RFID Signal Completion and Anti-Interference Method and System Based on Contrastive Learning in Small-Space Dense Storage Environment

Abstract

The present invention belongs to the technology of enhancing RFID signal quality, and is a method and system for RFID signal completion and anti-interference based on contrast learning in a small-space dense storage environment. The method includes the steps of: optimizing the automated reading of RFID signals, so as to optimize radio frequency identification in real time according to environmental changes and achieve adaptive reading of RFID tags; separating and extracting RFID signals, separating the strongly coupled part in the read RFID signals to reduce the generation of incorrect data due to the collision and coupling of RFID signals with each other; training a contrast learning model based on a convolutional neural network according to the spatial position relationship between historical RFID signals and RFID tags, so that the contrast learning model can learn the pattern and relationship features of RFID signals and complete incomplete RFID signals. The present invention can effectively solve the problem of missed or misread data caused by RFID signal collision in a small-space dense storage environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of radio frequency identification signal quality enhancement, and particularly relates to a method and system for RFID signal completion and anti-interference based on contrastive learning in a small-space dense storage environment. Background Art

[0002] With the continuous development of modern logistics and warehousing management technologies, intelligent warehousing has attracted much attention due to its high efficiency and accuracy; as one of the key technologies in intelligent warehousing, radio frequency identification (RFID) technology realizes automatic identification and tracking of items through wireless communication, bringing revolutionary changes to the construction of intelligent warehousing. RFID tags have the advantages of a wide reading range, non-contact identification, a large amount of information storage, high security, etc., and can meet the needs of warehouse managers for material inventory, tracking, refined management, etc.

[0003] However, in a small-space, multi-type, multi-material, small-item dense storage environment, it will interfere with the accurate reading of RFID signals, and there are still deficiencies in the accuracy of information acquisition, such as missed readings and misreadings. In addition, outer packaging materials such as metal and tin foil or liquid items will also cause great interference to the accurate reading of information. Accurately reading RFID signals in a small-space dense environment still faces the following challenges:

[0004] 1) Signal interference problem. In a dense space, multiple RFID tags overlap with each other, resulting in easy signal interference with each other, leading to a decrease in the accuracy and stability of data reading. Metals, tin foil, liquids, etc. will cause different degrees of reflection or attenuation to RFID signals, further exacerbating the signal interference problem.

[0005] 2) Multi-scale recognition problem. In a small-space dense storage environment, the sizes and shapes of items are diverse, and the random placement of items causes different angular deviations between the signal source and the reader / writer, resulting in signals not being accurately read and making it difficult to accurately identify signals at different angles in all directions.

[0006] 3) Complexity problem of data completion. Due to factors such as diverse tag types, uneven signal quality, and tag overlap, it becomes particularly complex to complete incomplete RFID data.

[0007] Existing technologies are difficult to effectively address these challenges, resulting in the accuracy and reliability of RFID data acquisition being affected.

[0008] Although some studies have tried to solve the above problems, most of them are aimed at improving the warehousing environment of batch-scale items in large warehouses. The RFID tag recognition scheme used in this kind of intelligent warehousing of bulk items is difficult to meet the demand for accurate signal reading in a small-space dense storage environment. Summary of the Invention

[0009] To solve the problems existing in the prior art, the present invention proposes a method and system for RFID signal completion and anti-interference based on contrast learning in a small-space dense environment. By using machine learning technology to perform pattern analysis and contrast learning on the collected multi-source data, and enhancing the signal recognition ability through pattern recognition and prediction models, the missing signals are completed and the interference is suppressed, thereby improving the signal quality and reading accuracy, effectively solving the problem of data leakage and misreading caused by RFID signal collision in a small-space dense storage environment, and realizing the efficient management and accurate tracking of small materials under the conditions of small-space multi-category dense storage.

[0010] On the one hand, the technical solution adopted in the embodiment of the present invention is: a method for RFID signal completion and anti-interference based on contrast learning in a small-space dense storage environment, including the following steps:

[0011] S1. Optimize the automatic reading of RFID signals, so as to optimize radio frequency identification in real time according to environmental changes and realize adaptive reading of RFID tags;

[0012] S2. Separate and extract RFID signals, separate the strongly coupled parts in the read RFID signals, and reduce the generation of incorrect data caused by the collision and coupling of RFID signals with each other;

[0013] S3. Train a contrast learning model based on a convolutional neural network according to the historical RFID signal and the spatial position relationship of RFID tags, so that the contrast learning model can learn the pattern and relationship characteristics of RFID signals and complete the incomplete RFID signals.

[0014] Based on the same inventive concept, the system for RFID signal completion and anti-interference based on contrast learning in a small-space dense storage environment provided in the embodiment of the present invention includes the following units:

[0015] A reading optimization unit for optimizing the automatic reading of RFID signals, so as to optimize radio frequency identification in real time according to environmental changes and realize adaptive reading of RFID tags;

[0016] A separation and extraction unit for separating and extracting RFID signals, separating the strongly coupled parts in the read RFID signals, and reducing the generation of incorrect data caused by the collision and coupling of RFID signals with each other;

[0017] A completion unit for training a contrast learning model based on a convolutional neural network according to the historical RFID signal and the spatial position relationship of RFID tags, so that the contrast learning model can learn the pattern and relationship characteristics of RFID signals and complete the incomplete RFID signals.

[0018] The technical solution of the present invention generally adopts a multi-source information fault-tolerant technology and a data completion strategy, effectively overcoming the interference caused by signal overlap, and improving the accuracy and reliability of RFID data reading in a small-space dense storage environment. Compared with the prior art, the technical effects achieved by the present invention specifically include:

[0019] 1. Optimized signal reading design: The anti-collision module of the RFID system is used to manage the communication between multiple RFID tags and the reader simultaneously, avoiding signal conflicts and collisions, with the aim of ensuring that each tag can be accurately and quickly read in a complex reading environment.

[0020] The anti-collision algorithms of existing RFID systems mainly include three categories: ALOHA-based anti-collision algorithms, tree-based anti-collision algorithms, and hybrid algorithms; these existing algorithms cannot adapt to the dynamic and complex changes of the environment.

[0021] The present invention proposes an RFID anti-collision optimization strategy based on MOPSO multi-objective optimization, which can perform radio frequency identification optimization in real time according to environmental changes, realizing adaptive reading of tag data. When the environment is complex or RFID signals are dense, the multi-objective particle swarm optimization (MOPSO) algorithm is used to intelligently process and analyze the acquired RFID data, and the radio frequency of the reader is adjusted in real time.

[0022] 2. Signal separation and extraction algorithm: After the reader extracts the RFID signal and stores it in the initial signal library of the system, due to tag overlap or signal collision, there are mixed data with mutual interference in these signals, and this part of the mixed signal needs to be separated.

[0023] Common signal separation algorithms include ICA, BSS, NMF, etc.; these algorithms can effectively separate mixed signals to a certain extent, but they usually rely on the statistical independence assumption of mixed signals, which does not always hold in practical applications. In addition, these algorithms may be sensitive to noise and outliers, resulting in poor separation effects.

[0024] The present invention proposes a strongly coupled relaxation decomposition technology, using the blind source separation (BSS) algorithm to decompose strongly coupled signals into independent signal sources, effectively solving the problem of incomplete separation existing in existing separation algorithms. In addition, further using graph theory and machine learning algorithms to model RFID signals, regarding the signals as nodes in the graph, and the edges represent the interactions between signals. By analyzing the correlation between signals, strongly coupled signals, that is, signals with serious mutual interference, are identified, and the strongly coupled signals are decomposed into independent signal sources, improving the quality and readability of the signals and reducing the interference between signals.

[0025] 3. Data Feature Learning and Completion Model Based on Contrastive Learning: RFID transmits signals in the form of electromagnetic waves, which belong to a type of energy. Energy will experience energy loss during transmission. Therefore, there is energy loss during the transmission of electromagnetic waves, that is, signal attenuation. Electromagnetic waves propagate in the form of direct wave, reflection, scattering, diffraction, etc. during propagation, and at the same time, an absorption effect will occur when blocked by obstacles. This will cause problems such as missing and omitted signals read by the reader.

[0026] To solve the above problems of missing and omission, etc., the present invention trains a convolutional neural network contrastive learning model to learn and complete the incorrect data with missing parts; by collecting a large number of historical RFID data signals obtained through multiple experiments and decomposition techniques, including tag reading intensity, timestamp, environmental noise, etc., and at the same time combining the spatial position relationship of RFID tags, including data such as the distance, angle, and relative position between tags, a convolutional neural network contrastive learning model is trained to enable it to learn the patterns and trends in RFID data and achieve precise completion of complex RFID signal data. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a schematic flowchart of the RFID signal completion and anti-interference method based on contrastive learning in an embodiment of the present invention;

[0028] Figure 2 is a schematic flowchart of the RFID anti-collision optimized reading based on the MOPSO algorithm in an embodiment of the present invention;

[0029] Figure 3 is a schematic diagram of the RFID signal completion principle in an embodiment of the present invention;

[0030] Figure 4 is a schematic diagram of the system architecture in an application example of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0031] The present invention will be further described in detail below in conjunction with the embodiments and the drawings, but the embodiments of the present invention are not limited thereto.

[0032] Embodiment

[0033] This embodiment provides an RFID signal completion and anti-interference method based on contrastive learning in a small-space dense storage environment, which is applicable to intelligent warehousing based on RFID technology in a small-space multi-category dense storage environment, such as Figure 1 shown, and the specific implementation steps include:

[0034] S1. Optimize the automated reading of RFID signals, so as to optimize radio frequency identification in real time according to environmental changes and achieve adaptive reading of RFID tags.

[0035] In this embodiment, an optimized design is made for the automatic reading of RFID signals. According to the change of the reading angle, the reading and writing radio frequency is adjusted through the MOPSO algorithm to minimize the signal missed reading.

[0036] The anti-collision module of the RFID system is used to manage the communication between multiple RFID tags and the reader simultaneously to avoid signal conflict and collision. Its goal is to ensure that each tag can be accurately and quickly read in a complex reading environment. This embodiment adopts an RFID anti-collision optimization strategy based on multi-objective particle swarm optimization (MOPSO), which can perform radio frequency identification optimization in real time according to environmental changes to achieve adaptive reading of RFID tags. When RFID signals are dense, the MOPSO algorithm is used to intelligently process and analyze the acquired RFID signals, and the radio frequency of the reader is adjusted in real time.

[0037] The RFID anti-collision optimized reading process based on the MOPSO algorithm is as Figure 2 shown, and the specific algorithm execution process is described as follows:

[0038] S11. Define the objective function of the optimization strategy to represent the optimization of the transmission power for reading accuracy, as shown in formula (1),

[0039] min{f0(y,z),f1(y,z),...,f m (y,z)} (1)

[0040] where f0 is the main objective function, representing the improvement rate of RFID signal reading accuracy; f i represents the i-th auxiliary objective function, 1 ≤ i ≤ m, which represents optimizing the transmission power or frequency of the RFID tag reader to reduce signal interference and improve the data acquisition accuracy of specific types of materials.

[0041] S12. Define the decision variables, including two parts: y and z:

[0042] y = {y1,y2,...,y n},z = {z1,z2,...,z n}(2)

[0043] where y represents the configuration parameters of the RFID system in a dense small space, including the layout of the RFID tag reader, the angle and position of the antenna; z represents the variables related to different types of materials or storage areas, including information such as the type of materials, tag characteristics, and position in the small space.

[0044] S13. Define the constraint conditions of the decision variables, as shown in formulas (3) and (4), to ensure that the decision variables meet the necessary constraint conditions:

[0045] g j ′(y, z) ≤ 0, j = 1, …, r (3)

[0046] h j ′(y, z) ≤ 0, j = 1, …, s (4)

[0047] S14. By performing weighted summation on the characteristic functions of the signal sources, approximately representing the read RFID signals, and realizing the effective extraction of multi-source signals, which is specifically expressed as shown in formula (6):

[0048]

[0049] where N represents the number of signal sources, w i is the weight of the i-th signal source, Φ i represents the characteristic function of the i-th signal source, and x represents the read RFID signal.

[0050] The RFID tags are deployed on the items, and the tag deployment strategy needs to ensure coverage of all storage units. In this embodiment, RFID tags of type NXPucode8 / ImpimjR6 are deployed on the items in the warehouse, with a size of 50x25mm and a chip power of 10dbm. The RFID reader is configured as follows: install 4 RFID readers of model MTK6763 - 8-core, which are respectively installed at the entrance, exit, and two corners of the warehouse to ensure that the RFID readers cover the warehouse comprehensively; the reader angle is set to 30 degrees to achieve the best reading effect. The mechanism of data reading is: the system acquisition frequency is once per second, and the acquisition points are evenly distributed to ensure that each storage unit can be covered. The synchronization mechanism between the RFID reader and the database adopts real-time synchronization, and the data format is uniformly JSON.

[0051] S2. Separate and extract the RFID signals, separate the strongly coupled part in the read RFID signals, and reduce the generation of incorrect data due to mutual collision and coupling of RFID signals.

[0052] In this embodiment, a strong coupling relaxation decomposition technology is specifically used, and the blind source separation (BSS) algorithm is adopted to decompose the strongly coupled part in the read RFID signals into independent signal sources, effectively solving the problem of incomplete separation existing in the prior art.

[0053] When separating and extracting the RFID signals in this embodiment, graph theory and machine learning algorithms are used to model the RFID signals, regarding the signals as nodes and the edges representing the interactions between signals. By analyzing the correlation between RFID signals, strongly coupled signals, that is, signals with serious mutual interference, are identified, and the strongly coupled signals are decomposed into independent signal sources to improve the signal quality and readability and reduce the interference between signals. The specific algorithm flow is as follows:

[0054] S21. Let the blind source separation (BSS) algorithm be independent component analysis (ICA). The objective function of independent component analysis can be expressed as:

[0055]

[0056] where W is the separation filter matrix, w i is the weight of the i-th signal source, represents the column vector of the separation filter matrix, and S is the covariance matrix of the RFID signal. In a small-space multi-category dense storage environment, this objective function is used to separate independent source signals from the RFID signal and reduce the interference between signals.

[0057] S22. Use the short-time Fourier transform (STFT) to extract the time-frequency features of the RFID signal, and analyze the correlation between RFID signals based on the extracted time-frequency features.

[0058] The time-frequency features are specifically expressed as follows:

[0059]

[0060] where X(t, f) is the time-frequency representation of the RFID signal, x RFID (τ) is the original RFID signal, w(t - τ) is the window function, t represents the time variable, and f represents the frequency variable.

[0061] Feature extraction is used to identify the unique attributes of the signal. When processing RFID signals, the short-time Fourier transform can extract the features of the signal in the time-frequency domain, which helps to identify and separate strongly coupled signals.

[0062] S3. According to the spatial position relationship between historical RFID signals and RFID tags, train the contrast learning model based on the convolutional neural network so that the contrast learning model can learn the patterns and relationship features of RFID signals and complete the incomplete RFID signals.

[0063] When data is lost or missed, analyze the patterns and reasons for data loss, construct a prediction model (linear regression, decision tree, neural network, etc.) to estimate the possible values of the missing data; use machine learning algorithms (k-nearest neighbor, SVM support vector machine, etc.) to predict the missing values and use them to complete the missing values.

[0064] In this embodiment, to solve the problem of missed RFID signal readings caused by storage media such as metals and liquids, a contrastive learning model based on the convolutional neural network (CNN) is used to learn and complete the missing data. By collecting a large amount of historical RFID signals, including tag reading intensity, timestamp, environmental noise, etc., and combining the spatial position relationships of RFID tags, including the distance, angle, and relative position between tags, etc., a contrastive learning model based on the convolutional neural network is trained to enable it to learn the patterns and relationship features of RFID signals and achieve precise completion of complex RFID signals. As Figure 3 shown, the specific execution process is introduced as follows:

[0065] S31. In the convolutional neural network (CNN), the convolutional layer performs a convolution operation on the input historical RFID signals by applying a series of learnable filters. The convolution operation is expressed as:

[0066]

[0067] where f represents the input historical RFID signals, including information such as RFID tag reading intensity, timestamp, environmental noise, etc., as well as spatial position relationship data of RFID tags, such as the distance, angle, and relative position between tags. g represents the learnable convolution kernel. By adjusting the parameters of the convolution kernel, the convolutional neural network (CNN) can learn the patterns and relationship features in the RFID data.

[0068] S32. The pooling layer is used to reduce the spatial size of the feature map of the RFID signal, thereby reducing the number of parameters and the amount of computation.

[0069] Among them, max pooling can be expressed as:

[0070] maxpooling(x)=max i ∈Nx i (10)

[0071] When processing RFID signals, the pooling operation is used to reduce the spatial size of the feature map, reduce the number of parameters and the amount of computation. For the input feature map, among the set of elements N within the pooling window, the maximum value is taken as the result after pooling to extract the main features of the RFID signal and reduce the data dimension at the same time. In this embodiment, a non-linear activation function (ReLU) is introduced into the convolutional neural network to enhance the expressive power of the model, which is specifically expressed as:

[0072] ReLU(x RFID )=max(0,x) (11)

[0073] For the input x RFID , if x RfID is greater than 0, then x RFIDitself; if x RFID is less than or equal to 0, then output 0. This can enable the CNN to learn more complex pattern and relationship features, improving the processing and analysis capabilities for RFID signals.

[0074] S33. Train a contrastive learning model based on a convolutional neural network.

[0075] When training the model, use the mean squared error (MSE) as the loss function to measure the difference between the model's predicted value and the actual value. The mean squared error (MSE) is specifically expressed as:

[0076]

[0077] where n represents the number of samples. When training the model for data completion, use the MSE loss function to measure the difference between the model's predicted value and the true value y i . By minimizing the MSE loss function, adjust the model's parameters so that the model can more accurately predict the values of the missing data.

[0078] During the training process, use the backpropagation algorithm to calculate the gradient of the loss function with respect to the network parameters. Specifically, calculate the gradient of each parameter with respect to the loss function through the chain rule. Then, combined with the gradient descent algorithm, update the network weights according to the direction and magnitude of the gradient, gradually optimizing the model performance. The trained model can predict the values of the missing data. This usually involves an autoencoder structure, where the encoder part compresses the input into a low-dimensional representation, and the decoder part reconstructs this representation into the original data. The goal is to minimize the reconstruction error, including both the encoder and decoder parts. The encoder compresses the input into a low-dimensional representation as follows:

[0079] z = Encoder(x)(13)

[0080] where x is the input RFID data and z is the low-dimensional representation.

[0081] The decoder reconstructs the low-dimensional representation into the original data as follows:

[0082]

[0083] The goal is to minimize the reconstruction error, usually using MSE or other relevant loss functions to measure the reconstruction error:

[0084]

[0085] By iteratively training the autoencoder, enable the model to learn the pattern and relationship features in the RFID signal, thereby achieving accurate completion of RFID signal data caused by metal or liquid.

[0086] The signal completion and interference resistance method of this embodiment firstly adopts a multi-dimensional signal acquisition method when reading signals to address the problems of dense distribution of RFID tags and signal overlap in a small space dense environment, optimizes the tag attachment medium, and performs radio frequency optimization in real time according to environmental changes; secondly, the strong coupling relaxation decomposition technology and blind source separation algorithm are used to model and separate the read RFID signals to improve signal quality and readability; finally, for incomplete data, a multi-scale linkage data fusion completion method based on MOPSO multi-objective optimization and contrast learning is used to train a convolutional neural network to learn data characteristics, thereby achieving data completion of incomplete RFID signals.

[0087] After performing signal completion and anti-interference processing, this embodiment can also perform intelligent data analysis and processing on the RFID data after data completion and fault tolerance processing to provide intelligent services.

[0088] The realization of intelligent services includes functions such as automatic classification, package location, usage prediction, and expiration date management; interacting with the user interface, such as providing search, screening, statistics and other operations. Specifically including:

[0089] Data hot backup technology: implement hot backup mechanism to ensure real-time replication and backup of data;

[0090] The database backup solution is regular backup, and the backup is triggered by data changes; the data storage solution includes local storage, remote data centers, and cloud storage data centers, with storage location performance optimization and cost balance;

[0091] Breakpoint recovery mechanism: detect hardware, software failures or network anomalies and switch to data backup;

[0092] During the fault handling process, minimize the service interruption time. The whole process from fault state to normal state includes data synchronization, service restart, status verification, etc.

[0093] Easy-to-expand technology: Adopt modular design to implement new functions or upgrade existing functions without rebuilding the entire system; use microservice architecture to improve the flexibility and scalability of the system.

[0094] In practical applications, a system architecture including a front-end user interface, a back-end server, a database management system, RFID reading devices, a sensor network, and a communication interface can be constructed first. The deployment process of the intelligent warehousing management system is as follows: First, evaluate the existing warehousing facilities to determine the locations suitable for deploying RFID readers and tags. Ensure that environmental factors such as light, temperature, and humidity in the warehouse are suitable to guarantee the stable operation of the RFID system. Then, according to the types and quantities of materials, select appropriate RFID tag configurations. For example, considering the particularity of the materials, the tags need to have strong anti-interference capabilities and waterproof performance. Subsequently, install and deploy RFID readers at the entrances, exits, and key storage areas of the warehouse to ensure that every storage unit can be covered; the installation height and angle of the readers should be carefully designed to achieve the best reading effect. Finally, system integration, integrate the RFID readers with the warehouse management system (WMS) to ensure that data can be transmitted and processed in real time. An application example after actual deployment can include an access layer, an application layer, a service layer, and an infrastructure layer in its system architecture, as Figure 4 shown.

[0095] The operation process of the actual intelligent warehousing management system can include the following processes:

[0096] Material inbound: After the materials arrive at the warehouse, scan the tags through the RFID readers and enter the material information into the system. The system automatically calculates the storage location and places the materials in place through automated equipment.

[0097] Storage management: The system regularly takes inventory of the stored materials, reads the tag information through the RFID readers, and compares it with the records in the system to ensure the accuracy and integrity of the materials.

[0098] Material retrieval: When materials need to be out of the warehouse, the operator can query the information of the required materials through the system. The system will automatically locate the material location and send the materials to the outbound port through automated equipment.

[0099] Material outbound: When the materials are out of the warehouse, scan the tags again through the RFID readers. The system records the outbound information and updates the inventory data.

[0100] The expected effects of the actual intelligent warehousing management system include:

[0101] Improve efficiency: Through automated equipment and intelligent management, reduce manual operations and improve the efficiency of material inbound and outbound.

[0102] Reduce costs: Reduce the costs of manual inventory and management, and reduce the waste of storage space through refined management; Enhance response speed: In case of emergencies, the system can quickly locate the required materials, improving the response speed and preparation efficiency of the unit.

[0103] Safe and reliable: Through RFID technology, real-time monitoring and tracking of materials are realized, reducing the risk of material loss and damage. At the same time, the system adopts strict data encryption and permission management mechanisms to ensure the security and reliability of material information.

[0104] Data-driven decision-making: By collecting and analyzing data such as material storage, inbound and outbound, the system can provide detailed data reports and analysis results to help decision-makers better understand the material management situation and make more reasonable decisions.

[0105] Scalability: The system adopts a modular design and can be functionally extended and customized according to actual needs to meet the requirements of different application scenarios.

[0106] Based on the same inventive concept, this embodiment also provides an RFID signal completion and anti-interference system based on contrast learning in a small-space intensive storage environment, which specifically includes the following units:

[0107] The reading optimization unit is used to optimize the automated reading of RFID signals, so as to optimize radio frequency identification in real time according to environmental changes and achieve adaptive reading of RFID tags;

[0108] The separation and extraction unit is used to separate and extract RFID signals, separate the strongly coupled part in the read RFID signals, and reduce the generation of incorrect data due to the collision and coupling of RFID signals with each other;

[0109] The completion unit is used to train a contrast learning model based on a convolutional neural network according to the historical RFID signals and the spatial position relationship of RFID tags, so that the contrast learning model can learn the patterns and relationship features of RFID signals and complete the incomplete RFID signals.

[0110] The completion unit can perform data completion and fault tolerance processing, preprocess and clean the read RFID signals, mainly including data completion and fault tolerance processing. Specifically, it mainly includes removing duplicate records, filling in null values, removing invalid data, consistency alignment processing, data error correction, etc.

[0111] The reading optimization unit realizes the automated reading of material information, including designing anti-interference algorithms to improve the reading accuracy of RFID signals. The completion unit studies data completeness strategies to complete and correct errors for data loss or missed reading caused by signal or material interference.

[0112] The system of this embodiment can also provide intelligent data analysis and services, develop data analysis models based on artificial intelligence, and provide intelligent services such as query, early warning, reminder and knowledge discovery; the system can set up high-availability and easy-to-expand maintenance modules, study data hot backup technology and breakpoint recovery mechanism, and ensure the continuous operation and adaptive expansion of the system.

[0113] It should be noted that the system of this embodiment corresponds to the aforementioned signal completion and anti-interference method, and each unit module of the system is respectively used to implement each step of the aforementioned signal completion and anti-interference method.

[0114] In general, this embodiment adopts multi-source information fault-tolerant technology and data completion strategy to effectively overcome signal interference and improve the accuracy and reliability of RFID data reading. First, in view of the dense distribution of RFID tags and signal overlap in a small space dense environment, a multi-dimensional signal acquisition method is adopted when reading the signal, the tag attachment medium is optimized, and the radio frequency optimization is performed in real time according to environmental changes; secondly, the strong coupling relaxation decomposition technology and blind source separation algorithm are used to model and separate the read RFID signals to improve signal quality and readability; finally, for incomplete data, a multi-scale linkage data fusion completion method based on MOPSO multi-objective optimization and contrast learning is used to train the convolutional neural network to learn data characteristics, and realize data completion of incomplete RFID signals, which can effectively solve the problem of data omission and misreading caused by RFID signal collision in a small space dense storage environment.

[0115] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A method for RFID signal completion and anti-interference based on contrastive learning in a small-space intensive storage environment, characterized in that, It includes the following steps: S1. Optimize the automated reading of RFID signals, thereby optimizing radio frequency identification in real time according to environmental changes to achieve adaptive reading of RFID tags; S2. Separate and extract RFID signals, separate the strongly coupled part in the read RFID signals, and reduce the generation of incorrect data due to collision and coupling between RFID signals; S3. Train a contrast learning model based on a convolutional neural network according to the historical RFID signal and the spatial position relationship of RFID tags, so that the contrast learning model can learn the patterns and relationship features of RFID signals and complete the incomplete RFID signals; In step S2, a blind source separation algorithm is adopted to decompose the strongly coupled part in the read RFID signals into independent signal sources; When separating and extracting RFID signals, graph theory and machine learning algorithms are used to model the RFID signals, regarding the signals as nodes and the edges representing the interactions between signals; by analyzing the correlation between RFID signals, strongly coupled signals are identified and decomposed into independent signal sources.

2. The RFID signal completion and anti-interference method according to claim 1, wherein In step S1, an RFID anti-collision optimization strategy based on multi-objective particle swarm optimization (MOPSO) is adopted to optimize radio frequency identification in real time according to environmental changes to achieve adaptive reading of RFID tags; when RFID signals are dense, the multi-objective particle swarm optimization (MOPSO) algorithm is used to intelligently process and analyze the obtained RFID signals and adjust the reading and writing radio frequency in real time.

3. The RFID signal completion and anti-interference method according to claim 2, wherein Step S1 includes: S11. Define the objective function of the optimization strategy to represent the optimization of the reading accuracy by the transmission power; S12. Define decision variables, including two parts: y and z: y = {y1, y2,..., y n}, z = {z1, z2,..., z n} Among them, y represents the configuration parameters of the RFID system in a dense small space, including the layout of RFID tag readers, the angles and positions of antennas; z represents the variables related to different types of materials or storage areas, including the types of materials, tag characteristics, and position information in the dense small space; S13. Define the constraint conditions of the decision variables: S14. Approximate the read RFID signals by weighted summation of the characteristic functions of the signal sources to achieve effective multi-source extraction of signals: where N represents the number of signal sources, w i is the weight of the i-th signal source, and Φ i represents the characteristic function of the i-th signal source, and x represents the read RFID signal.

4. The RFID signal completion and anti-interference method according to claim 1, characterized in that Step S2 includes: S21. Let the blind source separation algorithm be independent component analysis, and the objective function of independent component analysis is expressed as: where W is the separation filter matrix, and w i is the weight of the i-th signal source, represents the column vector of the separation filter matrix, and S is the covariance matrix of RFID signals; S22. Adopt short-time Fourier transform to extract the time-frequency characteristics of RFID signals, and analyze the correlation between RFID signals according to the extracted time-frequency characteristics.

5. The RFID signal completion and anti-interference method according to claim 4, characterized in that, The time-frequency characteristics of RFID signals are expressed as: where X(t, f) is the time-frequency representation of the RFID signal, x RFID (τ) is the original RFID signal, w(t - τ) is the window function, t represents the time variable, and f represents the frequency variable.

6. The RFID signal completion and anti-interference method according to claim 1, wherein Step S3 includes: S31. In the convolutional neural network, the convolutional layer performs convolution operations on the input historical RFID signals by applying a series of learnable filters; S32. Reduce the spatial size of the feature map of RFID signals through the pooling layer; S33. Train a contrast learning model based on a convolutional neural network.

7. The RFID signal completion and anti-interference method according to claim 1 or 6, characterized in that In step S3, the historical RFID signals include tag reading intensity, time stamp, and environmental noise; the spatial position relationship of RFID tags includes the distance, angle, and relative position between tags.

8. A RFID signal completion and anti-interference system based on contrastive learning in a small-space intensive storage environment, characterized in that, It includes the following units: A reading optimization unit for optimizing the automated reading of RFID signals, thereby optimizing radio frequency identification in real time according to environmental changes and achieving adaptive reading of RFID tags; A separation and extraction unit for separating and extracting RFID signals, separating the strongly coupled part of the read RFID signals to reduce the generation of incorrect data due to the collision and coupling of RFID signals with each other; A completion unit for training a contrast learning model based on a convolutional neural network according to the historical RFID signals and the spatial position relationship of RFID tags, enabling the contrast learning model to learn the patterns and relationship features of RFID signals and complete the incomplete RFID signals; The separation and extraction unit uses a blind source separation algorithm to decompose the strongly coupled part of the read RFID signals into independent signal sources; When separating and extracting RFID signals, graph theory and machine learning algorithms are used to model the RFID signals, regarding the signals as nodes and the edges representing the interactions between signals; by analyzing the correlation between RFID signals, strongly coupled signals are identified and decomposed into independent signal sources.

9. The RFID signal completion and anti-interference system according to claim 8, wherein The reading optimization unit adopts an RFID anti-collision optimization strategy based on multi-objective particle swarm optimization (MOPSO) to optimize radio frequency identification in real time according to environmental changes and achieve adaptive reading of RFID tags; when RFID signals are dense, the multi-objective particle swarm optimization (MOPSO) algorithm is used to intelligently process and analyze the obtained RFID signals and adjust the reading and writing radio frequencies in real time.

Images