A double-label dynamic compensation electric pole RFID detection method and system

Abstract

The present application relates to the technical field of electric pole detection, and particularly relates to a double-tag dynamic compensation electric pole RFID detection method and system, which specifically comprises the following steps: first, laying RFID tags in the preset area of the cement electric pole base and chuck, adopting eight-segment coding to complete double-tag coupling calibration, and establishing the physical coding correlation of the complete component; second, collecting the multi-frequency point background radio frequency data outside the electric pole root to construct a scene description vector, searching the parameter library or generating adaptive compensation parameters through a deep learning network; third, collecting multi-frequency point signals along multiple paths, extracting features after intensity, phase and frequency point differential compensation, and eliminating abnormal data through double-tag complete verification; and finally, outputting the component existence state and buried depth evaluation results through the model, and triggering active rescan and data fusion in a low confidence scene. The present application can effectively reduce the risk of cross-pole misreading and mismatch, improve the accuracy and robustness of detection in complex soil environments, and realize non-excavation efficient acceptance and inspection of electric pole foundations.

Description

Technical Field

[0001] This invention relates to the field of pole detection technology, and in particular to a dual-tag dynamic compensation RFID detection method and system for poles. Background Technology

[0002] After the construction of cement poles is completed, the base and clamp are usually buried below the backfill layer. On-site acceptance and inspection make it difficult to directly confirm whether they are installed in place and whether the burial depth meets the design requirements through visual inspection. Existing solutions mostly rely on manual experience judgment, partial excavation for verification, or sampling inspection using complex detection equipment such as ground-penetrating radar and ultrasound. These solutions suffer from problems such as low detection efficiency, poor consistency of results, destructive nature to the site, and insufficient digitalization.

[0003] In existing RFID detection solutions, the common practice is to read underground tags and determine the presence of components based on whether the tags are successfully read. These solutions typically only establish a general mapping between tag identifiers and component ledgers, without encoding and modeling the complete physical association between the chassis and chuck under the same pole. This leads to problems such as cross-pole mixed readings, chassis and chuck mismatches, and cross-component identification in actual acceptance scenarios. Furthermore, underground RFID signals are easily affected by factors such as soil medium, moisture conditions, nearby metal components, differences in scanning paths, and frequency response drift, manifesting as RSSI fluctuations, phase zero offset, multi-frequency differential instability, and localized missed readings. If fixed compensation parameters or a unified preprocessing procedure are still used, it is impossible to simultaneously meet the rapid processing needs of conventional scenarios and the adaptive processing needs of unfamiliar scenarios.

[0004] Therefore, this invention proposes a dual-tag dynamic compensation RFID detection method and system for utility poles to solve the above problems. On the one hand, a dual-tag coupled coding calibration model is used to differentiate the deployment and physical coding quantization of tags on the chassis and chuck under the same utility pole. On the other hand, a scene-adaptive dynamic compensation parameter library model is used to combine real-time baseline acquisition of the reference area, parameter library retrieval, and deep learning compensation parameter generation to improve the accuracy, robustness, and repeatability of underground chassis and chuck installation detection in complex scenarios. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a dual-tag dynamic compensation RFID detection method for utility poles. This invention can effectively reduce the risk of cross-pole mixed reading and mismatch, improve the accuracy and robustness of detection in complex soil environments, and realize efficient non-excavation acceptance and inspection of utility pole foundations.

[0006] On the one hand, the technical solution of this invention to solve the technical problem is a dual-tag dynamic compensation RFID detection method for utility poles, which includes the following steps: S1. RFID tags are deployed in the preset areas of the cement pole chassis and chuck, and dual-tag coupling coding calibration is performed. S2. Collect multi-frequency background radio frequency response data in real time in the reference area outside the base of the cement pole, construct a scene description vector, and determine the compensation parameters corresponding to the current scene. S3. Collect raw RFID signal data at multiple detection locations and multiple frequencies along the circumferential, radial, or combined paths at the base of the cement pole. S4. The compensation parameters are used to perform strength, phase and frequency differential compensation on the original RFID signal data. The original RFID signal data is clustered according to the tag ID to form a joint observation set. The component existence feature vector and burial depth evaluation feature vector are extracted. S5. Verify the matching of chassis RFID tags and chuck RFID tags based on dual-tag coupling coding calibration, and eliminate abnormal data; S6. Input the existence characteristics into the existence determination model, output the existence probability and existence status of the chassis and chuck. For components determined to exist, input the burial depth assessment characteristics into the burial depth assessment model, output the burial depth estimate, and combine the design parameters to determine whether the burial depth is qualified. S7. If the probability of the component is in the low confidence interval, or the dual-label verification fails, the scanning parameters are adjusted to trigger an active rescan. The first round of detection and rescan data are merged to obtain the final detection result and generate a detection record.

[0007] S1 is as follows: Assume the RFID tags deployed on the chassis and chuck are respectively and , Each tag uses an eight-segment physical code. Chassis RFID tags chuck RFID tag , Indicates the project section code. Indicates the line segment code. Indicates the main code of the pole number. Indicates a set of group codes. Indicates the component type code; This indicates the component role code used to distinguish the roles of chassis RFID tags and chuck RFID tags in the complete set of components. The chassis RFID tag corresponds to the main component role code, and the chuck RFID tag corresponds to the auxiliary component role code. This indicates a coupled calibration sequence, generated using a preset encoding rule. The preset encoding rule is: [The field is...] , , and Concatenate them sequentially to form the basic encoding string. Then through a fixed-length encoding function base encoding string Mapped to a length of Basic coupling sequence and combined with component type code and component role code For the basic coupling sequence Perform differential transformation to obtain the coupled calibration sequence. ; This represents the checksum, calculated according to preset checksum rules. The preset checksum rules are: for fields... , , , , , and The strings are concatenated sequentially to form the encoded string to be verified, and then processed according to a preset cyclic redundancy check function. The calculated length is Verification code ; RFID tags on the chassis of the same cement pole With chuck RFID tags satisfy , , , and and It is a complementary encoding.

[0008] S2 is as follows: Multi-frequency background radio frequency response data are collected in real time in the reference area outside the base of the cement pole, a scene description vector is constructed, and the compensation parameters corresponding to the current scene are determined. Multi-frequency background RF response data includes background received signal strength, background phase, background phase shift between adjacent frequency points, and background noise floor. Let the total number of frequency points be... ,use An index representing the total number of frequency points; a scene description vector composed of background RF response data collected from multiple frequency points. , means as follows: , in, , and They represent the first Background signal strength, background phase, and background noise floor at each frequency point This represents the amount of background phase shift between adjacent frequency points. Indicates scene category encoding; Based on the scene description vector The matching parameters are retrieved from the scene adaptive dynamic compensation parameter library to determine the corresponding compensation parameters. The scene adaptive dynamic compensation parameter library consists of historical scene description vectors. Each historical description vector has a corresponding compensation parameter. The similarity between the current scene description vector and the historical scene description vector is calculated. A scene similarity threshold is set. When the similarity is greater than or equal to the scene similarity threshold, the compensation parameter corresponding to the historical scene description vector is selected as the compensation parameter of the current scene description vector.

[0009] When the similarity is less than the scene similarity threshold, and there is no matching compensation parameter in the scene adaptive dynamic compensation parameter library, the deep learning compensation parameter generation network is called to generate compensation parameters. The deep learning compensation parameter generation network is a multilayer perceptron network, including an input layer, two fully connected hidden layers and an output layer. The hidden layers are set with normalization and non-linear activation functions. The network uses offline sample sets train, Indicates the first A historical scene description vector The optimal compensation parameters are obtained by solving the following objective function: , Among them, compensation parameters This represents the compensation parameter variable to be optimized in the objective function; Indicates compensation parameters To optimize variables, find the optimal solution that minimizes the objective function. Values; Indicates the use of compensation parameters The compensated signal strength fluctuation, Indicates the use of compensation parameters Compensated frequency-point differential phase fluctuation, Indicates the decoding success rate. Indicates the number of missed reads. This represents the squared L2 norm of the compensation parameter vector. to Indicates the balance coefficient; Training loss of deep learning compensated parameter generation network The calculation formula is as follows: , in, This indicates the compensation parameters generated by the deep learning compensation parameter generation network.

[0010] The specific compensation operation for the original RFID signal data in S4 is as follows: The raw RFID signal data consists of multiple raw RFID signal data frames. Each raw RFID signal data frame contains the tag ID, received signal strength, phase value, timestamp, frequency identifier, and decoding status flag. Strength, phase, and frequency differential compensation are performed on the raw RFID signal data according to compensation parameters, which are expressed as follows: , means as follows: , in, and They represent the first Background signal strength compensation and background phase compensation at each frequency point; , Indicates the first Phase drift compensation amount for adjacent frequency points; The raw signal strength and phase of the original RFID signal data at each detection point and frequency are respectively expressed as follows: and , The index representing the detection position number. The index representing the frequency point number, and the characteristics of the compensated signal strength, phase, and phase difference between adjacent frequency points are as follows: , , , in, This indicates the phase expansion operation. , and They represent the first The first detection location, the first The signal strength, phase, and phase difference characteristics of adjacent frequency points at each measurement point.

[0011] S4 is as follows: Clustering is performed based on the tag ID of the original RFID signal data. Compensated observations corresponding to the same tag ID at multiple detection locations and multiple frequency points are merged to form a joint observation set for the tag. The joint observation set includes the retrieval location index, frequency point index, compensated signal strength, compensated phase, compensated adjacent frequency point phase difference characteristics, and decoding status marker. The tag ID uniquely corresponds to one component, which is either a chassis or a chuck. Extracting existence feature vectors: The average signal strength and signal strength fluctuation after compensation are calculated based on the compensated signal strength. The mean phase difference and phase difference fluctuation of multiple frequencies are calculated based on the compensated frequency differential phase. Based on the distribution relationship of the compensated observations at different detection locations and frequencies, the position consistency index and frequency consistency index are calculated. The decoding success rate is statistically analyzed based on the decoding status markers, thus constructing an existence feature vector. ; Extracting feature vectors for burial depth assessment: Calculate the compensated average phase based on the compensated phase in the joint observation set. Based on the phase difference characteristics of adjacent frequency points after compensation Calculate the mean phase difference at multiple frequency points and multi-frequency phase difference fluctuation The slope of phase-frequency change is calculated based on the relationship between the changes in the mean phase after compensation at different frequency points. The intensity-frequency attenuation slope is calculated based on the relationship between the mean signal strength after compensation at different frequency points. This, in turn, constitutes the feature vector for burial depth assessment. , .

[0012] S5 is detailed below: Based on the dual-tag coupled coding calibration in step S1, the matching accuracy of the dual-tag sets is determined by the coupling coding consistency score. The consistency score is calculated using the following formula: , in, Indicates the consistency score. Indicates an indicator function, Indicates chassis RFID tag With chuck RFID tags The similarity function of the coupled calibration sequences, to Indicates the weighting coefficient. Represents the XOR operation; Preset set threshold ,when At that time, the chassis RFID tag is determined. With chuck RFID tags The components must be legally assembled and belong to the same cement pole; otherwise, they will be judged as incorrectly installed, misassembled, or mixed across poles.

[0013] S6 is detailed below: (1) Existence determination model: Existence eigenvectors The probability of a component's existence is calculated using the input logistic regression existence determination model, as shown in the following formula: , in, This indicates the probability of a component existing. Represents the model weight vector. Indicates the bias term; Preset existence probability threshold ,when Greater than or equal to When, determine that the corresponding component exists, when Less than The system determines that the corresponding component was not detected. (2) Burial depth assessment model: Use support vector regression or multilayer perceptron regression model; The epsilon-support vector regression model is used to evaluate the feature vectors of burial depth. Input model, output burial depth estimate ; Then according to the design burial depth and allowable deviation Output the project conclusions according to the following rules: when At that time, the depth was determined to be insufficient; when At that time, the depth was deemed acceptable; when When the depth exceeds the limit, it is determined that the depth has exceeded the limit. When the model's input features exceed the training distribution or the output credibility is lower than a preset threshold, the depth result is deemed unreliable.

[0014] S7 is detailed below: When the probability exists within the low confidence interval Internal, or complete set consistency score Below the complete set threshold At that time, an active rescan is triggered; During rescanning, the scaling factor is used. , and The number of interrogation rounds, single-point dwell time, and frequency set size are adjusted separately. Then, the burial depth results of the initial detection and the rescan detection are fused together. The calculation formula is as follows: , , in, and These represent the estimated burial depth values ​​obtained from the initial detection and the re-detection, respectively. and These represent the probability of the corresponding detection or its confidence level, respectively. This indicates taking the maximum value. This is the final estimated burial depth after fusion. To determine the final credibility of the test.

[0015] On the other hand, the present invention also provides a dual-tag dynamic compensation RFID detection system for utility poles, including a module for executing processing instructions for each step in a dual-tag dynamic compensation RFID detection method for utility poles, as follows: Dual-tag calibration and deployment module: used to deploy RFID tags in the preset area of ​​cement pole chassis and chuck and perform dual-tag coupling coding calibration. It adopts an eight-segment physical coding structure and generates a coupling calibration sequence through XOR transformation. Scene perception and compensation generation module: used to collect multi-frequency background radio frequency response data in the reference area outside the pole root, construct scene description vector; retrieve matching compensation parameters through scene adaptive dynamic compensation parameter library, and call multilayer perceptron network to generate current scene compensation parameters when no matching scene is found; Multi-path signal acquisition module: Used to acquire raw RFID signal data, including tag ID, signal strength, and phase, at multiple detection locations and multiple frequency points along the circumferential, radial, or combined paths at the base of the pole. Signal compensation and feature extraction module: Used to perform intensity, phase, and frequency differential compensation on the original data using compensation parameters, cluster by tag ID to form a joint observation set, and extract feature vectors for component existence and burial depth assessment. Dual-tag coupling verification module: used to calculate the complete set consistency score based on dual-tag coupling coding, verify the matching of dual tags, and eliminate abnormal data such as mis-installed, cross-component, and mixed reading across poles; The detection and evaluation module is used to input the existence characteristics into the judgment model and output the existence status of the component. For the components that are determined to exist, the module inputs the burial depth evaluation model and outputs the estimated burial depth value. The module combines the design parameters to determine the burial depth qualification. Active rescan fusion module: When the probability of existence is in the low confidence interval or the dual-label verification fails, the scanning parameters are adjusted to trigger a rescan, and the first and rescan data are fused to obtain the final detection result and generate a record.

[0016] The effects described in the invention are merely those of the embodiments, and not all the effects of the invention. The above technical solutions have the following advantages or beneficial effects: This invention discloses a dual-tag dynamic compensation RFID detection method and system for utility poles. Through an eight-segment dual-tag coupling coding calibration mechanism, the correspondence between the chassis and chuck tags is moved from a general ledger mapping to the physical coding level. Relying on multi-layered constraints of engineering section codes, complete set codes, component role codes, and coupled calibration sequences, combined with complete set consistency score verification rules, it can accurately identify abnormal working conditions such as cross-pole mixed readings and misinstalled components, significantly improving the accuracy of complete set component identification and fundamentally avoiding the risk of misidentification caused by single-tag detection. A scene-adaptive dynamic compensation system is constructed, forming a three-level processing mechanism: real-time background acquisition of the reference area, rapid retrieval of historical parameter databases, and online generation by a deep learning network. This ensures detection efficiency in conventional scenarios while possessing adaptive adaptability to unfamiliar and complex working conditions. Through triple redirection compensation of signal strength, phase, and frequency difference, it effectively offsets radio frequency interference caused by soil moisture content, adjacent metal components, and frequency drift, significantly improving the stability and repeatability of detection data.

[0017] This invention employs a multi-path, multi-frequency joint observation and feature-based hierarchical evaluation architecture. Based on the compensated net response data, it extracts existence features and burial depth assessment features, and combines them with logistic regression judgment and support vector regression estimation algorithms to achieve hierarchical output of qualitative judgment on component existence and quantitative assessment of burial depth. The detection conclusions are more aligned with the actual business needs of engineering acceptance. Furthermore, by setting up an active rescanning and data fusion closed-loop mechanism, it automatically adjusts scanning parameters for re-inspection in low-confidence intervals and scenarios where complete set verification fails. The final conclusion is output through weighted fusion, balancing detection efficiency and result reliability. Simultaneously, it generates structured detection records to support engineering quality traceability and digital operation and maintenance management. The overall solution requires no excavation, is easy to operate, and can be widely adapted to the construction acceptance and periodic inspection scenarios of power line pole foundations. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0019] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0020] To clearly illustrate the technical features of this solution, the invention will be described in detail below through specific implementation methods and in conjunction with the accompanying drawings.

[0021] Example 1 like Figure 1 As shown, a dual-tag dynamic compensation RFID detection method for utility poles includes the following steps: S1. RFID tags are deployed in the preset areas of the cement pole chassis and chuck, and dual-tag coupling coding calibration is performed. S2. Collect multi-frequency background radio frequency response data in real time in the reference area outside the base of the cement pole, construct a scene description vector, and determine the compensation parameters corresponding to the current scene. S3. Collect raw RFID signal data at multiple detection locations and multiple frequencies along the circumferential, radial, or combined paths at the base of the cement pole. S4. The compensation parameters are used to perform strength, phase and frequency differential compensation on the original RFID signal data. The original RFID signal data is clustered according to the tag ID to form a joint observation set. The component existence feature vector and burial depth evaluation feature vector are extracted. S5. Verify the matching of chassis RFID tags and chuck RFID tags based on dual-tag coupling coding calibration, and eliminate abnormal data; S6. Input the existence characteristics into the existence determination model, output the existence probability and existence status of the chassis and chuck. For components determined to exist, input the burial depth assessment characteristics into the burial depth assessment model, output the burial depth estimate, and combine the design parameters to determine whether the burial depth is qualified. S7. If the probability of the component is in the low confidence interval, or the dual-label verification fails, the scanning parameters are adjusted to trigger an active rescan. The first round of detection and rescan data are merged to obtain the final detection result and generate a detection record.

[0022] In a specific implementation, S1 is as follows: Assume the RFID tags deployed on the chassis and chuck are respectively and , Each tag uses an eight-segment physical code. Chassis RFID tags chuck RFID tag , Indicates the project section code. Indicates the line segment code. Indicates the main code of the pole number. Indicates a set of group codes. Indicates the component type code; This indicates the component role code used to distinguish the roles of chassis RFID tags and chuck RFID tags in the complete set of components. The chassis RFID tag corresponds to the main component role code, and the chuck RFID tag corresponds to the auxiliary component role code. This indicates a coupled calibration sequence, generated using a preset encoding rule. The preset encoding rule is: [The field is...] , , and Concatenate them sequentially to form the basic encoding string. Then through a fixed-length encoding function base encoding string Mapped to a length of Basic coupling sequence and combined with component type code and component role code For the basic coupling sequence Perform differential transformation to obtain the coupled calibration sequence. The calculation formula is as follows: , in, This represents the XOR operation. Indicated by component type code The corresponding type mask, Indicates the component role code Corresponding character mask; This represents the checksum, calculated according to preset checksum rules. The preset checksum rules are: for fields... , , , , , and The strings are concatenated sequentially to form the encoded string to be verified, and then processed according to a preset cyclic redundancy check function. The calculated length is Verification code The calculation formula is as follows: , in, This represents a string concatenation operation. Indicates the output length is Cyclic redundancy check function; RFID tags on the chassis of the same cement pole With chuck RFID tags satisfy , , , and and The system uses complementary coding. Therefore, during the detection phase, the system can not only identify "which tag was read," but also determine "whether the chassis tag and the chuck tag belong to a legitimate assembly of components for the same pole."

[0023] In a specific implementation, S2 is as follows: Multi-frequency background radio frequency response data are collected in real time in the reference area outside the base of the cement pole, a scene description vector is constructed, and the compensation parameters corresponding to the current scene are determined. Multi-frequency background RF response data includes background received signal strength, background phase, background phase shift between adjacent frequency points, and background noise floor. Let the total number of frequency points be... ,use An index representing the total number of frequency points; a scene description vector composed of background RF response data collected from multiple frequency points. , means as follows: , in, , and They represent the first Background signal strength, background phase, and background noise floor at each frequency point This represents the amount of background phase shift between adjacent frequency points. Indicates scene category encoding; Based on the scene description vector The matching parameters are retrieved from the scene adaptive dynamic compensation parameter library to determine the corresponding compensation parameters. The scene adaptive dynamic compensation parameter library consists of historical scene description vectors. Each historical description vector has a corresponding compensation parameter. The similarity between the current scene description vector and the historical scene description vector is calculated. A scene similarity threshold is set. When the similarity is greater than or equal to the scene similarity threshold, the compensation parameter corresponding to the historical scene description vector is selected as the compensation parameter of the current scene description vector.

[0024] In a specific implementation, when the similarity is less than the scene similarity threshold, if there is no matching compensation parameter in the scene adaptive dynamic compensation parameter library, then the deep learning compensation parameter generation network is called to generate compensation parameters. The deep learning compensation parameter generation network is a multilayer perceptron network, including an input layer, two fully connected hidden layers and an output layer. The hidden layers are set with normalization and non-linear activation functions. The network uses offline sample sets train, Indicates the first A historical scene description vector The optimal compensation parameters are obtained by solving the following objective function: , Among them, compensation parameters The compensation parameter variable in the objective function is a parameter vector consisting of the signal strength compensation amount at each frequency point, the phase compensation amount at each frequency point, and the differential compensation amount at adjacent frequency points. Indicates compensation parameters To optimize variables, find the optimal solution that minimizes the objective function. Values; Indicates the use of compensation parameters The compensated signal strength fluctuation, Indicates the use of compensation parameters Compensated frequency-point differential phase fluctuation, Indicates the decoding success rate. Indicates the number of missed reads. This represents the squared L2 norm of the compensation parameter vector. to Indicates the balance coefficient; Training loss of deep learning compensated parameter generation network The calculation formula is as follows: , in, This indicates the compensation parameters generated by the deep learning compensation parameter generation network.

[0025] In a specific implementation, the compensation operation for the original RFID signal data in S4 is as follows: The raw RFID signal data consists of multiple raw RFID signal data frames. Each raw RFID signal data frame contains the tag ID, received signal strength, phase value, timestamp, frequency identifier, and decoding status flag. Strength, phase, and frequency differential compensation are performed on the raw RFID signal data according to compensation parameters, which are expressed as follows: , means as follows: , in, and They represent the first Background signal strength compensation and background phase compensation at each frequency point; , Indicates the first Phase drift compensation amount for adjacent frequency points; The raw signal strength and phase of the original RFID signal data at each detection point and frequency are respectively expressed as follows: and , The index representing the detection position number. The index representing the frequency point number, and the characteristics of the compensated signal strength, phase, and phase difference between adjacent frequency points are as follows: , , , in, This indicates the phase expansion operation. , and They represent the first The first detection location, the first The signal strength, phase, and phase difference characteristics of adjacent frequency points at each measurement point.

[0026] In a specific implementation, S4 is as follows: Clustering is performed based on the tag ID of the original RFID signal data. Compensated observations corresponding to the same tag ID at multiple detection locations and multiple frequency points are merged to form a joint observation set for the tag. The joint observation set includes the retrieval location index, frequency point index, compensated signal strength, compensated phase, compensated adjacent frequency point phase difference characteristics, and decoding status marker. The tag ID uniquely corresponds to one component, which is either a chassis or a chuck. Extracting existence feature vectors: The existence feature vector is used to determine whether the component corresponding to the target tag exists. It mainly characterizes the readability, signal strength level, and reading stability of the target tag during multi-location and multi-frequency scanning. Based on the compensated signal strength, the average signal strength and signal strength fluctuation after compensation are calculated. Based on the compensated frequency differential phase, the mean phase difference and phase difference fluctuation of multiple frequencies are calculated. Based on the distribution relationship of the compensated observations at different detection locations and frequencies, the position consistency index and frequency consistency index are calculated. Based on the decoding status marker, the decoding success rate is statistically analyzed, thus forming the existence feature vector. ; Extracting feature vectors for burial depth assessment: The burial depth assessment feature vector is used to estimate the burial depth of identified target components, mainly characterizing the compensated phase response, frequency differential phase response, and signal attenuation with frequency. The compensated average phase is calculated based on the compensated phase from the joint observation set. Based on the phase difference characteristics of adjacent frequency points after compensation Calculate the mean phase difference at multiple frequency points and multi-frequency phase difference fluctuation The slope of phase-frequency change is calculated based on the relationship between the changes in the mean phase after compensation at different frequency points. The intensity-frequency attenuation slope is calculated based on the relationship between the mean signal strength after compensation at different frequency points. This, in turn, constitutes the feature vector for burial depth assessment. , .

[0027] In a specific implementation, S5 is as follows: Based on the dual-tag coupled coding calibration in step S1, the matching accuracy of the dual-tag sets is determined by the coupling coding consistency score. The consistency score is calculated using the following formula: , in, Indicates the consistency score. Indicates an indicator function, Indicates chassis RFID tag With chuck RFID tags The similarity function of the coupled calibration sequences, to Indicates the weighting coefficient. Represents the XOR operation; The calculation is as follows: , in, Indicates the chassis coupling calibration sequence Coupled calibration sequence with chuck Hamming distance between them Indicates the length of the coupling calibration sequence; Preset set threshold ,when At that time, the chassis RFID tag is determined. With chuck RFID tags The components must be legally assembled and belong to the same cement pole; otherwise, they will be judged as incorrectly installed, misassembled, or mixed across poles.

[0028] In a specific implementation, S6 is as follows: (1) Existence determination model: Existence eigenvectors The probability of a component's existence is calculated using the input logistic regression existence determination model, as shown in the following formula: , in, This indicates the probability of a component existing. Represents the model weight vector. Indicates the bias term; Preset existence probability threshold ,when Greater than or equal to When, determine that the corresponding component exists, when Less than The system determines that the corresponding component was not detected. (2) Burial depth assessment model: Use support vector regression or multilayer perceptron regression model; The epsilon-support vector regression model is used to evaluate the feature vectors of burial depth. Input model, output burial depth estimate ; Then according to the design burial depth and allowable deviation Output the project conclusions according to the following rules: when At that time, the depth was determined to be insufficient; when At that time, the depth was deemed acceptable; when When the depth exceeds the limit, it is determined that the depth has exceeded the limit. When the model's input features exceed the training distribution or the output credibility is lower than a preset threshold, the depth result is deemed unreliable.

[0029] In a specific implementation, when the probability exists in the low confidence interval Internal, or complete set consistency score Below the complete set threshold At that time, an active rescan is triggered; During rescanning, the scaling factor is used. , and The number of interrogation rounds, single-point dwell time, and frequency set size are adjusted separately. Then, the burial depth results of the initial detection and the rescan detection are fused together. The calculation formula is as follows: , , in, and These represent the estimated burial depth values ​​obtained from the initial detection and the re-detection, respectively. and These represent the probability of the corresponding detection or its confidence level, respectively. This indicates taking the maximum value. This is the final estimated burial depth after fusion. To determine the final credibility of the test.

[0030] Example 2 A dual-tag dynamic compensation RFID detection system for utility poles, comprising a dual-tag dynamic compensation RFID detection method, including: Dual-tag calibration and deployment module: used to deploy RFID tags in the preset area of ​​cement pole chassis and chuck and perform dual-tag coupling coding calibration. It adopts an eight-segment physical coding structure and generates a coupling calibration sequence through XOR transformation. Scene perception and compensation generation module: used to collect multi-frequency background radio frequency response data in the reference area outside the pole root, construct scene description vector; retrieve matching compensation parameters through scene adaptive dynamic compensation parameter library, and call multilayer perceptron network to generate current scene compensation parameters when no matching scene is found; Multi-path signal acquisition module: Used to acquire raw RFID signal data, including tag ID, signal strength, and phase, at multiple detection locations and multiple frequency points along the circumferential, radial, or combined paths at the base of the pole. Signal compensation and feature extraction module: Used to perform intensity, phase, and frequency differential compensation on the original data using compensation parameters, cluster by tag ID to form a joint observation set, and extract feature vectors for component existence and burial depth assessment. Dual-tag coupling verification module: used to calculate the complete set consistency score based on dual-tag coupling coding, verify the matching of dual tags, and eliminate abnormal data such as mis-installed, cross-component, and mixed reading across poles; The detection and evaluation module is used to input the existence characteristics into the judgment model and output the existence status of the component. For the components that are determined to exist, the module inputs the burial depth evaluation model and outputs the estimated burial depth value. The module combines the design parameters to determine the burial depth qualification. Active rescan fusion module: When the probability of existence is in the low confidence interval or the dual-label verification fails, the scanning parameters are adjusted to trigger a rescan, and the first and rescan data are fused to obtain the final detection result and generate a record.

[0031] Example 3 This invention's method is applied to the completion and acceptance of newly constructed power distribution lines. During the completion and acceptance phase of the power line project, after the construction unit completes the pole erection, chassis installation, clamp installation, and earthwork backfilling, the acceptance personnel arrive at the foundation location of a single pole with a wireless detection device. The system first reads the encoding information of the chassis tag and the clamp tag, and determines whether the two tags belong to the same legitimate set of components for the same pole based on the dual-tag coupled encoding calibration model.

[0032] Subsequently, the acceptance personnel performed real-time background data acquisition in the reference area outside the pole base. The system constructed a current scene description vector and prioritized searching for directly callable compensation parameters in the parameter library corresponding to the scene adaptive dynamic compensation parameter library model. If a sufficiently similar historical scene exists in the parameter library, its compensation parameters are directly called; otherwise, the deep learning compensation parameter generation network is invoked to output the current scene compensation parameters.

[0033] After obtaining the compensation parameters, the system collects raw RFID data frames of the target areas of the chassis and chuck along a preset scanning path at multiple detection locations and multiple frequency points, performs dynamic compensation on them, and extracts existence features and burial depth assessment features. If the existence probability of the chassis tag is higher than the threshold and the burial depth estimate meets the design requirements, the system outputs "Chassis installed in place"; if the existence probability of the chuck tag is lower than the threshold, the system outputs "Chuck not detected"; if both the chassis tag and the chuck tag are detected but their dual-tag set identification score is lower than the set threshold, the system outputs "Suspected cross-contamination or misinstallation".

[0034] For cases where the probability is in the low confidence interval, the system automatically performs an active rescan, fusing the initial and rescan results and writing the final inspection conclusion into a structured inspection record. This record includes the pole number, chassis label code, chuck label code, complete set identification results, called or generated compensation parameter numbers, component existence status, burial depth assessment conclusion, abnormal status, and re-inspection recommendations, used for project acceptance archiving and subsequent quality traceability.

[0035] Although the specific embodiments of the invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the invention. Based on the technical solutions of the invention, various modifications or variations that can be made by those skilled in the art without creative effort are still within the scope of protection of the invention.

Claims

1. A dual-tag dynamic compensation RFID detection method for utility poles, characterized in that, Includes the following steps: S1. RFID tags are deployed in the preset areas of the cement pole chassis and chuck, and dual-tag coupling coding calibration is performed. S2. Collect multi-frequency background radio frequency response data in real time in the reference area outside the base of the cement pole, construct a scene description vector, and determine the compensation parameters corresponding to the current scene. S3. Collect raw RFID signal data at multiple detection locations and multiple frequencies along the circumferential, radial, or combined paths at the base of the cement pole. S4. The compensation parameters are used to perform strength, phase and frequency differential compensation on the original RFID signal data. The original RFID signal data is clustered according to the tag ID to form a joint observation set. The component existence feature vector and burial depth evaluation feature vector are extracted. S5. Verify the matching of chassis RFID tags and chuck RFID tags based on dual-tag coupling coding calibration, and eliminate abnormal data; S6. Input the existence characteristics into the existence determination model, output the existence probability and existence status of the chassis and chuck. For components determined to exist, input the burial depth assessment characteristics into the burial depth assessment model, output the burial depth estimate, and combine the design parameters to determine whether the burial depth is qualified. S7. If the probability of the component is in the low confidence interval, or the dual-label verification fails, the scanning parameters are adjusted to trigger an active rescan. The first round of detection and rescan data are merged to obtain the final detection result and generate a detection record.

2. The dual-tag dynamic compensation RFID detection method for utility poles according to claim 1, characterized in that, S1 is as follows: Assume the RFID tags deployed on the chassis and chuck are respectively and , Each tag uses an eight-segment physical code. Chassis RFID tags CLASS RFID tag , Indicates the project section code. Indicates the line segment code. Indicates the main code of the pole number. Indicates a set of group codes. Indicates the component type code; This indicates the component role code used to distinguish the roles of chassis RFID tags and chuck RFID tags in the complete set of components. The chassis RFID tag corresponds to the main component role code, and the chuck RFID tag corresponds to the auxiliary component role code. This indicates a coupled calibration sequence, generated using a preset encoding rule. The preset encoding rule is: [The field is...] , , and Concatenate them sequentially to form the basic encoding string. Then through a fixed-length encoding function base encoding string Mapped to a length of Basic coupling sequence and combined with component type code and component role code For the basic coupling sequence Perform differential transformation to obtain the coupled calibration sequence. ; This represents the checksum, calculated according to preset checksum rules. The preset checksum rules are: for fields... , , , , , and The strings are concatenated sequentially to form the encoded string to be verified, and then processed according to a preset cyclic redundancy check function. The calculated length is Verification code ; RFID tags on the chassis of the same cement pole With chuck RFID tags satisfy , , , and and It is a complementary encoding.

3. The dual-tag dynamic compensation RFID detection method for utility poles according to claim 1, characterized in that, S2 is as follows: Multi-frequency background radio frequency response data are collected in real time in the reference area outside the base of the cement pole, a scene description vector is constructed, and the compensation parameters corresponding to the current scene are determined. Multi-frequency background RF response data includes background received signal strength, background phase, background phase shift between adjacent frequency points, and background noise floor. Let the total number of frequency points be... ,use An index representing the total number of frequency points; a scene description vector composed of background RF response data collected from multiple frequency points. , means as follows: ， in, , and They represent the first Background signal strength, background phase, and background noise floor at each frequency point This represents the amount of background phase shift between adjacent frequency points. Indicates scene category encoding; Based on the scene description vector The matching parameters are retrieved from the scene adaptive dynamic compensation parameter library to determine the corresponding compensation parameters. The scene adaptive dynamic compensation parameter library consists of historical scene description vectors. Each historical description vector has a corresponding compensation parameter. The similarity between the current scene description vector and the historical scene description vector is calculated. A scene similarity threshold is set. When the similarity is greater than or equal to the scene similarity threshold, the compensation parameter corresponding to the historical scene description vector is selected as the compensation parameter of the current scene description vector.

4. The RFID detection method for dual-tag dynamic compensation utility poles according to claim 3, characterized in that: When the similarity is less than the scene similarity threshold, and there is no matching compensation parameter in the scene adaptive dynamic compensation parameter library, the deep learning compensation parameter generation network is called to generate compensation parameters. The deep learning compensation parameter generation network is a multilayer perceptron network, including an input layer, two fully connected hidden layers and an output layer. The hidden layers are set with normalization and non-linear activation functions. The network uses offline sample sets train, Indicates the first A historical scene description vector The optimal compensation parameters are obtained by solving the following objective function: ， Among them, compensation parameters This represents the compensation parameter variable to be optimized in the objective function; Indicates compensation parameters To optimize variables, find the optimal solution that minimizes the objective function. Values; Indicates the use of compensation parameters The compensated signal strength fluctuation, Indicates the use of compensation parameters Compensated frequency-point differential phase fluctuation, Indicates the decoding success rate. Indicates the number of missed reads. This represents the squared L2 norm of the compensation parameter vector. to Indicates the balance coefficient; Training loss of deep learning compensated parameter generation network The calculation formula is as follows: ， in, This indicates the compensation parameters generated by the deep learning compensation parameter generation network.

5. The RFID detection method for dual-tag dynamic compensation utility poles according to claim 1, characterized in that, The specific compensation operation for the original RFID signal data in S4 is as follows: The raw RFID signal data consists of multiple raw RFID signal data frames. Each raw RFID signal data frame contains the tag ID, received signal strength, phase value, timestamp, frequency identifier, and decoding status flag. Strength, phase, and frequency differential compensation are performed on the raw RFID signal data according to compensation parameters, which are expressed as follows: , means as follows: ， in, and They represent the first Background signal strength compensation and background phase compensation at each frequency point; , Indicates the first Phase drift compensation amount for adjacent frequency points; The raw signal strength and phase of the original RFID signal data at each detection point and frequency are respectively expressed as follows: and , The index representing the detection position number. The index representing the frequency point number, and the characteristics of the compensated signal strength, phase, and phase difference between adjacent frequency points are as follows: ， ， ， in, This indicates the phase expansion operation. , and They represent the first The first detection location, the first The signal strength, phase, and phase difference characteristics of adjacent frequency points at each measurement point.

6. The RFID detection method for dual-tag dynamic compensation utility poles according to claim 1, characterized in that, S4 is as follows: Clustering is performed based on the tag ID of the original RFID signal data. Compensated observations corresponding to the same tag ID at multiple detection locations and multiple frequency points are merged to form a joint observation set for the tag. The joint observation set includes the retrieval location index, frequency point index, compensated signal strength, compensated phase, compensated adjacent frequency point phase difference characteristics, and decoding status marker. The tag ID uniquely corresponds to one component, which is either a chassis or a chuck. Extracting existence feature vectors: Based on the compensated signal strength, the average signal strength and signal strength fluctuation after compensation are calculated. Based on the compensated frequency differential phase, the mean phase difference and phase difference fluctuation of multiple frequencies are calculated. Based on the distribution relationship of the compensated observations at different detection locations and frequencies, the position consistency index and frequency consistency index are calculated. Based on the decoding status marker, the decoding success rate is statistically analyzed, thus constructing an existence feature vector. ; Extracting feature vectors for burial depth assessment: Calculate the compensated average phase based on the compensated phase in the joint observation set. Based on the phase difference characteristics of adjacent frequency points after compensation Calculate the mean phase difference at multiple frequency points and multi-frequency phase difference fluctuation The slope of phase-frequency change is calculated based on the relationship between the changes in the mean phase after compensation at different frequency points. The intensity-frequency attenuation slope is calculated based on the relationship between the mean signal strength after compensation at different frequency points. This, in turn, constitutes the feature vector for burial depth assessment. , .

7. The dual-tag dynamic compensation RFID detection method for utility poles according to claim 2, characterized in that, S5 is detailed below: Based on the dual-tag coupled coding calibration in step S1, the matching accuracy of the dual-tag sets is determined by the coupling coding consistency score. The consistency score is calculated using the following formula: ， in, Indicates the consistency score. Indicates an indicator function, Indicates chassis RFID tag With chuck RFID tags The similarity function of the coupled calibration sequences, to Indicates the weighting coefficient. This represents the XOR operation; Preset set threshold ,when At that time, the chassis RFID tag is determined. With chuck RFID tags The components must be legally assembled and belong to the same cement pole; otherwise, they will be judged as incorrectly installed, misassembled, or mixed across poles.

8. The RFID detection method for dual-tag dynamic compensation utility poles according to claim 1, characterized in that, S6 is detailed below: (1) Existence determination model: Existence eigenvectors The probability of a component's existence is calculated using the input logistic regression existence determination model, as shown in the following formula: ， in, Indicates the probability of a component's existence. Represents the model weight vector. Indicates the bias term; Preset existence probability threshold ,when Greater than or equal to When, determine that the corresponding component exists, when Less than The system determines that the corresponding component was not detected. (2) Burial depth assessment model: Use support vector regression or multilayer perceptron regression model; The epsilon-support vector regression model is used to evaluate the feature vectors of burial depth. Input model, output burial depth estimate ; Then according to the design burial depth and allowable deviation Output the project conclusions according to the following rules: when At that time, the depth was determined to be insufficient; when At that time, the depth was deemed acceptable; when When the depth exceeds the limit, it is determined that the depth has exceeded the limit. When the model's input features exceed the training distribution or the output credibility is lower than a preset threshold, the depth result is deemed unreliable.

9. The RFID detection method for dual-tag dynamic compensation utility poles according to claim 1, characterized in that, S7 is detailed below: When the probability exists within the low confidence interval Internal, or complete set consistency score Below the complete set threshold At that time, an active rescan is triggered; During rescanning, the scaling factor is used. , and The number of interrogation rounds, single-point dwell time, and frequency set size are adjusted separately. Then, the burial depth results of the initial detection and the rescan detection are fused together. The calculation formula is as follows: ， ， in, and These represent the estimated burial depth values ​​obtained from the initial detection and the re-detection, respectively. and These represent the probability of the corresponding detection or its confidence level, respectively. This indicates taking the maximum value. This is the final estimated burial depth after fusion. To determine the final credibility of the test.

10. A dual-tag dynamic compensation RFID detection system for utility poles, comprising executing the dual-tag dynamic compensation RFID detection method for utility poles as described in any one of claims 1-9, characterized in that, include: Dual-tag calibration and deployment module: used to deploy RFID tags in the preset area of ​​cement pole chassis and chuck and perform dual-tag coupling coding calibration. It adopts an eight-segment physical coding structure and generates a coupling calibration sequence through XOR transformation. Scene perception and compensation generation module: used to collect multi-frequency background radio frequency response data in the reference area outside the pole root, construct scene description vector; retrieve matching compensation parameters through scene adaptive dynamic compensation parameter library, and call multilayer perceptron network to generate current scene compensation parameters when no matching scene is found; Multi-path signal acquisition module: Used to acquire raw RFID signal data, including tag ID, signal strength, and phase, at multiple detection locations and multiple frequency points along the circumferential, radial, or combined paths at the base of the pole. Signal compensation and feature extraction module: Used to perform intensity, phase, and frequency differential compensation on the original data using compensation parameters, cluster by tag ID to form a joint observation set, and extract feature vectors for component existence and burial depth assessment. Dual-tag coupling verification module: used to calculate the complete set consistency score based on dual-tag coupling coding, verify the matching of dual tags, and eliminate abnormal data such as mis-installed, cross-component, and mixed reading across poles; The detection and evaluation module is used to input the existence characteristics into the judgment model and output the existence status of the component. For the components that are determined to exist, the module inputs the burial depth evaluation model and outputs the estimated burial depth value. The module combines the design parameters to determine the burial depth qualification. Active rescan fusion module: When the probability of existence is in the low confidence interval or the dual-label verification fails, the scanning parameters are adjusted to trigger a rescan, and the first and rescan data are fused to obtain the final detection result and generate a record.

Images