An RFID-based inventory tracking management system

Abstract

This invention relates to the field of item tracking technology, specifically to an RFID-based inventory tracking and management system. The system includes: a signal difference verification module, an inventory area mapping module, a status trajectory correction module, and an inventory movement determination module. In this invention, the sensitivity of anomaly detection is improved by calculating the signal power difference between adjacent areas and comparing it in real time against an attenuation threshold, avoiding the risk of misjudgment due to equipment aging or environmental interference under fixed threshold conditions. A Manhattan distance-matched frequency template library is used to screen candidate areas to solve positioning drift caused by multipath effects and environmental electromagnetic interference. Sliding window mean filtering and cubic spline interpolation are used to process historical trajectories, eliminating jumps caused by signal jitter and brief interruptions. Illegal movement is identified based on direction vector dot product, breaking through the limitations of traditional static rules and adapting to multi-directional legal paths in complex scenarios. A progressive data processing mechanism realizes closed-loop verification of abnormal positioning trajectory correction and behavior analysis, reducing manual intervention and improving the timeliness of inventory status updates and the reliability of decision-making.

Description

Technical Field

[0001] This invention relates to the field of item tracking technology, and more particularly to an RFID-based inventory tracking and management system. Background Technology

[0002] The field of item tracking technology encompasses various methods and systems for identifying, recording, and dynamically managing the location, status, and flow of physical goods, assets, or individuals. Its core lies in achieving end-to-end information visualization, data traceability, and status verification through technologies such as information coding, radio frequency identification (RFID), and image recognition. Typical applications include logistics and transportation, supply chain management, and medical sample management, involving data collection devices such as barcode scanners and RFID readers. Information is then centrally processed and updated to a database via a network. The inventory tracking management system, as a specific implementation of this technology, assigns unique RFID identification codes to inventory units, uses warehouse node readers to collect movement paths in real time, transmits the data wirelessly to the backend database for code matching and automatic updates, establishes item status marking rules to achieve identification management at each stage, and synchronizes inventory records through data interface integration with the enterprise resource planning system. This enables computerized control of item identification, location tracking, and status marking throughout the processes of warehousing, storage, relocation, and outbound.

[0003] Current technologies use fixed signal attenuation thresholds to determine location anomalies, failing to dynamically adjust the criteria based on warehouse temperature and humidity changes, equipment performance fluctuations, etc., and are prone to misinterpreting signal fluctuations caused by environmental interference as inventory displacement. For example, radio frequency signal reflections caused by metal shelves may trigger false alarms. Current technologies rely on single reader data to locate inventory positions, lacking a multi-node signal attenuation correlation analysis mechanism, making it difficult to distinguish between actual location offset and signal transmission path interference, leading to delayed or incorrect coordinate updates. Current technologies directly use raw coordinate points to generate movement trajectories without incorporating time-series data filtering algorithms, resulting in jagged paths when handling equipment stops abruptly or signals are lost, affecting the accuracy of movement direction determination. Current technologies pre-determine legal movement directions based on a static rule base, failing to automatically adapt to changes in scenarios such as the opening of temporary aisles or adjustments to shelf layout, potentially mismarking newly added compliant paths as illegal movements. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing an RFID-based inventory tracking and management system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: an RFID-based inventory tracking management system comprising:

[0006] The signal difference verification module is used to obtain the received power value of the tag in the adjacent shelf area, calculate the power difference between the current area and the previous area, compare the power attenuation threshold between shelves, generate an abnormal inventory location signal, associate the tag ID with the reader number to generate an abnormal signal identifier, and transmit it to the inventory area mapping module.

[0007] The inventory area mapping module is used to extract the frequency template library corresponding to the tag based on the abnormal signal identifier, match the current frequency with the reference frequency of the template library area through the Manhattan distance algorithm, filter the minimum distance to generate candidate inventory area identifiers, and pass them to the state trajectory correction module.

[0008] The state trajectory correction module is used to call the coordinates of three historical time series points through the candidate inventory area identifier, perform sliding window mean filtering and cubic spline interpolation to generate a smooth inventory coordinate sequence, and pass it to the inventory movement determination module.

[0009] As a further aspect of the present invention, the abnormal signal abnormality identifier specifically includes power difference, threshold range, tag ID, and reader number; the candidate inventory area identifier includes area number, reference frequency, and minimum distance value; and the smoothed inventory coordinate sequence specifically includes time-series point coordinates, filter coefficients, and interpolation parameters.

[0010] As a further aspect of the present invention, the power difference is the absolute value of the dBm difference between the tag received power of the current area and the previous area;

[0011] The unit of the power attenuation threshold between shelves is dBm, and the threshold range is determined by fitting experimental data.

[0012] The Manhattan distance algorithm introduces normalized weight coefficients for multi-dimensional frequency components, which are determined by optimization using the gradient descent method.

[0013] The sliding window mean filter has a window size of 3 time points, and the number of nodes in the cubic spline interpolation satisfies the Nyquist sampling theorem constraint.

[0014] As a further aspect of the present invention, the signal difference verification module includes:

[0015] The power difference calculation submodule obtains the tag receiving power values ​​of adjacent shelf areas, extracts the tag receiving power value recorded by the reader in the current area and the tag receiving power value recorded in the previous area, performs a subtraction operation on the two sets of power values, and obtains the power attenuation difference.

[0016] The attenuation threshold comparison submodule calls the power attenuation difference to compare with the power attenuation threshold between shelves. When the absolute value of the power attenuation difference exceeds the power attenuation threshold between shelves, an abnormal judgment state is marked. When the absolute value of the power attenuation difference does not exceed the threshold, a normal judgment state is marked and an abnormal judgment identifier is generated.

[0017] The abnormal signal generation submodule combines the abnormal judgment identifier to extract the tag ID that triggers the abnormal state and the corresponding reader number, and concatenates the tag ID and reader number into a string according to a preset format as the unique identifier of the abnormal signal to generate an abnormal signal identifier.

[0018] In the subtraction operation, the power value is in dBm, and the reader number is bound to the shelf area code.

[0019] As a further aspect of the present invention, the inventory area mapping module includes:

[0020] The frequency template library extraction submodule parses the abnormal signal identifier, extracts the tag code, retrieves regional reference frequency data that matches the tag from the pre-stored frequency template library, aggregates multi-dimensional frequency components, and generates a set of regional reference frequencies.

[0021] The regional distance calculation submodule calls the regional reference frequency set, collects the multi-dimensional components of the current frequency, and uses the following formula:

[0022]

[0023] Calculate the comprehensive distance value of the region, iterate through all regional reference frequencies, perform formula calculations, and generate a set of regional distances;

[0024] Among them, D r F represents the composite distance value of region r. a For the current frequency component in dimension a, F ra Let ω be the a-th dimension component of the reference frequency in region r. a Let V be the frequency difference weight in dimension a, and V be the current inventory capacity value. r ΔT represents the upper limit of the capacity of region r. b Let be the standard deviation of the fluctuation of the b-th type of signal, p be the total number of frequency dimensions, and q be the total number of signal categories;

[0025] The candidate region filtering submodule traverses the set of region distances, compares the comprehensive distance values ​​of multiple regions, selects the region identifier corresponding to the minimum value, maps it to the inventory space coding library, and generates candidate inventory region identifiers.

[0026] The multidimensional frequency components include the center frequency, frequency offset, and multipath effect compensation parameters of the tag signal;

[0027] The weighting coefficient ω aThe standard deviation of signal fluctuation ΔTb is determined by optimization using the training dataset, and is calculated based on historical signal stability data.

[0028] As a further aspect of the present invention, the state trajectory correction module includes:

[0029] Based on the candidate inventory area identifier, the coordinate data acquisition submodule calls the coordinate data of three historical time series points, extracts the X-axis, Y-axis and Z-axis coordinate components corresponding to each time series point, arranges them into a three-dimensional coordinate point set in chronological order, and generates a time series coordinate sequence.

[0030] The mean filtering submodule obtains the time-series coordinate sequence, defines a sliding window to cover the coordinates of three adjacent time-series points, sums the X-axis coordinate components within the window and divides them by the window size, performs the same operation on the Y-axis and Z-axis, replaces the coordinates of the window center point with the calculation results, traverses all time-series points to complete the coordinate component update, and generates the filtered coordinate sequence.

[0031] The interpolation sequence generation submodule calls the filtered coordinate sequence, selects adjacent time series points as interpolation nodes, uses the X, Y, and Z components at the nodes as boundary conditions, calculates the coefficients of a cubic polynomial, supplements the component values ​​of missing coordinate points between nodes, connects the interpolation points with the original nodes to form a continuous trajectory, and generates a smooth inventory coordinate sequence.

[0032] The time interval of the time-series coordinate sequence is 1 to 5 seconds, and the specific value is dynamically configured according to the inventory movement speed.

[0033] The sliding window moves in one time step, and the weights of the coordinate components within the window are equally weighted.

[0034] The coefficients of the cubic polynomial satisfy the condition of continuity of the first derivative at the nodes.

[0035] As a further aspect of the present invention, the system further includes:

[0036] The inventory movement determination module is used to extract the direction vectors of two consecutive coordinate points based on the smoothed inventory coordinate sequence, perform a dot product operation with the preset legal movement direction library vector, generate an illegal movement alarm event when the dot product result is negative, write the illegal movement alarm event to the log and trigger the database update.

[0037] As a further aspect of the present invention, the illegal movement alarm event includes a direction vector, a dot product result, and an alarm level.

[0038] As a further aspect of the present invention, the preset legal movement direction library is generated based on the direction vector clustering of historical compliant movement trajectories.

[0039] As a further aspect of the present invention, the inventory movement determination module includes:

[0040] The direction extraction submodule extracts two consecutive coordinate points based on the smoothed inventory coordinate sequence, calculates the vector value between the current coordinate point and the next coordinate point, arranges each pair of coordinate vectors in chronological order, obtains the inventory status movement direction vector sequence, and obtains the direction vector sequence value.

[0041] The direction comparison submodule calls all direction vectors in the preset legal movement direction library according to the direction vector sequence value, performs dot product operation for each group of stock movement direction vectors, obtains the dot product calculation value, filters the dot product results, and marks the direction vectors with dot product values ​​less than zero as illegal vectors to obtain the illegal direction judgment quantity.

[0042] The alarm generation submodule locates the corresponding inventory movement event data based on the illegal direction determination quantity, matches and records the coordinate index position of the illegal vector, generates illegal event log entries from the matching results and calls the database write operation, and simultaneously executes the trigger update operation to generate illegal movement alarm entry values.

[0043] The calculation accuracy of the direction vector is 0.1 m / s, and the resolution of the vector direction angle is 1 degree.

[0044] When the dot product result is negative, it is determined that the angle between the moving direction and the legal direction exceeds 90 degrees.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0046] This invention improves anomaly detection sensitivity by calculating the signal power difference between adjacent areas and comparing it with an attenuation threshold in real time, avoiding the risk of misjudgment due to equipment aging or environmental interference under fixed threshold conditions. A Manhattan distance-matched frequency template library is used to filter candidate areas, solving the positioning drift problem caused by multipath effects or environmental electromagnetic interference and enhancing position mapping accuracy. Historical trajectory data is processed using sliding window mean filtering and cubic spline interpolation to eliminate coordinate jumps caused by signal jitter and brief interruptions, generating smooth and continuous movement paths. Illegal movement is identified based on direction vector dot product operations, breaking through the limitations of traditional static rule bases and adapting to multi-directional legal path determination in complex scenarios. A progressive data processing mechanism achieves closed-loop verification of anomaly location, trajectory correction, and behavior analysis, reducing the need for manual intervention and improving the timeliness of inventory status updates and the reliability of decision-making. Attached Figure Description

[0047] Figure 1 This is a system flowchart of the present invention;

[0048] Figure 2 This is a flowchart of the signal difference verification module of the present invention;

[0049] Figure 3 This is a flowchart of the inventory area mapping module of the present invention;

[0050] Figure 4 This is a flowchart of the state trajectory correction module of the present invention;

[0051] Figure 5 This is a flowchart of the inventory movement determination module of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0053] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0054] Example 1

[0055] Please see Figure 1 An RFID-based inventory tracking management system includes:

[0056] The signal difference verification module is used to obtain the received power value of the tag in the adjacent shelf area, calculate the power difference between the current area and the previous area, compare the power attenuation threshold between shelves, generate an abnormal inventory location signal, associate the tag ID with the reader number to generate an abnormal signal identifier, and transmit it to the inventory area mapping module.

[0057] The inventory area mapping module is used to extract the frequency template library corresponding to the tag based on the abnormal signal identifier, match the current frequency with the reference frequency of the template library area through the Manhattan distance algorithm, filter the minimum distance to generate candidate inventory area identifiers, and pass them to the status trajectory correction module.

[0058] The state trajectory correction module is used to call the coordinates of three historical time series points through the candidate inventory area identifier, perform sliding window mean filtering and cubic spline interpolation to generate a smooth inventory coordinate sequence, and pass it to the inventory movement determination module.

[0059] The inventory movement determination module is used to extract the direction vectors of two consecutive coordinate points based on the smoothed inventory coordinate sequence, perform a dot product operation with the preset legal movement direction library vector, and generate an illegal movement alarm event when the dot product result is negative. The illegal movement alarm event is written to the log and triggers the database update.

[0060] The abnormal signal identification includes power difference, threshold range, tag ID, and reader number. The candidate inventory area identification includes area number, reference frequency, and minimum distance value. The smoothed inventory coordinate sequence includes time-series point coordinates, filter coefficients, and interpolation parameters. The illegal movement alarm event includes direction vector, dot product result, and alarm level.

[0061] The power difference is the absolute value of the dBm difference between the tag received power in the current area and the previous area;

[0062] The unit of the power attenuation threshold between shelves is dBm, and the threshold range is determined by fitting experimental data.

[0063] The Manhattan distance algorithm introduces normalized weight coefficients for multi-dimensional frequency components, which are determined by optimization using the gradient descent method.

[0064] The sliding window mean filter has a window size of 3 time points, and the number of nodes in the cubic spline interpolation satisfies the Nyquist sampling theorem constraint.

[0065] The preset legal movement direction library is generated by clustering the direction vectors of historical compliant movement trajectories.

[0066] Please see Figure 2 The signal difference verification module includes:

[0067] The power difference calculation submodule obtains the tag reception power values ​​of adjacent shelf areas through the reader hardware interface configured in the shelf area. Specifically, when the system detects that RFID tag Tag789 has moved from its current location in the first shelf area (covered by reader R001) to the adjacent second shelf area (covered by reader R002), the power difference calculation submodule first sends a command to reader R001 requesting the extraction of its most recently recorded reception power value for tag Tag789. This value is -65dBm, representing the signal strength of tag Tag789 when it was stably read by R001 in the first shelf area. Subsequently, the submodule sends a command to reader R002 requesting the extraction of its currently recorded reception power value for tag Tag789. This value is -72dBm, representing the signal strength of the tag after entering the second shelf area. Next, the submodule performs a subtraction operation on the two sets of power values. Specifically, it subtracts the power value of the previous region from the power value of the latter region, i.e., (-72dBm)-(-65dBm). After the operation, the power attenuation difference is -7dBm, which indicates that the signal strength has been attenuated.

[0068] The attenuation threshold comparison submodule calls the power attenuation difference calculated by the previous submodule, which is -7dBm, and reads the preset inter-shelf power attenuation threshold from the system configuration library. This threshold is set with reference to the typical attenuation range of signal power when tags move normally between shelves in the target warehousing environment. The specific setting process is as follows: Select 100 tags of different types. Under the condition that the readers are working properly and environmental interference factors are controlled, each tag is moved 10 times in a controlled manner between two adjacent predetermined test shelves. Record the signal power values ​​received by the two readers before and after each movement, calculate the absolute value of the difference, and form a total of 100 tags × 10 times / tag = 1000 absolute power attenuation value sample data. Perform statistical analysis on these data, such as calculating their cumulative distribution function. If it is found that 95% of the sample data values ​​are below 3.5dBm, then the inter-shelf power attenuation threshold is set to 3.5dBm. The experimental verification process collected data from experimental scenarios numbered S001 to S100, recording 20 attenuation data points for each scenario. The distribution characteristics were analyzed, and the data showed that using 3.5 dBm as the threshold effectively distinguishes between normal attenuation and abnormal attenuation caused by skipping or cross-reading. The false positive rate of misidentifying normal attenuation as abnormal was controlled within 2%, while the false negative rate of failing to detect abnormal attenuation was controlled within 3%, proving the rationality and effectiveness of this threshold setting. The submodule then compared the absolute value of this power attenuation difference with the threshold, i.e., compared |-7 dBm|.

[0069] The power attenuation difference is 3.5 dBm. The calculated value is |-7 dBm| = 7 dBm. Since the absolute value of the power attenuation difference, 7 dBm, exceeds the inter-shelf power attenuation threshold of 3.5 dBm (judgment rule: when the absolute value of the difference > 3.5 dBm, it exceeds the threshold; when the absolute value of the difference ≤ 3.5 dBm, it does not exceed the threshold), the current judgment state is marked as an abnormal judgment state. If the calculated absolute value of the power attenuation difference is 2 dBm, then since 2 dBm ≤ 3.5 dBm, it is marked as a normal judgment state. Finally, a corresponding abnormal judgment label is generated based on the comparison results; the abnormal judgment label generated here is "abnormal".

[0070] The abnormal signal generation submodule first combines the abnormal judgment identifier generated by the previous submodule. When this identifier is confirmed as "abnormal," the subsequent processing flow is initiated. It extracts the tag ID that triggered the abnormal state from the internal records or transmitted data of the power difference calculation submodule, identifying it as "Tag789," and extracts the current area reader number recorded to the abnormal power jump, identifying it as "R002." Subsequently, the submodule concatenates the obtained tag ID "Tag789" and reader number "R002" according to a system-preset fixed format. This preset format is defined as "TID:[Tag ID]_RID:[Reader Number]," ensuring the uniqueness and resolvability of the identifier. Following this format, the concatenated string is "TID:Tag789_RID:R002." This string will serve as the unique identifier for this specific abnormal signal event. Finally, the submodule generates a specific abnormal signal identification record based on this unique identifier, which is then stored or passed to the subsequent processing module.

[0071] In the subtraction operation, the power value is in dBm, and the reader number is bound to the shelf area code.

[0072] Please see Figure 3 The inventory area mapping module includes:

[0073] The frequency template library extraction submodule receives the abnormal signal identifier "TID:Tag789_RID:R002". First, the submodule performs precise parsing of this identifier, locating and extracting the tag code "Tag789" based on the "TID:" field in the preset format. Next, the submodule uses this tag code "Tag789" to perform a search operation in the pre-stored frequency template library. This frequency template library is a structured database table that stores the baseline values ​​of signal frequency characteristic parameters for each registered tag within various predefined physical areas, derived from statistical analysis of a large amount of historical normal communication data. The goal of the search operation is to find the regional baseline frequency data that matches the tag "Tag789" and is related to the physical area (and its adjacent potential areas) corresponding to the reader "R002" that triggered the abnormal signal. The query results show that the tag "Tag789" has pre-stored baseline frequency data in areas A01, A02, and A03. Each data point is composed of multi-dimensional frequency components, including the center frequency, frequency offset, and two independent multipath effect compensation parameters (denoted as Multipath Effect Compensation A and Multipath Effect Compensation B, respectively). The specific reference frequency data is as follows: The reference frequency data vector for region A01 is:

[0074] [2410MHz (center frequency), 12kHz (frequency offset), 0.75 (multipath compensation A), 0.15 (multipath compensation). The reference frequency data vector for region B) A02 is:

[0075] [2430MHz (center frequency), 10kHz (frequency offset), 0.60 (multipath compensation A), 0.20 (multipath compensation). Region B]

[0076] The reference frequency data vector for A03 is:

[0077] [2450MHz (center frequency), 11kHz (frequency offset), 0.65 (multipath compensation A), 0.18 (multipath compensation). Sub-module B) aggregates the multi-dimensional frequency components corresponding to each region extracted from the template library and organizes them into a set containing multiple regions and their corresponding reference frequency vectors. This set is the regional reference frequency set, which is used for subsequent regional distance calculations.]

[0078] Table 1: Reference Frequency Parameters for Tag Area

[0079] Tag Encoding Region ID Dimension Name Reference frequency component value unit Tag789 A01 Center frequency 2410 MHz Tag789 A01 Frequency offset 12 kHz Tag789 A01 Multipath effect compensation A 0.75 - Tag789 A01 Multipath effect compensation B 0.15 - Tag789 A02 Center frequency 2430 MHz Tag789 A02 Frequency offset 10 kHz Tag789 A02 Multipath effect compensation A 0.60 - Tag789 A02 Multipath effect compensation B 0.20 - Tag789 A03 Center frequency 2450 MHz Tag789 A03 Frequency offset 11 kHz Tag789 A03 Multipath effect compensation A 0.65 - Tag789 A03 Multipath effect compensation B 0.18 -

[0080] Table 1 details the four-dimensional reference frequency parameters of tag 789 in three different predefined regions (A01, A02, A03). These parameters form the basis for subsequent regional distance calculations.

[0081] The regional distance calculation submodule calls the regional reference frequency set generated by the previous submodule. Simultaneously, it collects the currently measured frequency multi-dimensional components of tag 789 from reader R002 (which last detected tag 789 and triggered an anomaly signal) via the communication interface. The currently collected frequency component vector c is: center frequency c1 = 2432MHz, frequency offset c2 = 9kHz, multipath effect compensation parameter Ac3 = 0.58, and multipath effect compensation parameter Bc4 = 0.22. Therefore, the total number of frequency dimensions N = 4.

[0082] The system presets the frequency difference weights for each dimension w. i The weights were determined through feature importance analysis on a training dataset containing 10,000 labeled samples (historical measurement frequency vectors, actual region labels). Specifically, an ensemble learning method based on decision trees (such as gradient boosting decision trees or random forests) was used to train a region classification model on this dataset. Then, the contribution metrics of each frequency dimension to the classification decision (such as the reduction in average impurity or the importance of feature permutation) were extracted. Taking the random forest model as an example, the analysis showed that the contribution of the center frequency was 0.38, the contribution of the frequency offset was 0.31, the contribution of multipath parameter A was 0.19, and the contribution of multipath parameter B was 0.12. These raw contributions were normalized (each item divided by the sum 0.38 + 0.31 + 0.19 + 0.12 = 1.00) to obtain the final weight coefficients: center frequency weight w1 = 0.38, frequency offset weight w2 = 0.31, multipath parameter A weight w3 = 0.19, and multipath parameter B weight w4 = 0.12. These weight values ​​ensure that ∑w i =1, which reflects the relative importance of different frequency components in distinguishing physical regions.

[0083] Signal fluctuation standard deviation σ ki It is obtained through statistical calculations based on a large amount of historical signal stability data for each frequency component i in each region k. For the center frequency of region A02 (its reference value b)... A02,1 =2430MHz), 200 historical center frequency readings of Tag789 under stable conditions in this region were collected, for example, the sequence [2428.5, 2431.0, 2430.5, ..., 2429.8MHz]. The arithmetic mean of these readings was calculated to be 2430.05MHz, and then the standard deviation formula was used. (where M = 200 is the sample size), the historical standard deviation σ of the center frequency in this region is calculated. A02,1 = 2.0MHz. Similarly, other standard deviations for region A02 are obtained: standard deviation of frequency offset σ A02,2 =1.0kHz, multipath parameter A standard deviation σ A02,3=0.05, standard deviation σ of multipath parameter B A02,4 =0.02. The same method was used to calculate σ for other regions A01 and A03. ki Value: σ A01 For [2.5MHz, 1.5kHz, 0.06, 0.03], σ A03 [2.2MHz, 1.2kHz, 0.04, 0.025].

[0084] Current inventory capacity value C cur Retrieved from the real-time inventory management system, currently at 80 units. Capacity limit C for each region. kmax Based on the physical space and pre-set fixed values ​​in the planning, the C value of area A01 A01max =100 units, C of region A02 A02max =120 units, C in region A03 A03max = 100 units.

[0085] This submodule uses formulas Calculate the overall distance value for the region. To ensure the accuracy of the calculation and the reasonableness of the contribution of each dimension, the current frequency component c of each dimension i in the formula is used. i Regional reference frequency component b ki and signal fluctuation standard deviation σ ki All data uses a standardized unit specific to each dimension. For example, center frequency (dimension 1) is uniformly expressed in MHz, frequency offset (dimension 2) in kHz, and multipath compensation parameters (dimensions 3 and 4) are unitless ratios. If the unit of the raw data for a certain dimension does not match the preset unit, a unit conversion is performed first. The conversion rules are set as follows: if the center frequency data is acquired in GHz, it is multiplied by 1000 to convert it to MHz (because 1 GHz = 1000 MHz); if the frequency offset data is acquired in MHz, it is multiplied by 1000 to convert it to kHz (because 1 MHz = 1000 kHz). This unit standardization process ensures... This term is a unitless normalized difference for each dimension i.

[0086] Taking region A02 as an example, its reference frequency component b A02 [2430MHz, 10kHz, 0.60, 0.20].

[0087] c1 = 2432MHz, b A02,1 =2430MHz,σ A02,1 =2.0MHzc2=9kHz,b A02,2 =10kHz,σ A02,2 =1.0kHz c3=0.58,bA02,3 =0.60,σ A02,3 =0.05c4=0.22,b A02,4 =0.20,σ A02,4 =0.02;

[0088]

[0089] Calculate the reference frequency b for region A01. A01 =[2410MHz,12kHz,0.75,0.15],σ A01 =[2.5MHz,1.5kHz,0.06,0.03].

[0090]

[0091] D A01 = (0.38·8.8+0.31·2.0+0.19·2.8333+0.12·2.3333)·0.8;

[0092] D A01 = (3.344 + 0.62 + 0.5383 + 0.28) · 0.8;

[0093] D A01 =4.7823·0.8≈3.8258;

[0094] Calculations are performed on region A03, and its reference frequency b is... A03 =[2450MHz,11kHz,0.65,0.18],σ A03 =[2.2MHz,1.2kHz,0.04,0.025].

[0095]

[0096] D A03 = (0.38·8.1818+0.31·1.6667+0.19·1.75+0.12·1.6)·0.8;

[0097] D A03 = (3.1091 + 0.5167 + 0.3325 + 0.192) · 0.8;

[0098] D A03 =4.1503·0.8≈3.3202;

[0099] After traversing all regional reference frequencies and performing the formula calculation, a set of regional distances is generated as {[A01:3.8258],[A02:0.5907],[A03:3.3202]}. The advantage of this formula is that by introducing weighted normalized differences of multi-dimensional frequency components and a storage capacity factor, it can comprehensively evaluate the similarity between the current signal characteristics and the reference characteristics of each region, while also taking into account the real-time availability of physical storage space in each region. The results show that the smaller the comprehensive distance value, the closer the current signal characteristics are to the reference characteristics of that region, and the more sufficient the storage capacity of that region, thus resulting in a higher matching degree.

[0100] The candidate region filtering submodule receives the region distance set generated by the previous submodule, namely {[A01:3.8258],[A02:0.5907],[A03:3.3202]}. This submodule iterates through each entry in this set, extracts the comprehensive distance value of each region, and performs a systematic comparison of these values. Specifically, it first sets the distance value of the first region A01 (3.8258) as the current minimum value, corresponding to region A01. Then, it takes the distance value of the next region A02 (0.5907) and compares it with the current minimum value 3.8258. Since 0.5907 < 3.8258, it updates the current minimum value to 0.5907, and the corresponding region is updated to A02. Next, it takes the distance value of the region A03 (3.3202) and compares it with the current minimum value 0.5907. Since 3.3202 > 0.5907, the current minimum value and the corresponding region remain unchanged. After traversing all regions, the minimum value was finally selected as 0.5907, and the region identifier corresponding to this minimum value is "A02". Subsequently, the submodule maps this selected region identifier "A02" to a predefined inventory space code library. This library is a mapping table that precisely associates logical region identifiers (such as A01, A02) with specific physical warehouse locations or spatial coordinate ranges. In this inventory space code library, the inventory space code corresponding to "A02" is "Z02-S03-B01". The submodule finally generates a candidate inventory region identifier "Z02-S03-B01", which will be passed to subsequent modules.

[0101] The multidimensional frequency components include the center frequency of the tag signal, the frequency offset, and the multipath effect compensation parameters;

[0102] Weighting coefficient ω a The standard deviation of signal fluctuation ΔTb is determined by optimization using the training dataset, and is calculated based on historical signal stability data.

[0103] Please see Figure 4 The state trajectory correction module includes:

[0104] The coordinate data acquisition submodule, based on the candidate inventory area identifier "Z02-S03-B01" generated by the previous module, sends a query request to the historical database, retrieving the coordinate data of the three most recent time-series points associated with the tag "Tag789" in that area or its immediate vicinity. The time interval is dynamically configured based on the movement speed of the inventory items: when the item's movement speed is considered high (above 0.5 m / s), the time interval is set to 1 second; when the movement speed is considered low (below 0.1 m / s), the time interval is set to 5 seconds; when the speed is between 0.1 m / s and 0.5 m / s, it is considered medium speed, and the time interval is set to 2 seconds. Currently, the monitored movement speed of the tag-associated item is 0.3 m / s, which is considered medium speed, therefore the time interval is configured to 2 seconds. The query operation retrieves the coordinate records in the database that are closest in timestamp to the tag "Tag789" and whose marked area is "Z02-S03-B01" or its immediate vicinity. The system will extract the X-axis, Y-axis, and Z-axis coordinate components corresponding to each time point. The three time point data obtained (arranged in reverse time, with P3 being the point relative to the current time of 0 seconds) are: P1 (relative to the current time - 4 seconds): (10.50m, 20.20m, 1.50m)

[0105] P2 (relative to current time - 2 seconds): (10.70m, 20.00m, 1.60m) P3 (relative to current time 0 seconds): (10.80m, 19.80m, 1.50m) These three-dimensional coordinate points will be arranged in chronological order (P1, P2, P3) to form a set of three-dimensional coordinate points, thus generating the time-series coordinate sequence Pts. raw =

[0106] [P1(10.50,20.20,1.50),P2(10.70,20.00,1.60),P3(10.80,19.80,1.50)]. For subsequent mean filtering, the system will also collect points preceding this sequence, or sequences with at least 5 points, for example, a longer sequence Pts. extended =

[0107] [P0(10.30,20.40,1.40),P1(10.50,20.20,1.50),P2(10.70,20.00,1.60),

[0108] P3(10.80,19.80,1.50),P4(11.00,19.60,1.60)].

[0109] Table 2: Example of Time-Series Coordinate Data Acquisition

[0110] Timing point identifier X-axis coordinate (m) Y-axis coordinate (m) Z-axis coordinate (m) Timestamp (relative to the current time) P0 10.30 20.40 1.40 -6s P1 10.50 20.20 1.50 -4s P2 10.70 20.00 1.60 -2s P3 10.80 19.80 1.50 0s P4 11.00 19.60 1.60 +2s (prediction or subsequent data collection)

[0111] Table 2 shows the coordinate data of five historical and current time series points collected for subsequent processing.

[0112] The mean filtering submodule obtains the time-series coordinate sequence generated by the previous submodule, using Pts here. extended =

[0113] [P0(10.30,20.40,1.40),P1(10.50,20.20,1.50),P2(10.70,20.00,1.60),

[0114] P3(10.80,19.80,1.50), P4(11.00,19.60,1.60)]. It defines a sliding window with a fixed size to cover the coordinates of three adjacent time points. The coordinate components within the window are weighted equally. The sliding window's step size is set to one time point. For point P1 in the sequence, its filtered coordinates P1′ are calculated as follows (window covers P0, P1, P2):

[0115] P1′ x =(P0) x +P1 x +P2 x ) / 3=(10.30+10.50+10.70)

[0116] / 3=31.50 / 3=10.50m;

[0117] P1′ y =(P0) y +P1 y +P2 y ) / 3=(20.40+20.20+20.00) / 3=60.60 / 3=20.20m;

[0118] P1′ z =(P0) z +P1 z +P2 z ) / 3=(1.40+1.50+1.60) / 3=4.50 / 3=1.50m;

[0119] Therefore, P1′ = (10.50, 20.20, 1.50). For point P2 in the sequence, its filtered coordinates P2′ are calculated as follows (window covers P1, P2, P3):

[0120] P2′ x =(P1) x +P2 x +P3 x) / 3=(10.50+10.70+10.80) / 3=32.00 / 3≈10.67m;

[0121] P2′ y =(P1) y +P2 y +P3 y ) / 3=(20.20+20.00+19.80) / 3=60.00 / 3=20.00m;

[0122] P2′ z =(P1) z +P2 z +P3 z ) / 3=(1.50+1.60+1.50) / 3=4.60 / 3≈1.53m;

[0123] Therefore, P2′=(10.67,20.00,1.53). For point P3 in the sequence, its filtered coordinates P3′ are calculated as follows (window covers P2, P3, P4):

[0124] P3′ x =(P2) x +P3 x +P4 x ) / 3=(10.70+10.80+11.00) / 3=32.50 / 3≈10.83m;

[0125] P3′ y =(P2) y +P3 y +P4 y ) / 3=(20.00+19.80+19.60) / 3=59.40 / 3=19.80m;

[0126] P3′ z =(P2) z +P3 z +P4 z ) / 3=(1.60+1.50+1.60) / 3=4.70 / 3≈1.57m; therefore, P3′=(10.83,19.80,1.57). The first point P0 and the last point P4 of the sequence are not filtered in this example, or boundary processing techniques (such as copying boundary points or using an adjusted window) are used. The filtered coordinate sequence Pts is obtained after processing. filtered = [P0,P1′,P2′,P3′,P4] (where P0 and P4 are the original points, only P1, P2, and P3 are updated) or a sequence of length 3 [P1′,P2′,P3′], depending on the boundary handling strategy. In this example, we use the filtered point sequence Pts. filtered= [P1′(10.50,20.20,1.50),P2′(10.67,20.00,1.53),P3′(10.83,19.80,1.57)], with corresponding timestamps of -4s, -2s, 0s.

[0127] The interpolation sequence generation submodule calls the filtered coordinate sequence Pts obtained after processing by the previous submodule. filtered =[P1′(-4s:10.50,20.20,1.50),P2′(-2s:10.67,20.00,1.53),

[0128] P3′(0s:10.83,19.80,1.57)]. This submodule selects adjacent time points in this sequence as nodes for interpolation calculation. Taking P2′ and P3′ as interpolation nodes as an example, the time starts from t A =-2s to t B =0s. The coordinate component values ​​are used as boundary conditions: X(-2) = 10.67, Y(-2) = 20.00, Z(-2) = 1.53 and X(0) = 10.83, Y(0) = 19.80, Z(0) = 1.57. To satisfy the condition of continuity of the first derivative at the node, it is necessary to calculate or estimate the value of the first derivative (velocity) at the node. The velocity at P2′ is estimated using the central difference method (data from P1′ and P3′ are required):

[0129] v P2′,x =(X P3′ -X P1′ ) / (t P3′ -t P1′ )=(10.83-10.50) / (0-(-4))=0.33 / 4=0.0825m / s;

[0130] v P2′,y =(Y P3′ -Y P1′ ) / (t P3′ -t P1′ )=(19.80-20.20) / 4=-0.40 / 4=-0.10m / s;

[0131] v P2′,z =(Z P3′ -Z P1′ ) / (t P3′ -t P1′ )=(1.57-1.50) / 4=0.07 / 4=0.0175m / s;

[0132] For boundary points P1′ and P3′, their derivatives can be obtained using forward or backward differencing, or set to 0 (if it is assumed that the sequence is stationary at the beginning / end). Assuming P3′ is the end of the sequence, its velocity is estimated using backward differencing: v P3′,x =(X P3′ -X P2′ ) / (t P3′ -t P2′ )=(10.83-10.67) / (0-(-2))=0.16 / 2=0.08m / s;

[0133] v P3′,y =(Y P3′ -Y P2′ ) / (t P3′ -t P2′ )=(19.80-20.00) / 2=-0.20 / 2=-0.10m / s;

[0134] v P3′,z =(Z P3′ -Z P2′ ) / (t P3′ -t P2′ )=(1.57-1.53) / 2=0.04 / 2=0.02m / s;

[0135] Solve the cubic polynomial Q(t) = at independently for each dimension X, Y, and Z in the interval [-2, 0]. 3 +bt 2 The coefficients a, b, c, and d of +ct+d. Taking the X dimension as an example, in the time interval [-2, 0] seconds, solve for the coefficients that satisfy the four conditions: X(-2) = 10.67, X(0) = 10.83, X′(-2) = 0.0825, X′(0) = 0.08. Once the coefficients are determined, the submodule supplements the missing coordinate component values ​​at these time points by selecting new time points (such as t = -1.5s, -1.0s, -0.5s) in the range t ∈ (-2, 0) and substituting them into the polynomials X(t), Y(t), and Z(t). For example, interpolation is performed at t = -1s (i.e., the time between P2′ and P3′), and the resulting coordinate point (let's assume it's P) is obtained. interp The distances are (10.75m, 19.90m, 1.55m). These newly generated interpolation points are connected to the original filter nodes in chronological order to form a more temporally dense and spatially continuous and smooth trajectory, ultimately generating a smooth inventory coordinate sequence, such as [P1′, P...]. interp1 ,P2′,P interp2 ,P3′].

[0136] The time interval of the time-series coordinate sequence is 1 to 5 seconds, and the specific value is dynamically configured according to the inventory movement speed;

[0137] The sliding window moves in one time step, and the weights of the coordinate components within the window are equally weighted.

[0138] The coefficients of the cubic polynomial satisfy the condition of continuity of the first derivative at the nodes.

[0139] Please see Figure 5 The inventory movement determination module includes:

[0140] The direction extraction submodule is based on the smoothed inventory coordinate sequence generated by the previous module. This sequence consists of a series of coordinate points arranged in chronological order, such as [S1(-2.0s:10.67,20.00,1.53),S2(-1.0s:10.75,19.90,1.55),S3(0.0s:10.83,19.80,1.57)], where the time interval Δt = 1.0 second (determined by the interpolation density). This submodule sequentially extracts two consecutive coordinate points from the sequence. First, it extracts S1 and S2, and then calculates the displacement vector from the current coordinate point S1 to the next coordinate point S2. Then calculate the instantaneous velocity vector. Since Δt = 1.0s:

[0141] The velocity vector is calculated with an accuracy of 0.01 m / s (two decimal places), and the vector direction angle is calculated using inverse trigonometric functions, achieving a resolution of 1 degree. Next, S2 and S3 are extracted to calculate the displacement vector. Instantaneous velocity vector The submodule performs this vector calculation for each pair of consecutive points in the entire smoothed inventory coordinate sequence, resulting in a series of inventory state movement direction vectors arranged in chronological order. This sequence is the value of the direction vector sequence, which is here.

[0142] The direction comparison submodule uses the direction vector sequence values ​​obtained from the previous submodule. The currently processed inventory movement direction vector is... Meters per second. This submodule calls a preset library of legal movement directions from the system. This library is defined according to the passage rules of the warehouse aisle or work area where the current item is located, and contains unit vectors of all allowed movement directions at that location. For the aisle where the current tag Tag789 is located (assuming it mainly extends along the positive X-axis, the Y-axis is horizontal, and the Z-axis is vertically upward), the legal movement directions are set as follows: (Proceed along the passageway) (Small adjustments to the left are allowed) (Small adjustments to the right are permitted) (Small upward adjustments are allowed) (Small downward adjustments are allowed) These legal direction vectors are set during the system configuration phase through path planning analysis of warehouse CAD drawings and in conjunction with actual logistics operation processes. The submodule specifies the current inventory movement direction vector. With each legal direction unit vector in the legal movement direction library Perform dot product operation.

[0143] After obtaining these dot product values, the submodule filters the dot product results. When the dot product value is less than zero, it indicates that the angle between the current movement direction vector and the specific legal direction unit vector is greater than 90 degrees. In this example, and A movement vector is considered valid if the dot product of its vector and all valid direction vectors in the library is not less than zero. A movement vector is considered valid if at least one dot product is less than zero. The entire movement is judged as an illegal directional movement. Here, due to the presence of a negative dot product, The condition was deemed illegal. The submodule records the specific circumstances leading to the illegal determination; for example, the illegal direction is relative to... (Moved too far to the left or incorrectly) and (Move downwards). Finally, an illegal direction determination quantity is generated, indicating... It is an illegal movement vector.

[0144] The alarm generation submodule is based on the illegal direction determination value output by the previous submodule. It detects a movement vector between time points t1 = -2.0s and t2 = -1.0s. Because it is consistent with the preset legal direction and The dot product is negative and therefore marked as illegal. The submodule first uses this determination to locate the specific inventory movement event data that triggered this illegal determination. This includes the time period of the movement event (from -2.0s to -1.0s), the related tag ID "Tag789", and the starting coordinates S1 (10.67, 20.00, 1.53) and ending coordinates S2 (10.75, 19.90, 1.55) of the movement. Next, the submodule analyzes this illegal vector... Match and record the start and end points (S1 and S2, or corresponding time periods) in the original smoothed coordinate sequence. Record this information, including the event timestamp (take the midpoint of the interval or the start time, such as -2.0s), tag ID, and illegal movement vector. The component values, the detected illegal referenced legal directions (e.g.) The entry, along with the starting coordinates S1 at the time of the movement, is combined to generate a structured illegal event log entry. The entry format is: "[Timestamp:2025-05-1410:30:03Z]EventID:EVT001,TagID:Tag789,IllegalMoveVector:(0.08,-0.10,0.02)m / s,StartCoord:(10.67,20.00,1.53)m,ViolatedRefs:[L2,L5]". Then, the submodule calls the database interface service to write this log entry to the specified illegal event database table or system log file. Simultaneously, the submodule executes a trigger update operation, which sends an immediate alarm notification to the central monitoring system, containing the key information of the log entry. Furthermore, the system updates the status indicator associated with the tag "Tag789", marking its current status as "illegal movement alarm", and increments an illegal movement event counter associated with that tag. This process ultimately generates and records the values ​​and related statuses of illegal movement alarm entries.

[0145] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An RFID-based inventory tracking management system, characterized by, The system includes: The signal difference verification module is used to obtain the tag receiving power value of adjacent shelf areas through the reader configured in the shelf area, calculate the power difference between the tag receiving power value recorded by the reader in the current area and the tag receiving power value recorded in the previous area, compare the power difference with the power attenuation threshold between shelves, generate an abnormal inventory location signal, associate the tag ID with the reader number to generate an abnormal signal identifier, and transmit it to the inventory area mapping module. The inventory area mapping module is used to extract the frequency template library corresponding to the tag based on the abnormal signal identifier, match the current frequency of the tag with the reference frequency of the frequency template library through the Manhattan distance algorithm, filter the reference frequency data of the area with the smallest distance to the current frequency of the tag in the frequency template library, generate candidate inventory area identifiers, and pass them to the state trajectory correction module. The state trajectory correction module is used to call the coordinates of three historical time series points through the candidate inventory area identifier, perform sliding window mean filtering and cubic spline interpolation to generate a smooth inventory coordinate sequence, and pass it to the inventory movement determination module.

2. The RFID-based inventory tracking management system according to claim 1, characterized in that, The abnormal signal identifiers include power difference, power attenuation threshold between shelves, tag ID, and reader number. The candidate inventory area identifiers include area number, reference frequency, and minimum distance value. The smoothed inventory coordinate sequence specifically includes time-series point coordinates, filtering coefficients, and interpolation parameters.

3. The RFID-based inventory tracking management system according to claim 2, characterized in that, The power difference is the absolute value of the dBm difference between the tag received power in the current area and the previous area; The unit of the power attenuation threshold between shelves is dBm, and the power attenuation threshold between shelves is determined by fitting experimental data. The Manhattan distance algorithm introduces normalized weight coefficients for multi-dimensional frequency components, which are determined by optimization using the gradient descent method. The sliding window mean filter has a window size of 3 time points, and the number of nodes in the cubic spline interpolation satisfies the Nyquist sampling theorem constraint.

4. The RFID-based inventory tracking management system according to claim 3, characterized in that, The signal difference verification module includes: The power difference calculation submodule obtains the tag receiving power values ​​of adjacent shelf areas, extracts the tag receiving power value recorded by the reader in the current area and the tag receiving power value recorded in the previous area, performs a subtraction operation on the two sets of power values, and obtains the power attenuation difference. The attenuation threshold comparison submodule calls the power attenuation difference to compare with the power attenuation threshold between shelves. When the absolute value of the power attenuation difference exceeds the power attenuation threshold between shelves, an abnormal judgment state is marked. When the absolute value of the power attenuation difference does not exceed the threshold, a normal judgment state is marked. The abnormal signal generation submodule extracts the tag ID and the corresponding reader number that trigger the abnormal judgment state, and concatenates the tag ID and reader number into a string according to a preset format to serve as the unique identifier of the abnormal signal, thereby generating an abnormal signal identifier. In the subtraction operation, the power value is in dBm, and the reader number is bound to the shelf area code.

5. The RFID-based inventory tracking management system according to claim 4, characterized in that, The inventory area mapping module includes: The frequency template library extraction submodule parses the abnormal signal identifier, extracts the tag ID, retrieves regional reference frequency data that matches the current frequency of the tag in the pre-stored frequency template library, aggregates multi-dimensional frequency components, and generates a set of regional reference frequencies. The regional distance calculation submodule calls the regional reference frequency set, collects the multi-dimensional frequency components of the tag's current frequency, and uses the following formula: ； Calculate the comprehensive distance value of the region, iterate through all regional reference frequencies, perform formula calculations, and generate a set of regional distances; in, Representative area The overall distance value, The current frequency of the tag dimensional components, For the region Reference frequency dimensional components, For the first Weights of the current frequency differences of the labels of the dimensional components. This represents the current inventory capacity. For the region Capacity limit, For the first Standard deviation of signal fluctuation, The total number of frequency dimensions. Total number of signal categories; The candidate region filtering submodule traverses the set of region distances, compares the comprehensive distance values ​​of multiple regions, selects the region identifier corresponding to the minimum value, maps it to the inventory space coding library, and generates candidate inventory region identifiers. The multidimensional frequency components include the center frequency, frequency offset, and multipath effect compensation parameters of the tag signal; The weighting coefficient The standard deviation of the signal fluctuation was determined through optimization using the training dataset. Calculated based on historical signal stability data.

6. The RFID-based inventory tracking management system according to claim 5, characterized in that, The state trajectory correction module includes: Based on the candidate inventory area identifier, the coordinate data acquisition submodule calls the coordinate data of three historical time series points, extracts the X-axis, Y-axis and Z-axis coordinate components corresponding to each time series point, arranges them into a three-dimensional coordinate point set in chronological order, and generates a time series coordinate sequence. The mean filtering submodule obtains the time-series coordinate sequence, defines a sliding window to cover the coordinates of three adjacent time-series points, sums the X-axis coordinate components within the window and divides them by the window size, performs the same operation on the Y-axis and Z-axis, replaces the coordinates of the window center point with the calculation results, traverses all time-series points to complete the coordinate component update, and generates the filtered coordinate sequence. The interpolation sequence generation submodule calls the filtered coordinate sequence, selects adjacent time series points as interpolation nodes, uses the X, Y, and Z components at the nodes as boundary conditions, calculates the coefficients of a cubic polynomial, supplements the component values ​​of missing coordinate points between nodes, connects the interpolation points with the original nodes to form a continuous trajectory, and generates a smooth inventory coordinate sequence. The time interval of the time-series coordinate sequence is 1 to 5 seconds, and the specific value is dynamically configured according to the inventory movement speed. The sliding window moves in one time step, and the weights of the coordinate components within the window are equally weighted. The coefficients of the cubic polynomial satisfy the condition of continuity of the first derivative at the nodes.

7. The RFID-based inventory tracking management system according to claim 6, characterized in that, The system also includes: The inventory movement determination module is used to extract the direction vectors of two consecutive coordinate points based on the smoothed inventory coordinate sequence, perform a dot product operation with the preset legal movement direction library vector, and generate an illegal movement alarm event when the dot product result is negative. The illegal movement alarm event is written to the log and triggers a database update.

8. The RFID-based inventory tracking management system according to claim 7, characterized in that, The illegal movement alarm event includes the direction vector, dot product result, and alarm level.

9. The RFID-based inventory tracking management system according to claim 8, characterized in that, The preset legal movement direction library is generated based on the direction vector clustering of historical compliant movement trajectories.

10. The RFID-based inventory tracking management system according to claim 9, characterized in that, The inventory movement determination module includes: The direction extraction submodule extracts two consecutive coordinate points based on the smoothed inventory coordinate sequence, calculates the vector value between the current coordinate point and the next coordinate point, arranges each pair of coordinate vectors in chronological order, obtains the inventory status movement direction vector sequence, and obtains the direction vector sequence value. The direction comparison submodule calls all direction vectors in the preset legal movement direction library according to the direction vector sequence value, performs dot product operation for each group of stock movement direction vectors, obtains the dot product calculation value, filters the dot product results, and marks the direction vectors with dot product values ​​less than zero as illegal vectors to obtain the illegal direction judgment quantity. The alarm generation submodule locates the corresponding inventory movement event data based on the illegal direction determination quantity, matches and records the coordinate index position of the illegal vector, generates illegal event log entries from the matching results and calls the database write operation, and simultaneously executes the trigger update operation to generate illegal movement alarm entry values. The calculation accuracy of the direction vector is 0.1 m / s, and the resolution of the vector direction angle is 1 degree. When the dot product result is negative, it is determined that the angle between the moving direction and the legal direction exceeds 90 degrees.

Images