolutional neural network to extract feature vectors from the image data of the target light pole, and calculate the cosine similarity between the feature vectors of the structured data and the feature vectors of the image data of the target light pole;

[0090] S107211: If the cosine similarity is less than 90%, the comparison verification fails, an anomaly is marked and pushed to the application, waiting for the application to re-feed back the image data and positioning data of the target light pole;

[0091] S107212: If the cosine similarity is greater than or equal to 90%, the comparison and verification pass. Logical verification is performed on the collected data, including: the preset rule base includes required fields, geographical range and associated verification groups, and the associated groups include projects and the information range corresponding to the projects; the structured data includes data groups, and the data groups include names and the information corresponding to the names;

[0092] S1072121: Determine whether structured data includes all required fields, determine whether the calibration coordinates are within the geographic range, and determine whether, for any set of identical names and items, the information corresponding to the name is within the information range corresponding to the item.

[0093] S10721211: If the structured data includes all required fields, the calibration coordinates are within the geographical range, and for any set of the same name and item, the information corresponding to the name is within the information range corresponding to the item, then the logical verification passes, the collected data and the hash value corresponding to the collected data form a standardized archive, and the standardized archive is updated to the cloud database 8.

[0094] S10721212: If the structured data does not include all required fields, the correction coordinates are not within the geographical range, or there is at least one set of the same name and item, and the information corresponding to the name is not within the information range corresponding to the item, then the logical verification will fail, an anomaly will be marked, and the data will be pushed to the application. The application will then re-feed back the image data and location data of the target light pole.

[0095] It should be noted that in step S101, an application terminal will be issued at least one light pole survey task. The actual number of light pole survey tasks received by the application terminal can be adjusted according to actual needs, and this embodiment does not impose specific limitations on this. In step S102, the target light pole image data returned by the application terminal can be at least 3 images, and this embodiment does not impose specific limitations on this. In step S104, a neural network architecture is used to perform visual recognition on the target light pole image data. The neural network architecture can be a trained YOLOv8 deep learning framework. The YOLOv8 deep learning framework is trained using more than 100,000 light pole image samples covering different materials, heights, and environments. Through transfer learning optimization, a trained YOLOv8 deep learning framework is obtained. The trained YOLOv8 deep learning framework can automatically extract key information such as light pole number, diameter, damage location, damage area, and external equipment type. In step S105, encrypted data can be temporarily stored in a local encrypted storage area. In step S106, the blockchain node is a node in the blockchain consortium chain. In step S10721, the English name for Convolutional Neural Network is Convolutional Neural Network, which can also be abbreviated as CNN network. (Refer to...) Figure 2 The application can be a mobile application client 1; the correction processing of the positioning data of the target light pole can be implemented by the multi-mode positioning and calibration module 2; the visual recognition of the image data of the target light pole can be implemented by the visual recognition module 3; the hash value consistency check of the census execution node 5, the management and supervision node 6 and the data application node 7, the comparison and verification of the structured data and the image data of the target light pole, and the logical verification of the collected data can be implemented by the dynamic verification module 4, but it is not limited to this.

[0096] Understandably, a neural network architecture is used to perform visual recognition on the image data of the target light poles to obtain structured data; the image data, correction coordinates, and structured data of the target light poles are combined to form the collected data; the structured data is compared and verified with the image data of the target light poles; when the comparison and verification pass, the collected data undergoes logical verification; when the logical verification passes, the collected data and its corresponding hash value are combined to form a standardized file, and the standardized file is updated to the cloud database. This achieves automatic identification of the characteristics of the target light poles and the formation of standardized files, which are then uploaded to the cloud server. This can replace the traditional manual acquisition and entry of features, optimize the target light pole survey process, improve the efficiency of the target light pole survey, and effectively reduce errors in the target light pole survey. After obtaining the collected data, it is encrypted. After the encrypted data is uploaded, the hash value consistency is checked by the blockchain nodes. After the consistency check is passed, the data is decrypted to obtain the collected data for subsequent data verification. By verifying the hash value consistency across nodes, the integrity and tamper-proofness of the data in distributed storage can be verified from the bottom layer. The risk of tampering is eliminated from the data storage level, ensuring that the data source for subsequent image feature comparison and logical verification is the original real data. This effectively solves the core problem of untrustworthy data sources in traditional verification.

[0097] Example 2

[0098] Reference Figure 3 , Figure 4 , Figure 5 and Figure 6 , Figure 3 This is another flowchart illustrating the intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by the present invention. Figure 4 This is another flowchart illustrating the intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by the present invention. Figure 5 This is a flowchart illustrating a process for visually recognizing image data of target light poles using a neural network architecture. Figure 6 This is a flowchart illustrating the data encryption and verification process, used to explain another specific embodiment of the intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by this invention, including:

[0099] S201: Distribute the light pole survey task to the application terminal. The light pole survey task includes the target light poles.

[0100] S202: Receive image data and positioning data of the target light pole from the application terminal;

[0101] S203: Correct the positioning data of the target light pole to obtain the corrected coordinates, including: the positioning data of the target light pole includes the coordinates of the ultra-wideband positioning point and the initial coordinates of the wide-area positioning.

[0102] S2031: Obtain a 3D map, which includes feature points and their corresponding coordinates;

[0103] S2032: Based on the coordinates of the ultra-wideband positioning point, generate a preset area in the 3D map, and select three feature points in the preset area;

[0104] S2033: Calculate the relative distance between the coordinates of the ultra-wideband positioning point and the coordinates corresponding to the selected feature point, using the following method:

[0105]

[0106] in, For the coordinates of the ultra-wideband positioning point and the selected first The relative distance between the coordinates of each feature point. The x-coordinate of the ultra-wideband positioning point. For the selected number The x-coordinate of each feature point. The ordinate of the ultra-wideband positioning point is y. For the selected number The ordinate of the coordinates corresponding to each feature point. The vertical coordinate of the ultra-wideband positioning point. For the selected number The vertical coordinates of the coordinates corresponding to each feature point;

[0107] S2034: Calculate the theoretical distance between the initial coordinates of the wide-area positioning and the coordinates corresponding to the selected feature points;

[0108] S2035: Calculate the corrected abscissa, corrected ordinate, and corrected vertical ordinate based on the initial coordinates of wide-area positioning, the coordinates corresponding to the selected feature points, the relative distance between the coordinates of the ultra-wideband positioning point and the coordinates corresponding to the selected feature points, and the theoretical distance between the initial coordinates of wide-area positioning and the coordinates corresponding to the selected feature points.

[0109] S2036: Determine the restricted position on the 3D map based on the corrected abscissa and corrected ordinate, and determine whether the deviation between the terrain elevation corresponding to the restricted position on the 3D map and the corrected ordinate exceeds 0.5 meters;

[0110] S20361: If the distance is less than 0.5 meters, the corrected horizontal coordinate, corrected vertical coordinate, and corrected vertical coordinate will be used to form the corrected coordinates;

[0111] S20362: If the distance exceeds 0.5 meters, the vertical coordinate is updated and corrected using the terrain elevation corresponding to the limited location on the 3D map. The corrected horizontal coordinate, the corrected vertical coordinate, and the updated corrected vertical coordinate constitute the corrected coordinates.

[0112] S204: A neural network architecture is used to perform visual recognition on the image data of the target light pole to obtain structured data, including:

[0113] S2041: Preprocess the image data of the target light pole to obtain a preprocessed image;

[0114] S2042: Identify feature information based on the preprocessed image;

[0115] S2043: Map feature information to standard field representations to obtain identification data;

[0116] S2044: Preprocess the identification data to obtain preprocessed data, which includes empty fields and non-empty fields;

[0117] S2045: Mark empty fields in the preprocessed data as unrecognized;

[0118] S2046: Convert the marked preprocessed data into structured data according to the preset format;

[0119] S205: The image data, correction coordinates and structured data of the target light pole are combined to form the acquired data. The acquired data is then encrypted to obtain encrypted data, which includes a hash value.

[0120] In step S205, the collected data is encrypted to obtain encrypted data, including:

[0121] S2051: Use a random number generator to generate the encryption key;

[0122] S2052: Encrypt the collected data using an encryption key and a symmetric encryption algorithm to obtain ciphertext;

[0123] S2053: Generate a hash value from the ciphertext;

[0124] S2054: The management and supervision node includes a public key, and uses the public key and asymmetric encryption algorithms to encrypt the encryption key to obtain an encrypted packet;

[0125] S2055: Combining ciphertext, hash value, and encrypted packet to form encrypted data;

[0126] S206: Upload encrypted data to blockchain nodes, which include census execution nodes, management and supervision nodes, and data application nodes;

[0127] S207: Perform consistency checks on hash values ​​for census execution nodes, management and supervision nodes, and data application nodes;

[0128] S2071: If the consistency check fails, mark it as abnormal and push it to the application, waiting for the application to re-feed back the image data and positioning data of the target light pole;

[0129] S2072: If the consistency check passes, the control and management supervision node decrypts the encrypted data to obtain the collected data, and compares and verifies the structured data with the image data of the target light pole, including:

[0130] S20721: Use a convolutional neural network to extract feature vectors from the structured data, and use a convolutional neural network to extract feature vectors from the image data of the target light pole. Calculate the cosine similarity between the feature vectors of the structured data and the feature vectors of the image data of the target light pole, using the following method:

[0131] ;

[0132] in, For cosine similarity, For the feature vectors of structured data, For the feature vectors of structured data, This is the vector dot product operation (inner product). This is a multiplication operation (scalar multiplication).

[0133] S207211: If the cosine similarity is less than 90%, the comparison verification fails, an anomaly is marked and pushed to the application, and the application waits for the application to re-feed back the image data and positioning data of the target light pole;

[0134] S207212: If the cosine similarity is greater than or equal to 90%, the comparison and verification pass. Logical verification is performed on the collected data, including: the preset rule base includes required fields, geographical range and associated verification groups, and the associated groups include projects and the information range corresponding to the projects; the structured data includes data groups, and the data groups include names and the information corresponding to the names;

[0135] S2072121: Determine whether structured data includes all required fields, determine whether the calibration coordinates are within the geographic range, and determine whether, for any set of identical names and items, the information corresponding to the name is within the information range corresponding to the item.

[0136] S20721211: If the structured data includes all required fields, the calibration coordinates are within the geographical range, and for any set of identical names and items, the information corresponding to the name is within the information range corresponding to the item, then the logical verification passes, the collected data and the hash value corresponding to the collected data form a standardized archive, and the standardized archive is updated to the cloud database.

[0137] S20721212: If the structured data does not include all required fields, the correction coordinates are not within the geographical range, or there is at least one set of the same name and item, and the information corresponding to the name is not within the information range corresponding to the item, then the logical verification will fail, an anomaly will be marked, and the data will be pushed to the application. The application will then wait for the application to re-feed back the image data and location data of the target light pole.

[0138] It should be noted that in step S203, the positioning data of the target light pole can be obtained by fusing GPS, BeiDou wide-area positioning, and Ultra Wide Band (UWB) positioning. The error of BeiDou wide-area positioning is ±5 meters, and this fusion method can reduce the error to within 3 centimeters. In step S2031, the acquired 3D map can be a city GIS 3D map. Introducing a 3D map allows for terrain matching to eliminate positioning deviations. In step S2032, limiting the selection range of feature points to a preset area is to select feature points around the coordinates of the UWB positioning point. Feature points can be road corners or fixed landmarks, etc. In step S2034, the theoretical distance between the initial coordinates of the wide-area positioning and the coordinates corresponding to the selected feature points is calculated. This can be done by referring to the calculation method in step S2033, where the selected feature points are... The coordinates corresponding to each feature point are replaced with the initial coordinates of the wide-area positioning system. In step S2035, the corrected abscissa is calculated based on the initial coordinates of the wide-area positioning system, the coordinates corresponding to the selected feature points, the relative distance between the coordinates of the ultra-wideband positioning point and the coordinates corresponding to the selected feature points, and the theoretical distance between the initial coordinates of the wide-area positioning system and the coordinates corresponding to the selected feature points. The equation system can be solved using the least squares method, calculated as follows:

[0139]

[0140] in, To correct the x-axis, The x-coordinate of the initial coordinates for wide-area positioning. The total number of feature points selected. For the selected number The x-coordinate of each feature point. For the coordinates of the ultra-wideband positioning point and the selected first The relative distance between the coordinates of each feature point. For wide-area positioning, the initial coordinates and the selected first... The theoretical distance to the coordinates of the selected feature point. Similarly, the corrected ordinate and corrected vertical coordinate can be calculated in the same way as the corrected horizontal coordinate. The x-coordinate of the coordinates corresponding to the first feature point is replaced with the selected first feature point. The corrected ordinate is obtained by taking the ordinate of the coordinates corresponding to the nth feature point. The x-coordinate of the coordinates corresponding to the first feature point is replaced with the selected first feature point. The vertical coordinates of the coordinates corresponding to each feature point are corrected to obtain the vertical coordinates, thus forming a unique mapping relationship between a lamp post and its coordinates.

[0141] Reference Figure 5 In step S2041, the image data of the target light pole is preprocessed, including at least one of the following: noise reduction, enhancement, cropping, and normalization. In step S2043, the identified feature information, taking "metal pole diameter 30 cm" as an example, is mapped to standard field descriptions to obtain "light pole material: metal; light pole diameter: 30 cm," forming identification data. In step S2044, the identification data is preprocessed, which may include data format normalization, unifying units for numerical data, standardizing representation for textual data, and standardizing enumeration values ​​for status data, at least one of the following. In step S2046, the marked preprocessed data is converted into structured data according to a preset format. The preset format may be JSON format conforming to cloud database storage specifications, facilitating the storage of the converted JSON format structured data.

[0142] In step S205, the symmetric encryption algorithm can be AES-256, and the asymmetric encryption algorithm can be RSA2048. In step S2051, the random number generator generates an encryption key based on a true random number algorithm.

[0143] In step S206, each node of the blockchain is authenticated through a digital certificate. When uploading encrypted data to the blockchain node, it is simultaneously uploaded to the cloud database. The application and backend can also be updated in real time. Identification codes are used to avoid duplicate data collection. These codes can be standardized codes specifically generated for the census according to the rule of "city code - area code - road code - light pole number". A one-to-one correspondence can be established between the identification code and the identification number, which is the original physical number of the light pole. The identification code and identification number can be printed simultaneously on the sign on the light pole. Signs are installed on the light poles beforehand, and the image data of the target light pole also includes the sign. The content on the sign can be obtained by using a neural network architecture to perform visual recognition on the image data of the target light pole.

[0144] In step S207, during node data transmission, situations may arise such as data packet loss / tampering, data corruption due to node local storage failure, or malicious nodes manually modifying locally stored data. Subsequent image feature comparison alone cannot resolve the tampering / corruption issues during data transmission and storage. Hash values, however, possess "unique correspondence," meaning even minor modifications to any data will result in a complete change in the hash value. Cross-node hash value consistency verification can verify the integrity and tamper-proof nature of data in distributed storage from the underlying level, serving as a reliable pre-verification for image feature comparison. The combination of these two methods provides dual protection of "data trustworthiness + content accuracy," eliminating the risk of tampering at the data storage level and ensuring that the data source for subsequent image feature comparison and logical verification is the original, authentic data, thus solving the core problem of "untrustworthy data source" in traditional verification. Through distributed storage across nodes, smart contracts pre-set rules for on-chain verification and traceability, ensuring that data is immutable and fully traceable.

[0145] In step S20721, the management and supervision node decrypts the encrypted data to obtain the collected data, including:

[0146] The management and oversight nodes include private keys;

[0147] The management and monitoring node uses its private key to decrypt the encrypted packet and obtain the encryption key.

[0148] The management and monitoring nodes use an encryption key to decrypt the ciphertext and obtain the collected data.

[0149] During decryption, only the management and supervision node can decrypt using its own private key, ensuring the uniqueness and security of decryption permissions.

[0150] In step S2072121, if the item is the height of the light pole, the information range corresponding to the light pole height is 3 meters to 5 meters. If the structured data includes the light pole height, and the information corresponding to the light pole height in the structured data is not within the range of 3 meters to 5 meters, the verification fails. If the item is external power supply, the information range corresponding to external power supply should be the type of external power supply. If the structured data includes external power supply, and the information corresponding to external power supply in the structured data is not the type of external power supply, the verification fails. Similarly, if the item is that the light pole has been removed, the information range corresponding to the light pole being removed should be that device association information cannot be filled in. If the structured data includes the light pole being removed, and any information corresponding to the light pole being removed in the structured data is filled in, the verification fails. The preset rule base can be adjusted according to the actual situation, and this embodiment does not impose specific restrictions on it.

[0151] After step S07, the standardized archives are updated to the cloud database, and can also be automatically updated to the city's smart lighting monitoring system, supporting multi-terminal queries and export in Excel and CSV formats. Model iteration can also be performed, such as optimizing the neural network architecture based on new samples, or adjusting the standard field descriptions and preset rule bases according to the effectiveness of the standardized archives.

[0152] Understandably, the intelligent light pole survey method based on the fusion of visual recognition and blockchain notarization provided by this invention uses a neural network architecture to visually recognize image data of target light poles, replacing traditional manual data entry and reducing on-site operation time by more than 80%. Intelligent task assignment optimizes the workflow, and offline data collection overcomes network limitations, improving overall survey efficiency by more than 50%. The neural network architecture maintains a stable recognition accuracy of over 95%. Combined with a triple verification mechanism of image feature comparison, data logic verification, and cross-node consistency verification, the data accuracy can reach over 98%. The automatic hazard level assessment function further improves the timeliness of risk management. Blockchain consortium blockchain notarization technology ensures that survey data is tamper-proof and traceable throughout the entire process from collection to application, resolving the trust crisis of traditional survey data and providing highly reliable data support for the asset management of light poles. The multi-terminal collaborative architecture enables real-time data synchronization between the application, backend, and cloud databases, and provides visualized monitoring of the census process; seamless offline / online switching ensures data integrity, and blockchain cross-node permission management supports collaborative use by multiple departments; census templates are flexibly configurable, and the neural network architecture supports online iterative optimization, making it suitable for census needs of cities of different sizes and different types of light poles, and can also be extended to other urban infrastructure census scenarios.

[0153] Example 3

[0154] The present invention also provides a specific example of correcting the positioning data of a target light pole to obtain corrected coordinates, including:

[0155] Ultra-wideband positioning point coordinates: =100.0, =200.0, =5.0;

[0156] Wide-area positioning initial coordinates: =102.0, =203.0, =6.0;

[0157] Coordinates of the three feature points: (105, 205, 5.5); (98, 198, 4.8); :(101, 202, 5.2);

[0158] Calculate the relative distance between the coordinates of the ultra-wideband positioning point and the coordinates corresponding to the selected feature point:

[0159] ;

[0160] ;

[0161] ;

[0162] Calculate the theoretical distance between the initial coordinates of the wide-area positioning and the coordinates corresponding to the selected feature points:

[0163] ;

[0164] ;

[0165] ;

[0166] Calculate the corrected x-coordinate, corrected y-coordinate, and corrected vertices:

[0167]

[0168]

[0169]

[0170] The terrain elevation on the 3D map is 6.1m, and the vertical coordinates are corrected. =6.028m, the deviation between the two is less than 0.5m, and the final corrected coordinates (102.048, 203.318, 6.028) are obtained by correcting the abscissa, ordinate and ordinate.

[0171] As can be seen from the above embodiments, the intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by the present invention achieves at least the following beneficial effects:

[0172] 1. The intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage provided by this invention includes: using a neural network architecture to perform visual recognition on the image data of the target light pole to obtain structured data; combining the image data of the target light pole, correction coordinates, and structured data to form collected data; comparing and verifying the structured data with the image data of the target light pole; when the comparison and verification pass, performing logical verification on the collected data; when the logical verification passes, combining the collected data with the hash value corresponding to the collected data to form a standardized file, updating the standardized file to the cloud database, realizing automatic identification of the characteristics of the target light pole and forming a standardized file uploaded to the cloud server, which can replace the traditional manual acquisition of features and manual feature entry, optimize the survey process of the target light pole, improve the survey efficiency of the target light pole, and effectively reduce the errors in the survey of the target light pole.

[0173] 2. The intelligent light pole survey method based on the fusion of visual recognition and blockchain notarization provided by this invention encrypts the collected data after it is obtained. After the encrypted collected data is uploaded, the hash value consistency of the blockchain nodes is checked. After the consistency check is passed, the collected data is decrypted and then used for subsequent data verification. By verifying the hash value consistency across nodes, the integrity and tamper-proofness of the data in the distributed storage can be verified from the bottom layer. The risk of tampering is eliminated from the data notarization level, ensuring that the data source for subsequent image feature comparison and logical verification is the original real data. This effectively solves the core problem of untrustworthy data sources in traditional verification methods.

[0174] While specific embodiments of the invention have been described in detail by way of examples, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of the invention. Those skilled in the art should understand that modifications can be made to the above embodiments without departing from the scope and spirit of the invention. The scope of the invention is defined by the appended claims.

Claims

1. A method for surveying smart light poles based on the fusion of visual recognition and blockchain evidence storage, characterized in that, include: The task of surveying light poles is distributed to the application terminal, and the task of surveying light poles includes target light poles; Receive image data and positioning data of the target light pole from the application terminal; The positioning data of the target light pole is corrected to obtain the corrected coordinates; A neural network architecture is used to perform visual recognition on the image data of the target light pole to obtain structured data. This includes: preprocessing the image data of the target light pole to obtain a preprocessed image; identifying feature information based on the preprocessed image; mapping the feature information to standard field representations to obtain recognition data; preprocessing the recognition data to obtain preprocessed data, which includes empty and non-empty fields; marking the empty fields of the preprocessed data as unrecognized; and converting the marked preprocessed data into the structured data according to a preset format. The image data of the target light pole, the correction coordinates, and the structured data constitute the acquired data. The acquired data is then encrypted to obtain encrypted data, which includes a hash value. The encrypted data is uploaded to a blockchain node, which includes a census execution node, a management and supervision node, and a data application node. Perform a consistency check on the hash values ​​of the census execution node, the management and supervision node, and the data application node. If the consistency check fails, an anomaly is marked and pushed to the application, which then waits for the application to re-feed back the image data and positioning data of the target light pole. If the consistency check passes, the management and supervision node is controlled to decrypt the encrypted data to obtain the collected data. The structured data is then compared and verified with the image data of the target light pole, including: The feature vector of the structured data is extracted using a convolutional neural network, and the feature vector of the image data of the target lamp post is extracted using the same convolutional neural network. The cosine similarity between the feature vector of the structured data and the feature vector of the image data of the target lamp post is then calculated. If the cosine similarity is less than 90%, the comparison verification fails, an anomaly is marked, and the error is pushed to the application. The application then waits for the application to re-feed back the image data and location data of the target light pole. If the cosine similarity is greater than or equal to 90%, the comparison verification passes, and logical verification is performed on the collected data, including: The preset rule base includes required fields, geographical range, and associated verification groups. The associated groups include projects and the information range corresponding to the projects. The structured data includes data groups, and the data groups include names and the information corresponding to the names. Determine whether the structured data includes all the required fields, determine whether the corrected coordinates are within the geographic range, and determine whether, for any set of identical names and items, the information corresponding to the name is within the information range corresponding to the item; If the structured data includes all the required fields, the correction coordinates are within the geographical range, and for any set of identical names and items, the information corresponding to the name is within the information range corresponding to the item, then the logical verification passes, the collected data and the hash value corresponding to the collected data form a standardized archive, and the standardized archive is updated to the cloud database; If the structured data does not include all the required fields, the corrected coordinates are not within the geographical range, or there is at least one set of the same name and item, and the information corresponding to the name is not within the information range corresponding to the item, then the logical verification fails, an anomaly is marked, and the error is pushed to the application, waiting for the application to re-feed back the image data and location data of the target light pole.

2. The intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage as described in claim 1, characterized in that, The step of encrypting the collected data to obtain encrypted data includes: The encryption key is generated using a random number generator; The collected data is encrypted using the aforementioned encryption key and symmetric encryption algorithm to obtain ciphertext; Generate the hash value based on the ciphertext; The management and supervision node includes a public key, and the encryption key is encrypted using the public key and an asymmetric encryption algorithm to obtain an encrypted packet; The ciphertext, the hash value, and the encryption packet constitute the encrypted data.

3. The intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage as described in claim 2, characterized in that, The management and supervision node decrypts the encrypted data to obtain the collected data, including: The management and supervision node includes a private key; The management and supervision node uses the private key to decrypt the encrypted packet to obtain the encryption key; The management and supervision node uses the encryption key to decrypt the ciphertext to obtain the collected data.

4. The intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage as described in claim 1, characterized in that, The cosine similarity between the feature vector of the structured data and the feature vector of the image data of the target lamppost is calculated in the following manner: ； in, The cosine similarity is... The feature vector of the structured data, The feature vector of the structured data, For vector dot product operation, This refers to multiplication operations.

5. The intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage as described in claim 1, characterized in that, The process of correcting the positioning data of the target light pole to obtain corrected coordinates includes: The positioning data of the target light pole includes the coordinates of the ultra-wideband positioning point and the initial coordinates of the wide-area positioning. Obtain the three-dimensional map, which includes feature points and the coordinates corresponding to the feature points; Based on the coordinates of the ultra-wideband positioning point, a preset area is generated in the three-dimensional map, and three feature points are selected in the preset area. Calculate the relative distance between the coordinates of the ultra-wideband positioning point and the coordinates corresponding to the selected feature point; Calculate the theoretical distance between the initial coordinates of the wide-area positioning and the coordinates corresponding to the selected feature point; The corrected horizontal coordinate, corrected vertical coordinate, and corrected vertical coordinate are calculated based on the initial coordinates of the wide-area positioning, the coordinates corresponding to the selected feature point, the relative distance between the coordinates of the ultra-wideband positioning point and the coordinates corresponding to the selected feature point, and the theoretical distance between the initial coordinates of the wide-area positioning and the coordinates corresponding to the selected feature point. Based on the corrected abscissa and the corrected ordinate, a defined position is determined on the 3D map, and it is determined whether the deviation between the terrain elevation corresponding to the defined position on the 3D map and the corrected ordinate exceeds 0.5 meters; If the distance is less than 0.5 meters, the corrected horizontal coordinate, the corrected vertical coordinate, and the corrected vertical coordinate shall constitute the corrected coordinate. If the distance exceeds 0.5 meters, the corrected vertical coordinate is updated using the terrain elevation corresponding to the defined location on the 3D map, and the corrected horizontal coordinate, the corrected vertical coordinate, and the updated corrected vertical coordinate constitute the corrected coordinates.

6. The intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage as described in claim 5, characterized in that, The relative distance between the coordinates of the ultra-wideband positioning point and the coordinates corresponding to the selected feature point is calculated in the following manner: ； in, The coordinates of the ultra-wideband positioning point and the selected first The relative distance between the coordinates of the feature points. Let x be the x-coordinate of the ultra-wideband positioning point. For the selected number The x-coordinate of the coordinates corresponding to each of the aforementioned feature points. Let be the ordinate of the ultra-wideband positioning point. For the selected number The ordinate of the coordinates corresponding to each of the aforementioned feature points. Let be the vertical coordinate of the ultra-wideband positioning point. For the selected number The vertical coordinates of the coordinates corresponding to each of the feature points.

7. The intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage as described in claim 5, characterized in that, The corrected abscissa is calculated based on the initial coordinates of the wide-area positioning, the coordinates corresponding to the selected feature point, the relative distance between the coordinates of the ultra-wideband positioning point and the coordinates corresponding to the selected feature point, and the theoretical distance between the initial coordinates of the wide-area positioning and the coordinates corresponding to the selected feature point, in the following manner: ； in, The corrected abscissa, The x-coordinate of the initial coordinates for the wide-area positioning. The total number of selected feature points. For the selected number The x-coordinate of the coordinates corresponding to each of the aforementioned feature points. The coordinates of the ultra-wideband positioning point and the selected first The relative distance between the coordinates of the feature points. The initial coordinates for the wide-area positioning and the selected first The theoretical distance between the coordinates of the feature points.

8. The intelligent light pole survey method based on the fusion of visual recognition and blockchain evidence storage as described in claim 1, characterized in that, The preprocessing of the image data of the target light pole includes: The image data of the target lamp post is subjected to at least one of the following processing methods: noise reduction, enhancement, cropping, and normalization.

Images