A method and system for RFID-based tool inventory

Abstract

The application provides an RFID-based tool inventory method and system, obtains multiple sets of original phase values and signal strength RSSI values returned by RFID tags of each power tool in a preset inventory area within a continuous time window, calculates phase dispersion entropy, calculates RSSI mean value to generate RSSI weighted phase compensation factor, and completes phase unfolding after correcting adjacent phase difference value; constructs the unfolded phase, phase dispersion entropy and RSSI mean value into a three-dimensional feature vector, divides the tags into at least one tool group by using a hyperbolic tangent kernel principal component clustering algorithm; calculates the global inventory environment stability index by using the mean value of phase dispersion entropy of all tags, determines a confidence threshold, calculates the group confidence according to the number of tags in the group and the cohesion degree of the feature vector, and uniformly identifies all tool tags in the group as in-stock when the group confidence is greater than the confidence threshold, and generates an inventory list.

Description

Technical Field

[0001] This application belongs to the field of tool inventory, and in particular relates to a tool inventory method and system based on RFID. Background Technology

[0002] Radio Frequency Identification (RFID) technology, with its advantages of non-contact and rapid reading, has been widely used in warehousing and logistics, asset management, and other fields, especially in the inventory of electrical tools, where it can improve efficiency and accuracy. Traditional RFID inventory checks primarily confirm the presence of items by reading the tag's ID. However, in workshops or warehouses with dense tools and complex metal environments, multipath effects caused by signal reflection and diffraction are severe, leading to frequent missed reads and misreads. Existing technologies utilize physical layer information of RFID signals, such as Received Signal Strength Indicator (RSSI) and phase. RSSI can roughly reflect the distance relationship between the tag and the reader, but its stability is poor. Phase information is more sensitive to small displacement changes in the tag and has high accuracy potential; however, the raw phase data suffers from periodic entanglement. Existing technologies typically use phase unwrapping algorithms to recover continuous phase values ​​to address this issue. However, traditional methods such as linear unwrapping are prone to cumulative errors due to misjudging phase transition points when processing signals mixed with strong noise, resulting in distorted unwrapping results. In terms of data quality control, using RSSI thresholds to filter weak signal data cannot identify and remove data points with high signal strength but unstable phases. During the inventory decision-making phase, the strategy ignores the potential clustering characteristics between tools, making the inventory results susceptible to the impact of accidental reading failures of individual tags. Furthermore, in power systems, strong electromagnetic interference makes it difficult to maintain high accuracy under varying interference levels. Therefore, how to utilize phase information and establish an intelligent inventory model capable of adapting to environmental changes and making collective decisions is a pressing challenge in the current technological field. Summary of the Invention

[0003] This invention proposes an RFID-based inventory method to address the problem in power systems where RFID is susceptible to electromagnetic interference, leading to inventory results being easily affected by the failure of a single tag and the method being unable to adapt to changing inventory environments. The method includes: The system acquires multiple sets of original phase values ​​and corresponding RSSI values ​​returned by RFID tags of each power tool within a preset inventory area within a continuous time window. For each tag, it calculates the phase dispersion entropy of the multiple sets of original phase values ​​and compares it with a preset threshold. If the phase dispersion entropy is lower than the preset threshold, all sets of original phase values ​​and corresponding RSSI values ​​are used as the valid data sequence of the tag. If it is not lower than the preset threshold, the two sets of data with the lowest RSSI values ​​are removed, and the remaining phase values ​​and RSSI values ​​are used as the valid data sequence of the tag. Based on the effective data sequence for each tag, the mean RSSI of the sequence is calculated and an RSSI-weighted phase compensation factor is generated; the difference between adjacent phase values ​​in the effective data sequence is corrected using the compensation factor, and then the phase expansion is completed by accumulating the corrected phase difference to obtain the expanded phase of the tag; The expanded phase, phase dispersion entropy, and RSSI mean of the effective data sequence of each label are used to construct a three-dimensional feature vector; and the hyperbolic tangent kernel principal component clustering algorithm is used to process the three-dimensional feature vectors of all labels to divide the tool labels into at least one tool group. Calculate the mean of the phase dispersion entropy of all tags to obtain the global inventory environment stability index, and determine a confidence threshold based on the index and a preset nonlinear mapping function; for each tool group, calculate a group confidence score based on the number of tags in the group and the cohesion of the feature vectors in the group; when the group confidence score is greater than the confidence threshold, mark all tool tags in the group as in stock and generate an inventory list.

[0004] Furthermore, the present invention also relates to an RFID-based tool inventory system, comprising the following modules: The comparison module is used to acquire multiple sets of original phase values ​​and corresponding RSSI values ​​returned by RFID tags of each power tool within a preset inventory area within a continuous time window; for each tag, the phase dispersion entropy of the multiple sets of original phase values ​​of the tag is calculated and compared with a preset threshold: if the phase dispersion entropy is lower than the preset threshold, all multiple sets of original phase values ​​and corresponding RSSI values ​​are used as the valid data sequence of the tag; if it is not lower than the preset threshold, the two sets of data with the lowest RSSI values ​​are removed, and the remaining phase values ​​and RSSI values ​​are used as the valid data sequence of the tag; The correction module is used to calculate the mean RSSI of the effective data sequence for each tag and generate an RSSI-weighted phase compensation factor; use the compensation factor to correct the difference between adjacent phase values ​​in the effective data sequence, and then complete the phase expansion by accumulating the corrected phase difference to obtain the expanded phase of the tag; The partitioning module is used to construct a three-dimensional feature vector from the expanded phase, the phase dispersion entropy, and the mean RSSI of the effective data sequence for each label; and to process the three-dimensional feature vectors of all labels using the hyperbolic tangent kernel principal component clustering algorithm to divide the tool labels into at least one tool group. The generation module is used to calculate the mean of the phase dispersion entropy of all tags to obtain the global inventory environment stability index, and determine a confidence threshold based on the index and a preset nonlinear mapping function; for each tool group, a group confidence is calculated based on the number of tags in the group and the cohesion of the feature vectors in the group; when the group confidence is greater than the confidence threshold, all tool tags in the group are uniformly marked as in stock and an inventory list is generated.

[0005] Compared with existing technologies, this invention uses phase dispersion entropy to assess and filter the stability of raw data, eliminating unstable and noisy data and ensuring data quality. RSSI weighted compensation is used in the phase unfolding stage to reduce cumulative errors caused by noise jumps. This invention provides a holistic evaluation of tool groups by constructing multi-dimensional features that integrate phase, signal stability, and intensity. It then uses clustering algorithms to identify tool groups and makes collective decisions based on group cohesion and the number of labels. Leveraging the physical correlation between tools, even if individual label signals within a group are poor, the overall high confidence level can be corrected, enhancing the anti-interference capability of inventory conclusions and improving the overall accuracy of tool inventory in complex electromagnetic environments. Attached Figure Description

[0006] Figure 1 A flowchart of the first embodiment; Figure 2 This is a schematic diagram of the phase expansion process; Figure 3 A schematic diagram illustrating the determination of the confidence threshold; Figure 4 This is a diagram illustrating the determination of group confidence. Detailed Implementation

[0007] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with relevant laws, regulations, and standards, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0008] In the first embodiment, the present invention proposes an RFID-based tool inventory method, such as... Figure 1 ,include: S1, acquire multiple sets of original phase values ​​and corresponding RSSI values ​​returned by RFID tags of each power tool within a preset inventory area within a continuous time window; for each tag, calculate the phase dispersion entropy of the multiple sets of original phase values ​​of the tag and compare it with a preset threshold: if the phase dispersion entropy is lower than the preset threshold, then all multiple sets of original phase values ​​and corresponding RSSI values ​​are used as the valid data sequence of the tag; if it is not lower than the preset threshold, then after removing the two sets of data with the lowest RSSI values, the remaining phase values ​​and RSSI values ​​are used as the valid data sequence of the tag; Specifically, an UHF RFID reader is deployed in the inventory area, such as a tool cabinet or designated workstation. The reader operates on a designated channel and continuously emits electromagnetic waves into the inventory area. Within a set continuous time window, such as 3 seconds, the reader cyclically reads the passive RFID tags on all power tools in the area at a frequency of 50 times per second. For each identified tag ID, the reader receives the backscattered signal returned by the tag and parses the original carrier phase value and the Received Signal Strength Indicator (RSSI) value from it. Thus, within this 3-second time window, each tag generates an original data sequence consisting of 150 data pairs, each data pair containing a phase value and an RSSI value.

[0009] The phase value range from 0 to 2π is uniformly divided into, for example, 16 intervals, each interval representing a phase state. The 150 original phase values ​​for a given label are iterated through, and the number of phase values ​​falling within each phase state interval is counted. Based on the statistical results, the probability of each state is calculated. Where i ranges from 1 to 16. The calculated entropy value H is compared with a preset threshold obtained through experimental calibration, such as 2.8. If H is less than 2.8, it indicates that the tag signal is stable, and all 150 sets of data from the tag are adopted as valid data sequences. If H is not less than 2.8, it indicates that the signal has significant jitter. In this case, the two data pairs with the lowest RSSI values ​​in the tag data sequence are found and removed from the sequence. The remaining 148 sets of data constitute the valid data sequence of the tag.

[0010] S2, based on the effective data sequence of each tag, calculate the mean RSSI of the sequence and generate an RSSI-weighted phase compensation factor; use the compensation factor to correct the difference between adjacent phase values ​​in the effective data sequence, and then complete the phase expansion by accumulating the corrected phase difference to obtain the expanded phase of the tag; Calculate the arithmetic mean of all RSSI values ​​in a tag's valid data sequence, denoted as the RSSI mean. Based on a preset mapping relationship, such as compensation factor = 1 + (reference RSSI value - current RSSI mean) × scaling factor 0.05, generate an RSSI-weighted phase compensation factor. This factor increases as the RSSI mean decreases. During phase unwrapping, starting from the second phase value in the valid data sequence, calculate the difference between the current phase value and the previous phase value. If the absolute value of the difference is greater than the product of the compensation factor and π, phase entanglement has occurred, requiring 2π compensation. Starting from the first phase value in the sequence, accumulate the compensated phase differences to obtain a continuously changing phase curve. Take the arithmetic mean of all points on this curve as the unwrapped phase of the tag.

[0011] S3, construct a three-dimensional feature vector from the expanded phase, the phase dispersion entropy, and the mean RSSI of the effective data sequence for each label; and use the hyperbolic tangent kernel principal component clustering algorithm to process the three-dimensional feature vectors of all labels, dividing the tool labels into at least one tool group; Specifically, for each tag read within the inventory area, the calculated phase, phase dispersion entropy, and RSSI mean of the tag are combined into a three-dimensional vector. Assuming there are N tags within the inventory area, a dataset containing N three-dimensional vectors is obtained. Hyperbolic tangent kernel principal component analysis is applied to calculate the similarity between all vector pairs using the hyperbolic tangent kernel function, constructing an N×N kernel matrix. Eigenvalue decomposition is performed on this kernel matrix, extracting the eigenvectors corresponding to the two largest eigenvalues, thus mapping the original three-dimensional feature data to a two-dimensional space. In the dimensionality-reduced two-dimensional space, the K-means clustering algorithm is used to divide the data points corresponding to all tags into different clusters, each cluster being a tool group.

[0012] S4. Calculate the mean of the phase dispersion entropy of all tags to obtain the global inventory environment stability index, and determine a confidence threshold based on the index and a preset nonlinear mapping function; for each tool group, calculate a group confidence based on the number of tags in the group and the cohesion of the feature vectors in the group; when the group confidence is greater than the confidence threshold, mark all tool tags in the group as in stock and generate an inventory list.

[0013] The phase dispersion entropy values ​​calculated for all tags in this inventory are summed, and then divided by the total number of tags to obtain an average entropy value, which is the global inventory environment stability index. A higher index indicates more severe electromagnetic interference or poorer overall signal quality in the current inventory environment. A preset sigmoid logistic function is used as a nonlinear mapping function to map this environment stability index to a value between 0 and 1. For example, when the environment stability index is low, the function outputs a higher threshold, such as 0.8, requiring strict judgment criteria; when the index is high, the function outputs a lower threshold, such as 0.6, to adapt to harsh environments. This output value is the confidence threshold for this inventory.

[0014] For each tool group formed by clustering, the centroid of all three-dimensional feature vectors within the group is calculated, i.e., the average value is calculated for each dimension. The Euclidean distance from each vector within the group to the centroid is calculated, and all distances are averaged. The reciprocal of this average distance is used as the cohesion of the group; the smaller the distance, the higher the cohesion. In one embodiment, the formula for calculating the group confidence is: cohesion × weight coefficient 0.6 + weight coefficient 0.4 × number of tags in the group / preset total number of tools. The calculated group confidence is compared with the above confidence threshold. If the group confidence is greater than the threshold, all tools within the group are determined to be in the database. After traversing all groups and completing the judgment, the tag IDs of all tools determined to be in the database are summarized to form a successful inventory list.

[0015] In an optional embodiment, calculating the phase dispersion entropy of multiple sets of original phase values ​​for each tag includes: The multiple sets of original phase values ​​of the label Convert to N-1 adjacent phase difference values ;Will The interval is divided into 16 equal-width sub-intervals; the frequency of phase difference values ​​falling into each sub-interval is counted, and the frequency is calculated. The formula for calculating the phase dispersion entropy is: .

[0016] Assuming the preset threshold is set to 0.85, N sets of raw phase data from one tag are collected, such as... Figure 2 For example, when N is 100, a sequence of 100 phase values ​​is obtained, assuming they are 1.2, 2.5, 3.8, 5.0, ..., 0.5, etc., in radians. The difference between 99 adjacent phases of these 100 values ​​is calculated; for example, the first difference is 1.3, the second is 1.3, and all differences are calculated using modulo 2π to ensure the result falls between 0 and 2π.

[0017] Divide the interval from 0 to 2π into 16 small squares, each approximately 0.39 pixels wide. Count which square each of the 99 phase differences calculated earlier falls into, and calculate the frequency of each square. For example, assuming 30 differences fall into the 5th square, then the frequency of the 5th square... The value is approximately 0.303. The total entropy is calculated using the above formula for all 16 cells. If the calculated entropy value H is less than the preset value of 0.85, it indicates that the phase data of the tag is sufficiently stable.

[0018] In an optional embodiment, the step of calculating the RSSI mean of the valid data sequence based on each tag and generating an RSSI-weighted phase compensation factor includes: Calculate the arithmetic mean of the M RSSI values ​​in the valid data sequence. The formula for calculating the RSSI-weighted phase compensation factor C is as follows:

[0019] in This is the reference signal strength.

[0020] Specifically, the reference signal strength is set to -60 dBm. For a tag that has passed the aforementioned entropy test, all M RSSI values, i.e., signal strength indication values, are extracted from the tag's valid data sequence. Assume a sequence containing 10 RSSI values ​​is obtained, for example, -55, -56, -55, -57, -54, -56, -55, -58, -55, -54, in dBm. The calculated arithmetic mean is -55.5 dBm, which is the average RSSI of the tag. The average RSSI value is then substituted into a given formula to calculate the compensation factor C. The reference signal strength in the formula... The value is fixed at -60dBm. The calculated compensation factor C is approximately 3.836. The stronger the signal, the closer C is to 2π; the weaker the signal, the closer C is to π.

[0021] To eliminate the ambiguity caused by phase jumps between 0 and 2π, in an optional embodiment, the step of correcting the difference between adjacent phase values ​​in the effective data sequence using the compensation factor, and then completing phase unrolling by accumulating the corrected phase differences, includes: The phase value in the effective data sequence is denoted as For any adjacent phase values and Calculate the difference ; like The corrected phase difference ; like The corrected phase difference ; otherwise, ; Where C is the RSSI-weighted phase compensation factor; the expanded phase of the tag The result is obtained through cumulative calculation: .

[0022] Specifically, using the n phase values ​​from the valid data sequence obtained in the previous step, for example, a sequence of 1.5, 5.8, 2.1, 5.5, and the compensation factor C calculated above (assuming a value of 3.836), the differences between adjacent phase values ​​are calculated sequentially. The first difference is 4.3. Because 4.3 is greater than π, the correction condition is triggered, and the corrected difference is 0.464. The second difference is -3.7, which also triggers the correction condition, and the corrected difference is 0.136. Differences that do not exceed the ±π threshold remain unchanged. After calculating and correcting all differences, the first phase value of 1.5 is added to all corrected phase differences; the sum is the expanded phase of the label, representing the continuously changing quantity.

[0023] In an optional embodiment, the processing of the three-dimensional feature vectors of all labels using the hyperbolic tangent kernel principal component clustering algorithm includes: The hyperbolic tangent kernel function is:

[0024] in and For any two three-dimensional feature vectors, Here, is the kernel function parameter, and c is a parameter; a kernel matrix is ​​constructed using the kernel function, and the kernel matrix is ​​subjected to feature decomposition to extract the first two principal components. Then, the K-means clustering algorithm is used to cluster all labels in the two-dimensional principal component space, wherein the number of clusters K is iteratively increased from 2 until the preset silhouette coefficient is greater than 0.7.

[0025] For example, the kernel function parameters The value is set to 1.0, and the parameter c is set to -0.5. A three-dimensional feature vector containing phase, Doppler shift, and RSSI is generated for each label. Assume there are 10 labels, i.e., 10 three-dimensional vectors. The similarity between every two vectors is calculated using a given hyperbolic tangent kernel function. For example, for vector... and First, calculate their dot product, then multiply the dot product by the parameter α (1.0), add the parameter c (-0.5), and take the hyperbolic tangent of the entire result. This yields a 10×10 kernel matrix, where each element represents the similarity of a pair of label features.

[0026] Principal component analysis (PCA) is performed on the kernel matrix, i.e., eigenvalue decomposition is performed to extract the eigenvectors corresponding to the two largest eigenvalues. These two eigenvectors constitute a new two-dimensional space. The original 10 three-dimensional eigenvectors are then mapped to 10 points in this two-dimensional space. K-means clustering is performed on these 10 two-dimensional points. Initially, the number of clusters K is set to 2, dividing the points into two clusters, and the silhouette coefficient is calculated to evaluate the clustering effect. If the silhouette coefficient is less than 0.7, K is increased to 3, and the clustering is re-clustered and the silhouette coefficient is calculated again. This process is repeated until a K value is found that makes the silhouette coefficient greater than 0.7; the clustering result at this point is considered a tool group partition.

[0027] To adjust the judgment criteria based on the stability of the current inventory environment, in an optional embodiment, determining a confidence threshold based on the index and a preset nonlinear mapping function includes: Let the global inventory environment stability index be denoted as . The nonlinear mapping function is a piecewise linear function: when At that time, confidence threshold ; when At that time, confidence threshold ; when At that time, confidence threshold .

[0028] Specifically, a global inventory environment stability index is calculated by analyzing the overall fluctuations of all tag data. The higher the index, the more stable the environment; conversely, the lower the index, the more unstable the environment.

[0029] The corresponding calculation formula is selected based on the value of the index. For example, if the calculated value is... The value is 1.2, falling between 0.5 and 1.5, therefore the middle formula is chosen. The confidence threshold is then calculated. It is 0.81, such as Figure 3 If the environment is very unstable, for example... If the value is 0.3, then the threshold will be set directly to the highest value of 0.95, which is a very strict requirement. If the environment is very stable, for example... The standard is set at 2.0, while the threshold is set to a minimum of 0.75, indicating a relatively lenient standard. This threshold is used to assess the reliability of each tool group.

[0030] In an optional embodiment, calculating a group confidence score for each tool group based on the number of labels within the group and the cohesion of the feature vectors within the group includes: For the set of three-dimensional feature vectors of all labels, perform min-max normalization on each feature dimension to linearly map each dimensional component of all feature vectors to the interval [0, 1]. Based on the normalized feature vectors, the centroids of all three-dimensional feature vectors within the group are calculated. Then, the Euclidean distance from each feature vector to the centroid is calculated, and the average of all distances is defined as the group divergence D. The group confidence level... The calculation formula is: in This is the number of tags in the current group. It is the maximum allowable divergence; and These are the weighting coefficients.

[0031] Specifically, for each tool group divided by the clustering algorithm, the confidence score of that group needs to be calculated. The original three-dimensional feature vectors of all labels are normalized; that is, for the phase, Doppler shift, and RSSI dimensions, the maximum and minimum values ​​are found among all labels, respectively. The values ​​of the corresponding dimensions for each label are scaled to the range of 0 to 1, which can eliminate the influence of differences in feature scales. Only the labels within the current group are considered, and their normalized feature vectors are used. Assume a group contains... Given 5 labels, first calculate the average vector of the 5 3D vectors, i.e., the centroid. Then calculate the Euclidean distance from each vector to the centroid, and average the 5 distances to obtain the group divergence D, which is assumed to be 0.1. Substituting α=5 and D=0.1 into the confidence calculation formula, where the weighting coefficients α and β are both set to 0.5, the maximum allowable divergence is... Set it to 0.3. A group confidence score between 0 and 1 is calculated, which comprehensively reflects the number of labels and feature similarity within the group. A higher score indicates a more reliable group. Figure 4 .

[0032] In a second embodiment, the present invention also provides an RFID-based tool inventory system, comprising the following modules: The comparison module is used to acquire multiple sets of original phase values ​​and corresponding RSSI values ​​returned by RFID tags of each power tool within a preset inventory area within a continuous time window; for each tag, the phase dispersion entropy of the multiple sets of original phase values ​​of the tag is calculated and compared with a preset threshold: if the phase dispersion entropy is lower than the preset threshold, all multiple sets of original phase values ​​and corresponding RSSI values ​​are used as the valid data sequence of the tag; if it is not lower than the preset threshold, the two sets of data with the lowest RSSI values ​​are removed, and the remaining phase values ​​and RSSI values ​​are used as the valid data sequence of the tag; The correction module is used to calculate the mean RSSI of the effective data sequence for each tag and generate an RSSI-weighted phase compensation factor; use the compensation factor to correct the difference between adjacent phase values ​​in the effective data sequence, and then complete the phase expansion by accumulating the corrected phase difference to obtain the expanded phase of the tag; The partitioning module is used to construct a three-dimensional feature vector from the expanded phase, the phase dispersion entropy, and the mean RSSI of the effective data sequence for each label; and to process the three-dimensional feature vectors of all labels using the hyperbolic tangent kernel principal component clustering algorithm to divide the tool labels into at least one tool group. The generation module is used to calculate the mean of the phase dispersion entropy of all tags to obtain the global inventory environment stability index, and determine a confidence threshold based on the index and a preset nonlinear mapping function; for each tool group, a group confidence is calculated based on the number of tags in the group and the cohesion of the feature vectors in the group; when the group confidence is greater than the confidence threshold, all tool tags in the group are uniformly marked as in stock and an inventory list is generated.

[0033] In an optional embodiment, calculating the phase dispersion entropy of multiple sets of original phase values ​​for each tag includes: The multiple sets of original phase values ​​of the label Convert to N-1 adjacent phase difference values ;Will The interval is divided into 16 equal-width sub-intervals; the frequency of phase difference values ​​falling into each sub-interval is counted, and the frequency is calculated. The formula for calculating the phase dispersion entropy is: .

[0034] In an optional embodiment, the step of calculating the RSSI mean of the valid data sequence based on each tag and generating an RSSI-weighted phase compensation factor includes: Calculate the arithmetic mean of the M RSSI values ​​in the valid data sequence. The formula for calculating the RSSI-weighted phase compensation factor C is as follows:

[0035] in This is the reference signal strength.

[0036] In an optional embodiment, the step of using the compensation factor to correct the difference between adjacent phase values ​​in the effective data sequence, and then completing phase unwrapping by accumulating the corrected phase differences, includes: The phase value in the effective data sequence is denoted as For any adjacent phase values and Calculate the difference ; like The corrected phase difference ; like The corrected phase difference ; otherwise, ; Where C is the RSSI-weighted phase compensation factor; the expanded phase of the tag The result is obtained through cumulative calculation: .

[0037] In an optional embodiment, the processing of the three-dimensional feature vectors of all labels using the hyperbolic tangent kernel principal component clustering algorithm includes: The hyperbolic tangent kernel function is:

[0038] in and For any two three-dimensional feature vectors, Here, is the kernel function parameter, and c is a parameter; a kernel matrix is ​​constructed using the kernel function, and the kernel matrix is ​​subjected to feature decomposition to extract the first two principal components. Then, the K-means clustering algorithm is used to cluster all labels in the two-dimensional principal component space, wherein the number of clusters K is iteratively increased from 2 until the preset silhouette coefficient is greater than 0.7.

[0039] In an optional embodiment, determining a confidence threshold based on the exponent and a preset nonlinear mapping function includes: Let the global inventory environment stability index be denoted as . The nonlinear mapping function is a piecewise linear function: when At that time, confidence threshold ; when At that time, confidence threshold ; when At that time, confidence threshold .

[0040] In an optional embodiment, calculating a group confidence score for each tool group based on the number of labels within the group and the cohesion of the feature vectors within the group includes: For the set of three-dimensional feature vectors of all labels, perform min-max normalization on each feature dimension to linearly map each dimensional component of all feature vectors to the interval [0, 1]. Based on the normalized feature vectors, the centroids of all three-dimensional feature vectors within the group are calculated. Then, the Euclidean distance from each feature vector to the centroid is calculated, and the average of all distances is defined as the group divergence D. The group confidence level... The calculation formula is: in This is the number of tags in the current group. It is the maximum allowable divergence; and These are the weighting coefficients.

[0041] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0042] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A tool inventory method based on RFID, characterized in that, Includes the following steps: S1, acquire multiple sets of original phase values ​​and corresponding RSSI values ​​returned by the RFID tags of each power tool in the preset inventory area within a continuous time window; for each tag, calculate the phase dispersion entropy of the multiple sets of original phase values ​​of the tag and compare it with a preset threshold: if the phase dispersion entropy is lower than the preset threshold, then all multiple sets of original phase values ​​and corresponding RSSI values ​​are used as the valid data sequence of the tag. If the value is not lower than the preset threshold, the two sets of data with the lowest RSSI values ​​are removed, and the remaining phase values ​​and RSSI values ​​are used as the valid data sequence of the tag. The step of calculating the phase dispersion entropy of multiple sets of original phase values ​​for each tag includes: The multiple sets of original phase values ​​of the label Convert to N-1 adjacent phase difference values ;Will The interval is divided into 16 equal-width sub-intervals; the frequency of phase difference values ​​falling into each sub-interval is counted, and the frequency is calculated. The formula for calculating the phase dispersion entropy is: ; S2, based on the effective data sequence of each tag, calculate the mean RSSI of the sequence and generate an RSSI-weighted phase compensation factor; use the compensation factor to correct the difference between adjacent phase values ​​in the effective data sequence, and then complete the phase expansion by accumulating the corrected phase difference to obtain the expanded phase of the tag; The calculation of the RSSI mean of the effective data sequence based on each label and the generation of an RSSI-weighted phase compensation factor include: Calculate the arithmetic mean of the M RSSI values ​​in the valid data sequence. The formula for calculating the RSSI weighted phase compensation factor C is as follows: ， in Reference signal strength; S3, construct a three-dimensional feature vector from the expanded phase, the phase dispersion entropy, and the mean RSSI of the effective data sequence for each label; and use the hyperbolic tangent kernel principal component clustering algorithm to process the three-dimensional feature vectors of all labels, dividing the tool labels into at least one tool group; The process of using the hyperbolic tangent kernel principal component clustering algorithm to process the three-dimensional feature vectors of all labels includes: The hyperbolic tangent kernel function is: ， in and Let a be any two three-dimensional feature vectors, where a is the kernel function parameter and c is the parameter; construct a kernel matrix through the kernel function, perform feature decomposition on the kernel matrix to extract the first two principal components, and then use the K-means clustering algorithm to cluster all labels in the two-dimensional principal component space, wherein the value of the number of clusters K starts from 2 and increases iteratively until the preset silhouette coefficient is greater than 0.

7. S4, calculate the mean of the phase dispersion entropy of all tags to obtain the global inventory environment stability index, and determine a confidence threshold based on the index and a preset nonlinear mapping function; for each tool group, calculate a group confidence based on the number of tags in the group and the cohesion of the feature vectors in the group; when the group confidence is greater than the confidence threshold, mark all tool tags in the group as in stock and generate an inventory list; For each tool group, a group confidence score is calculated based on the number of labels within the group and the cohesion of the feature vectors within the group, including: For the set of three-dimensional feature vectors of all labels, perform min-max normalization on each feature dimension to linearly map each dimensional component of all feature vectors to the interval [0, 1]. Based on the normalized feature vectors, the centroids of all three-dimensional feature vectors within the group are calculated. Then, the Euclidean distance from each feature vector to the centroid is calculated, and the average of all distances is defined as the group divergence D. The group confidence level... The calculation formula is: in This is the number of tags in the current group. It is the maximum allowable divergence; and These are the weighting coefficients.

2. The method according to claim 1, characterized in that, The step of correcting the difference between adjacent phase values ​​in the effective data sequence using the compensation factor, and then completing phase expansion by accumulating the corrected phase differences, includes: The phase value in the effective data sequence is denoted as... For any adjacent phase values and Calculate the difference ; like The corrected phase difference ; like The corrected phase difference ; otherwise, ; Where C is the RSSI-weighted phase compensation factor; the expanded phase of the tag The result is obtained through cumulative calculation: .

3. The method according to claim 1, characterized in that, Determining a confidence threshold based on the exponent and a preset nonlinear mapping function includes: Let the global inventory environment stability index be denoted as . The nonlinear mapping function is a piecewise linear function: when At that time, confidence threshold ; when At that time, confidence threshold ; when At that time, confidence threshold .

4. An RFID-based tool inventory system, characterized in that, Includes the following modules: The comparison module is used to acquire multiple sets of original phase values ​​and corresponding signal strength RSSI values ​​returned by the RFID tags of each power tool in the preset inventory area within a continuous time window; for each tag, the phase dispersion entropy of the multiple sets of original phase values ​​of the tag is calculated and compared with a preset threshold; if the phase dispersion entropy is lower than the preset threshold, then all multiple sets of original phase values ​​and corresponding RSSI values ​​are used as the valid data sequence of the tag. If the value is not lower than the preset threshold, the two sets of data with the lowest RSSI values ​​are removed, and the remaining phase values ​​and RSSI values ​​are used as the valid data sequence of the tag. The step of calculating the phase dispersion entropy of multiple sets of original phase values ​​for each tag includes: The multiple sets of original phase values ​​of the label Convert to N-1 adjacent phase difference values ;Will The interval is divided into 16 equal-width sub-intervals; the frequency of phase difference values ​​falling into each sub-interval is counted, and the frequency is calculated. The formula for calculating the phase dispersion entropy is: ; The correction module is used to calculate the mean RSSI of the effective data sequence for each tag and generate an RSSI-weighted phase compensation factor; use the compensation factor to correct the difference between adjacent phase values ​​in the effective data sequence, and then complete the phase expansion by accumulating the corrected phase difference to obtain the expanded phase of the tag; The calculation of the RSSI mean of the effective data sequence based on each label and the generation of an RSSI-weighted phase compensation factor include: Calculate the arithmetic mean of the M RSSI values ​​in the valid data sequence. The formula for calculating the RSSI weighted phase compensation factor C is as follows: ， in Reference signal strength; The partitioning module is used to construct a three-dimensional feature vector from the expanded phase, the phase dispersion entropy, and the mean RSSI of the effective data sequence for each label; and to process the three-dimensional feature vectors of all labels using the hyperbolic tangent kernel principal component clustering algorithm to divide the tool labels into at least one tool group. The process of using the hyperbolic tangent kernel principal component clustering algorithm to process the three-dimensional feature vectors of all labels includes: The hyperbolic tangent kernel function is: ， in and Let a be any two three-dimensional feature vectors, where a is the kernel function parameter and c is the parameter; construct a kernel matrix through the kernel function, perform feature decomposition on the kernel matrix to extract the first two principal components, and then use the K-means clustering algorithm to cluster all labels in the two-dimensional principal component space, wherein the value of the number of clusters K starts from 2 and increases iteratively until the preset silhouette coefficient is greater than 0.

7. The generation module is used to calculate the mean of the phase dispersion entropy of all tags to obtain the global inventory environment stability index, and determine a confidence threshold based on the index and a preset nonlinear mapping function; for each tool group, a group confidence is calculated based on the number of tags in the group and the cohesion of the feature vectors in the group; when the group confidence is greater than the confidence threshold, all tool tags in the group are uniformly marked as in stock and an inventory list is generated. For each tool group, a group confidence score is calculated based on the number of labels within the group and the cohesion of the feature vectors within the group, including: For the set of three-dimensional feature vectors of all labels, perform min-max normalization on each feature dimension to linearly map each dimensional component of all feature vectors to the interval [0, 1]. Based on the normalized feature vectors, the centroids of all three-dimensional feature vectors within the group are calculated. Then, the Euclidean distance from each feature vector to the centroid is calculated, and the average of all distances is defined as the group divergence D. The group confidence level... The calculation formula is: in This is the number of tags in the current group. It is the maximum allowable divergence; and These are the weighting coefficients.

Images