Multi-sensor fusion intelligent detection method and system for 5G infrastructure

Abstract

The invention relates to the technical field of 5G infrastructure intelligent detection, in particular to a multi-sensor fusion intelligent detection method and system for a 5G infrastructure, which realizes real-time monitoring of the appearance and environmental parameters of 5G infrastructure equipment by integrating an image sensor, a temperature and humidity sensor and a vibration sensor, establishes a data interaction protocol, and improves the detection accuracy of the 5G infrastructure equipment. Processing and analyzing the collected data, generating a joint certificate, checking and verifying the integrity and availability of the data, then carrying out fault detection and health state evaluation by using a visual fault detection model and a state variable prediction model, and carrying out intelligent decision making according to an evaluation result and a decision fusion rule base. Comprising self-diagnosis operation, fault information recording, emergency alarm sending and emergency plan triggering, collected data, fault levels and intelligent scheduling decisions are displayed in real time through a graphical interface, and a visual monitoring interface is provided for maintenance personnel. According to the invention, the operation efficiency and safety of the 5G infrastructure are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of 5G infrastructure intelligent detection technology, and in particular to a multi-sensor fusion intelligent detection method and system for 5G infrastructure. Background Art

[0002] In recent years, multi-sensor fusion technology has attracted widespread attention because of its ability to provide more comprehensive and accurate monitoring data. By integrating multiple types of sensors, it can achieve all-round monitoring of 5G infrastructure, timely detect potential problems and take preventive measures.

[0003] A Chinese invention patent with announcement number CN106973086B discloses an intelligent detection system based on multi-sensor fusion, including multiple sensor fusion detection devices, each of which sends data to the cloud via a wireless network. Several sensor fusion detection devices are self-networked through the ZigBee protocol. After networking, the sensor fusion detection device acts as a node to transmit data to a coordinator, and the coordinator transmits the data to a host computer or sends it to the cloud via a wireless network.

[0004] However, the operation of 5G infrastructure faces multiple challenges, the most critical of which is the difficulty in timely detection, diagnosis and handling of equipment failures. Traditional monitoring methods often rely on manual inspections or single sensor data, which is not only inefficient but also prone to missing potential hidden dangers. For example, key information such as equipment appearance damage, temperature and humidity changes, and vibration abnormalities, if not monitored in real time and accurately, will directly affect the judgment of the health status of the equipment. In addition, the incompleteness and availability of data are also important factors that restrict the accuracy of fault detection. In the absence of effective data interaction protocols and processing mechanisms, the collected data may not be effectively integrated and analyzed, thus affecting the accuracy and timeliness of decision-making. Summary of the invention

[0005] The purpose of the present invention is to address the problems existing in the background technology and to propose a multi-sensor fusion intelligent detection method and system for 5G infrastructure.

[0006] The technical solution of the present invention is a multi-sensor fusion intelligent detection method for 5G infrastructure, comprising the following specific implementation steps:

[0007] S1. Collect appearance image data, temperature data, humidity data and vibration data of 5G infrastructure, and review the integrity and availability of the image data, temperature data, humidity data and vibration data, and then pre-process the collected image data, temperature data, humidity data and vibration data;

[0008] S2. Build a visual fault detection model, use image data as model input parameters, input it into the trained visual fault detection model, detect the 5G infrastructure, identify whether there is a fault, and output image fault information if a fault is detected;

[0009] S3: Fusion analysis of temperature data, humidity data, and vibration data to build a fault risk prediction model and predict the health status of 5G infrastructure;

[0010] S4. Combine image fault information and predicted health status of 5G infrastructure, automatically classify the 5G infrastructure status according to the decision fusion rule base, and make automatic intelligent scheduling decisions based on the 5G infrastructure status level;

[0011] S5, real-time display of temperature data, humidity data, vibration data, image data, 5G infrastructure status level and intelligent scheduling decisions.

[0012] Preferably, the analysis process of fusing and analyzing the temperature data, humidity data and vibration data is as follows:

[0013] S21. Define the state variable x representing the current health status i and the measured variables z representing the real-time collected temperature, humidity, and vibration data i =[Tdata i , Hdata i , Vdata i ] T ;

[0014] Among them, Tdata i 、Hdata i 、Vdata i Respectively represent temperature data, humidity data and vibration data;

[0015] S22. Predict the state variables at the next moment And corresponds to the uncertainty of the predicted state, using the covariance matrix P i|i-1 Indicates the error range of the predicted state:

[0016]

[0017] In the formula, Indicates time T i The predicted state, that is, at the last moment T i-1 Under the observation conditions, for the current moment T i The predicted value of the health status of the 5G infrastructure; A represents the state transition matrix; Indicates the last moment T i-1 The estimated state, that is, based on the last moment T i-1The estimated value of health status after data processing; w i represents process noise, i.e., state changes caused by unknown interference factors; P i|i-1 represents the covariance matrix of the predicted state; P i-1|i-1 Indicates the last moment T i-1 The state estimation covariance matrix is ​​the prediction uncertainty of the previous moment; Q represents the process noise covariance matrix, which measures the process noise w i uncertainty;

[0018] S23, calculate Kalman gain K i , and use the measured variable z i Update 5G infrastructure status and get status estimate Then update the covariance matrix P of the state estimate i|i , reflecting the confidence of the current state estimate:

[0019] K i =P i|i-1 H T (HP i|i-1 H T +R) -1 ;

[0020]

[0021] P i|i =(IK i H)P i|i-1 ;

[0022] In the formula, K i represents the Kalman gain matrix, which represents the weight of the measurement data in the current state estimation; H represents the measurement matrix, which is used to describe the temperature data Tdata i 、Humidity data Hdata i And vibration data Vdata i The relationship between and state variables is represented by the unit matrix; R represents the measurement noise covariance matrix, which reflects the degree of measurement error of the sensor; Indicates the current time T i The state estimation of , i.e., the health state estimation after combining the prediction and the current measurement; represents the residual, that is, the difference between the prediction and the measurement; P i|i Indicates the current time T i The covariance matrix of is the updated uncertainty range; I represents the identity matrix;

[0023] S24, construct the health status time series data Y of 5G infrastructure = {y1, y2, ..., y i , ..., y n};

[0024] Among them, y i Indicates time T i Health status data of 5G infrastructure;

[0025] S25. Based on the trained fault risk prediction model, the health status time series data of 5G infrastructure Y = {y1, y2, ..., y i , ..., y n} is the input parameter, predicting the future T i+h The health of 5G infrastructure at all times The prediction formula is as follows:

[0026]

[0027] In the formula, Represents the predicted future T i+h Keep track of the health of your 5G infrastructure. represents the predicted value of past health status; ε i+h-j represents the past forecast error.

[0028] Preferably, the training process of the fault risk prediction model is as follows:

[0029] S31. Collect time series data related to the health status of the equipment, including temperature, humidity and vibration data, ensure that the frequency of the data is consistent, and merge the temperature, humidity and vibration data into a time series, which is defined as training data;

[0030] S32. Use the Augmented Dickey-Fuller test to check whether the time series is stationary. If the series is not stationary, perform difference processing to eliminate the trend. The number of differences is d.

[0031] S33, draw the autocorrelation function and partial autocorrelation function graphs, take the 95% confidence interval as a reference, observe the significance, and select appropriate p' and q' values;

[0032] Among them, the autocorrelation function is used to identify the order q' of the moving average part; the partial autocorrelation function is used to identify the order p' of the autoregressive part;

[0033] S34. Define the ARIMA model: ARIMA(p', d, q');

[0034] S35, train the ARIMA model, use the training data to fit the ARIMA model, and estimate the model parameters: autoregressive coefficient and moving average coefficient by minimizing the error, so that the model fits the existing training data;

[0035] S36. Repeat steps S32 to S35, use the Akaike Information Criterion to evaluate the performance of different ARIMA models, and select the model with the lowest AIC value as the fault risk prediction model.

[0036] Preferably, the visual fault detection model integrates the coordinate attention module in the backbone feature extraction network based on the YOLOv8 network structure.

[0037] Preferably, the implementation steps of the coordinate attention module are as follows:

[0038] S51, perform average pooling operations on the input feature map F in the horizontal and vertical directions respectively, compressing the features into information in the height direction and compressing the features into information in the width direction respectively:

[0039]

[0040] In the formula, F X and F Y Respectively represent the compressed horizontal feature representation and the compressed vertical feature representation; AvgPool W () means that the feature map of each channel is averaged in the width direction (i.e. each row) to generate features in the height direction; AvgPool H () indicates that the feature map of each channel is averaged in the height direction (i.e., each column) to generate features in the width direction; F represents the input feature map;

[0041] S52, through the convolution operation, the horizontal feature representation F X and the longitudinal feature representation F Y Mapped to the attention weight matrix respectively, we get the horizontal attention weight matrix W X and the vertical attention weight matrix W Y :

[0042]

[0043] Where σ represents the activation function; Conv 1×1 () represents a 1×1 convolution kernel, which is used to generate attention weights; W X and W Y Represent the horizontal attention weight matrix and the vertical attention weight matrix respectively;

[0044] S53, generated horizontal attention weight matrix W X and the vertical attention weight matrix W Y The weights of the input feature map F are rescaled horizontally and vertically respectively, resulting in a spatially enhanced feature representation:

[0045] Multiply the horizontal weights by the input feature map F element by element to obtain the horizontal weighted feature F ′X :

[0046] F′ X =F⊙W X ;

[0047] Multiply the vertical weight by the input feature map F element by element to obtain the vertical weighted feature F′ Y :

[0048] F′ Y =F⊙W Y ;

[0049] In the formula, F′ X and F′ Y Respectively represent horizontal weighted features and vertical weighted features; ⊙ represents element-by-element multiplication operation;

[0050] S54, the horizontal weighted feature F' X and the longitudinal weighted feature F′ Y After fusion, we get the final feature F' with spatial attention enhancement:

[0051] F'=F' X +F′ Y .

[0052] The technical solution of the present invention is a multi-sensor fusion intelligent detection system for 5G infrastructure, which is used to execute the above-mentioned multi-sensor fusion intelligent detection method for 5G infrastructure, including:

[0053] The sensor data acquisition module is used to collect and transmit the status data of the 5G infrastructure, including image sensors, temperature and humidity sensors, vibration sensors, and data acquisition and transmission units;

[0054] Image sensors for real-time monitoring of appearance image data of 5G infrastructure;

[0055] Temperature and humidity sensors, used to collect temperature and humidity data of 5G infrastructure;

[0056] Vibration sensors to collect vibration data from 5G infrastructure;

[0057] A data acquisition and transmission unit, used to obtain real-time image data, temperature data, humidity data and vibration data from the image sensor, the temperature and humidity sensor and the vibration sensor, and generate a joint certificate, and transmit {joint certificate, image data, temperature data, humidity data, vibration data} to the data processing module;

[0058] The data processing module is used to build a data interaction protocol with the data acquisition and transmission unit in the system initialization phase, and to review the received joint certificate in the multi-sensor fusion intelligent detection phase, verify the integrity and availability of image data, temperature data, humidity data, and vibration data, and perform denoising and standardization processing on the image data, temperature data, humidity data, and vibration data;

[0059] The data fusion and analysis module is used to build a fault risk prediction model, fuse and analyze temperature data, humidity data, and vibration data, predict the health status of 5G infrastructure, and build a visual fault detection model to detect 5G infrastructure and identify whether there is a fault. If a fault is detected, the image fault information is output and the predicted health status of the 5G infrastructure and the image fault information are transmitted to the intelligent decision-making module;

[0060] An intelligent decision-making module, which is used to classify the status of 5G infrastructure and automatically generate maintenance recommendations based on the predicted health status and image fault information of 5G infrastructure;

[0061] The system visualization module is used to display the collected temperature data, humidity data, vibration data, image data, 5G infrastructure status level and intelligent scheduling decisions in real time through a graphical interface.

[0062] Preferably, the specific implementation process of the data interaction protocol is as follows:

[0063] S71, the data processing module implements the following operations:

[0064] S7101. Define a cyclic group G of order q, where P is a generator of the cyclic group G;

[0065] Where q>2 k , k is an arbitrarily selected security parameter;

[0066] S7102. Define the following secure hash function:

[0067]

[0068] Wherein, L1 is the length of the binary string of the set image sensor, temperature and humidity sensor or vibration sensor device identifier, and L2 is the length of the binary string of the specified collected image data, temperature data, humidity data or vibration data; is the multiplicative group:

[0069] S7103, Randomly select the first-level audit factor Calculate the secondary audit factor F s2 =F s1 ×P;

[0070] S7104, save the first level audit factor F s1 and common parameters {q, P, G, F s2 , H1, H2}, and disclose the public parameters {q, P, G, F s2 , H1, H2};

[0071] S72, the data acquisition and transmission unit performs the following operations:

[0072] S7201, select a unique random number for the image sensor Select a unique random number for the temperature and humidity sensor Select a unique random number for the vibration sensor And calculate the permission application parameter P A1 =α1×P, temperature and humidity sensor permission application parameter P A2 =α2×P, vibration sensor permission application parameter P A3 =α3×P;

[0073] S7202, obtain the device identification of the image sensor, temperature and humidity sensor, and vibration sensor, and establish an associated group <ID1,P A1 >, Association Group <ID2,P A2 >, Association Group <ID3,P A3 >

[0074] Among them, ID1, ID2 and ID3 are the device identifiers of the image sensor, temperature and humidity sensor and vibration sensor respectively;

[0075] S7203, save the image sensor exclusive random number α1, the temperature and humidity sensor exclusive random number α2, the vibration sensor exclusive random number α3, and the association group <ID1,P A1 >, Association Group <ID2,P A2 >, Association Group <ID3,P A3 > and associate the group <ID1,P A1 >, Association Group <ID2,P A2 >, Association Group <ID3,P A3 >Transmit to data processing module;

[0076] S73, the data processing module implements the following operations:

[0077] S7301. Select auxiliary authority control factor for image sensor Select auxiliary permission control factors for temperature and humidity sensors Select auxiliary permission control factors for vibration sensors

[0078] S7302. Calculate the primary data collection authority parameter γ of the image sensor 11 =β1×P, secondary data collection authority parameter γ 12 =β1+F s1 ×H1(ID1,P A1 , γ 11 );

[0079] Calculate the primary data collection authority parameter γ of the temperature and humidity sensor 21 =β2×P, secondary data collection authority parameter γ 22 =β2+F s1 ×H1(ID2,P A2 , γ 21 );

[0080] Calculate the primary data collection authority parameter γ of the vibration sensor 31 =β3×P, secondary data collection authority parameter γ 32 =β3+F s1 ×H1(ID3,P A3 , γ 31 );

[0081] S7303, create and save association group <ID1,P A1 , γ 11 , γ 12 >, Association Group <ID2,P A2 , γ 21 , γ 22 > and associated groups <ID3,P A3 , γ 31 , γ 32 >, then associate the group <ID1,P A1 , γ 11 , γ 12 >, Association Group <ID2,P A2 , γ 21 , γ 22 > and associated groups <ID3,P A3 , γ 31 , γ 32 >Transmit to the data acquisition and transmission unit.

[0082] Preferably, the generation process of the joint certificate is as follows:

[0083] S81, image data Pdata i Generate a data certificate. The generation process is as follows:

[0084] S8101, select random number

[0085] S8102. Calculate the following certificate generation factors:

[0086] CG P1 =κ1×P;

[0087] CG P2 =H2(ID1, Pdata i , C.G. P1 );

[0088] CG P3 =κ1+(α1+γ 12 )×CG P2 ;

[0089] S8103, generate image data Pdata i Data certificate Cert1 = (CG P1 , C.G. P3 ); S82, temperature data Pdata i Generate a data certificate. The generation process is as follows:

[0090] S8201, select random number

[0091] S8202. Calculate the following certificate generation factors:

[0092] CG T1 =κ2×P;

[0093] CG T2 =H2(ID2,Tdata i , C.G. T1 );

[0094] CG T3 =κ2+(α2+γ 22 )×CG T2 ;

[0095] S8203, generate temperature data Tdata i Data certificate Cert2 = (CG T1 , C.G. T3 ); S83, humidity data Hdata i Generate a data certificate. The generation process is as follows:

[0096] S8301, select random number

[0097] S8302. Calculate the following certificate generation factors:

[0098] CG H1 =κ3×P;

[0099] CG H2 =H2(ID2,Hdatai , C.G. H1 );

[0100] CG H3 =κ3+(α2+γ 22 )×CG H2 ;

[0101] S8303, generate humidity data Hdata i Data certificate Cert3 = (CG H1 , C.G. H3 ); S84, vibration data Vdata i Generate a data certificate. The generation process is as follows:

[0102] S8401, select random number

[0103] S8402. Calculate the following certificate generation factors:

[0104] CG V1 =κ4×P;

[0105] CG V2 =H2(ID3, Vdata i , C.G. V1 );

[0106] CG V3 =κ4+(α3+γ 32 )×CG V2 ;

[0107] S8403, generate vibration data Vdata i Data certificate Cert3 = (CG V1 , C.G. V3 );

[0108] S85, Computing Joint Certificate Level 1 JCert1 = CG P1 +CG T1 +CG H1 +CG V1 , calculate the second level joint certificate JCert2 = CG P3 +CG T3 +CG H3 +CG V3 ;

[0109] S86. Generate a joint certificate JCert={JCert1, JCert2}.

[0110] Preferably, the audit process for auditing the received joint certificate is as follows:

[0111] S9101. Calculate the following audit factors:

[0112] F AP1 =H2(ID1,P A1 , γ 11 );

[0113] F AP2 =H2(ID1, Pdata i , C.G. P1 );

[0114] F AT1 =H2(ID2,P A2 , γ 21 );

[0115] F AT2 =H2(ID2,Tdata i , C.G. T1 );

[0116] F AH1 =H2(ID2,P A2 , γ 21 );

[0117] F AH2 =H2(ID2,Hdata i , C.G. H1 );

[0118] F AV1 =H2(ID3,P A3 , γ 31 );

[0119] F AV2 =H2(ID3, Vdata i , C.G. V1 );

[0120] S9102, calculate the first-level audit parameter A1 = JCert2 × P, calculate the second-level audit parameter A2 = [F AP2 ×(P A1 +γ 11 +F s2 ×F AP1 )]+[F AT2 ×(P A2 +γ 21 +F s2 ×F AT1 )]+[F AH2 ×(P A2 +γ 21 +F s2 ×F AH1 )]+[F AV2 ×(P A3 +γ 31 +Fs2 ×F AV1 )]+JCert1;

[0121] If A1=A2, the joint certificate is approved, which means the received image data Pdata is confirmed. i 、Tdata i 、Humidity data Hdata i And vibration data Vdata i completeness and availability.

[0122] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects:

[0123] The present invention designs a multi-sensor fusion intelligent detection method and system for 5G infrastructure; the method realizes real-time monitoring of the appearance and environmental parameters of 5G infrastructure equipment by integrating image sensors, temperature and humidity sensors and vibration sensors. The data processing module establishes a data interaction protocol with the data acquisition and transmission unit, processes and analyzes the collected data, and generates a joint certificate to verify the integrity and availability of the data. The data fusion and analysis module receives the pre-processed data and uses the visual fault detection model and the state variable prediction model to perform fault detection and health status assessment. The intelligent decision-making module makes intelligent decisions based on the evaluation results and the decision fusion rule base, including self-diagnosis operations, recording fault information, issuing emergency alarms and triggering emergency plans. The system visualization module displays the collected data, fault levels and intelligent scheduling decisions in real time through a graphical interface, providing an intuitive monitoring interface for maintenance personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0124] Figure 1 This is a system architecture diagram of a multi-sensor fusion intelligent detection system for 5G infrastructure proposed in the present invention. DETAILED DESCRIPTION

[0125] Embodiment 1, as Figure 1 As shown, the present invention proposes a multi-sensor fusion intelligent detection system for 5G infrastructure, including: a sensor data acquisition module, a data analysis module, a data fusion and analysis module, an intelligent decision-making module and a system visualization module.

[0126] The sensor data acquisition module is used to collect and transmit the status data of the 5G infrastructure, including image sensors, temperature and humidity sensors, vibration sensors, and data acquisition and transmission units;

[0127] Image sensors are used to monitor the appearance image data of 5G infrastructure in real time, including but not limited to the appearance and surface status of equipment;

[0128] Temperature and humidity sensors are used to collect temperature and humidity data of 5G infrastructure;

[0129] Vibration sensors are used to collect vibration data of 5G infrastructure;

[0130] The data acquisition and transmission unit is used to obtain real-time image data, temperature data, humidity data and vibration data from the image sensor, the temperature and humidity sensor and the vibration sensor, and generate a joint certificate, and transmit {joint certificate, image data, temperature data, humidity data, vibration data} to the data processing module;

[0131] The data processing module is used to build a data interaction protocol with the data acquisition and transmission unit in the system initialization phase, and to review the received joint certificate in the multi-sensor fusion intelligent detection phase, and to perform denoising and standardization processing on image data, temperature data, humidity data, and vibration data;

[0132] The data fusion and analysis module is used to build a fault risk prediction model, fuse and analyze temperature data, humidity data, and vibration data, predict the health status of 5G infrastructure, and build a visual fault detection model to detect 5G infrastructure and identify whether there is a fault. If a fault is detected, the image fault information is output and the predicted health status of the 5G infrastructure and the image fault information are transmitted to the intelligent decision-making module;

[0133] The intelligent decision-making module is used to classify the status of the 5G infrastructure based on the predicted health status and image fault information of the 5G infrastructure, and automatically generate maintenance recommendations;

[0134] The system visualization module is used to display the collected temperature data, humidity data, vibration data, image data, 5G infrastructure status level and intelligent scheduling decisions in real time through a graphical interface.

[0135] Embodiment 2: A multi-sensor fusion intelligent detection method for 5G infrastructure proposed in the present invention is applied to a multi-sensor fusion intelligent detection system for 5G infrastructure proposed in Embodiment 1. The specific implementation steps are as follows:

[0136] Initialization phase:

[0137] Construct a data interaction protocol between the data processing module and the data acquisition and transmission unit. The specific process is as follows:

[0138] S1. The data processing module implements the following operations:

[0139] S11. Define a cyclic group G of order q, where P is a generator of the cyclic group G;

[0140] Where q>2 k, k is an arbitrarily selected security parameter;

[0141] S12. Define the following secure hash function:

[0142]

[0143] Wherein, L1 is the length of the binary string of the set image sensor, temperature and humidity sensor or vibration sensor device identifier, and L2 is the length of the binary string of the specified collected image data, temperature data, humidity data or vibration data; is the multiplicative group {1, 2, ..., q-1};

[0144] S13. Randomly select the first-level audit factor Calculate the secondary audit factor F s2 =F s1 ×P;

[0145] S14. Save the first-level audit factor F s1 and common parameters {q, P, G, F s2 , H1, H2}, and disclose the public parameters {q, P, G, F s2 , H1, H2};

[0146] S2. The data acquisition and transmission unit performs the following operations:

[0147] S21. Selecting a unique random number for the image sensor Select a unique random number for the temperature and humidity sensor Select a unique random number for the vibration sensor And calculate the permission application parameter P A1 =α1×P, temperature and humidity sensor permission application parameter P A2 =α2×P, vibration sensor permission application parameter P A3 =α3×P;

[0148] S22, obtain the device identification of the image sensor, temperature and humidity sensor and vibration sensor, and establish an associated group <ID1,P A1 >, Association Group <ID2,P A2 >, Association Group <ID3,P A3 >

[0149] Among them, ID1, ID2 and ID3 are the device identifiers of the image sensor, temperature and humidity sensor and vibration sensor respectively;

[0150] S23, save the image sensor exclusive random number α1, the temperature and humidity sensor exclusive random number α2, the vibration sensor exclusive random number α3, and the associated group <ID1,P A1>, Association Group <ID2,P A2 >, Association Group <ID3,P A3 > and associate the group <ID1,P A1 >, Association Group <ID2,P A2 >, Association Group <ID3,P A3 >Transmit to data processing module;

[0151] S3, the data processing module implements the following operations:

[0152] S31. Selecting auxiliary authority control factors for image sensors Select auxiliary permission control factors for temperature and humidity sensors Select auxiliary permission control factors for vibration sensors

[0153] S32, calculate the primary data collection authority parameter γ of the image sensor 11 =β1×P, secondary data collection authority parameter γ 12 =β1+F s1 ×H1(ID1,P A1 , γ 11 );

[0154] Calculate the primary data collection authority parameter γ of the temperature and humidity sensor 21 =β2×P, secondary data collection authority parameter γ 22 =β2+F s1 ×H1(ID2,P A2 , γ 21 );

[0155] Calculate the primary data collection authority parameter γ of the vibration sensor 31 =β3×P, secondary data collection authority parameter γ 32 =β3+F s1 ×H1(ID3,P A3 , γ 31 );

[0156] S33. Create and save association groups <ID1,P A1 , γ 11 , γ 12 >, Association Group <ID2,P A2 , γ 21 , γ 22 > and associated groups <ID3,P A3 , γ 31 , γ 32 > and associate the group <ID1,P A1 , γ 11 , γ 12 >, Association Group <ID2,PA2 , γ 21 , γ 22 > and associated groups <ID3,P A3 , γ 31 , γ 32 >Transmit to the data acquisition and transmission unit.

[0157] Multi-sensor fusion intelligent detection stage:

[0158] S1. The image sensor captures the appearance image of the 5G infrastructure and sends the collected image data Pdata i Transmit to the data acquisition and transmission unit;

[0159] The temperature and humidity sensor monitors the temperature and humidity changes of 5G infrastructure and environment in real time, and transmits the temperature data Tdata i 、Humidity data Hdata i Converted into digital signals and sent to the data acquisition and transmission unit;

[0160] Vibration sensors measure vibration data Vdata for 5G infrastructure i , and the vibration data Vdata i Transmit to the data acquisition and transmission unit.

[0161] S2. The data acquisition and transmission unit receives the image data Pdata i 、Tdata i 、Humidity data Hdata i And vibration data Vdata i , generate a joint certificate for the received image data, temperature data, humidity data and vibration data, and {joint certificate, image data Pdata i , temperature data Tdata i , humidity data Hdata i , vibration data Vdata i}Transmit to data processing module;

[0162] The process of generating a joint certificate is as follows:

[0163] S21, is the image data Pdata i Generate a data certificate. The generation process is as follows:

[0164] S2101, select random number

[0165] S2102. Calculate the following certificate generation factors:

[0166] CG P1 =κ1×P;

[0167] CG P2 =H2(ID1, Pdata i , C.G. P1 );

[0168] CG P3 =κ1+(α1+γ 12 )×CG P2 ;

[0169] S2103, generate image data Pdata i Data certificate Cert1 = (CG P1 , C.G. P3 ); S22, temperature data Pdata i Generate a data certificate. The generation process is as follows:

[0170] S2201, select random number

[0171] S2202. Calculate the following certificate generation factors:

[0172] CG T1 =κ2×P;

[0173] CG T2 =H2(ID2,Tdata i , C.G. T1 );

[0174] CG T3 =κ2+(α2+γ 22 )×CG T2 ;

[0175] S2203, generate temperature data Tdata i Data certificate Cert2 = (CG T1 , C.G. T3 ); S23, humidity data Hdata i Generate a data certificate. The generation process is as follows:

[0176] S2301, select random number

[0177] S2302. Calculate the following certificate generation factors:

[0178] CG H1 =κ3×P;

[0179] CG H2 =H2(ID2,Hdata i , C.G. H1 );

[0180] CG H3=κ3+(α2+γ 22 )×CG H2 ;

[0181] S2303, generate humidity data Hdata i Data certificate Cert3 = (CG H1 , C.G. H3 ); S24, vibration data Vdata i Generate a data certificate. The generation process is as follows:

[0182] S2401, select random number

[0183] S2402. Calculate the following certificate generation factors:

[0184] CG V1 =κ4×P;

[0185] CG V2 =H2(ID3, Vdata i , C.G. V1 );

[0186] CG V3 =κ4+(α3+γ 32 )×CG V2 ;

[0187] S2403, generate vibration data Vdata i Data certificate Cert3 = (CG V1 , C.G. V3 );

[0188] S25, calculation of the first joint certificate JCert1 = CG P1 +CG T1 +CG H1 +CG V1 , calculate the second level joint certificate JCert2 = CG P3 +CG T3 +CG H3 +CG V3 ;

[0189] S26. Generate a joint certificate JCert={JCert1, JCert2}.

[0190] S3, the data processing unit receives {joint certificate JCert, image data Pdata i , temperature data Tdata i , humidity data Hdata i , vibration data Vdata i} for preprocessing, the specific implementation process is as follows:

[0191] S31, review the joint certificate and confirm the received image data Pdata i 、Tdata i 、Humidity data Hdata i And vibration data Vdata i The completeness and availability of the audit process are as follows:

[0192] S3101. Calculate the following audit factors:

[0193] F AP1 =H2(ID1,P A1 , γ 11 );

[0194] F AP2 =H2(ID1, Pdata i , C.G. P1 );

[0195] F AT1 =H2(ID2,P A2 , γ 21 );

[0196] F AT2 =H2(ID2,Tdata i , C.G. T1 );

[0197] F AH1 =H2(ID2,P A2 , γ 21 );

[0198] F AH2 =H2(ID2,Hdata i , C.G. H1 );

[0199] F AV1 =H2(ID3,P A3 , γ 31 );

[0200] F AV2 =H2(ID3, Vdata i , C.G. V1 );

[0201] S3102, calculate the first-level audit parameter A1 = JCert2 × P, calculate the second-level audit parameter A2 = [F AP2 ×(P A1 +γ 11 +F s2 ×F AP1 )]+[F AT2×(P A2 +γ 21 +F s2 ×F AT1 )]+[F AH2 ×(P A2 +γ 21 +F s2 ×F AH1 )]+[F AV2 ×(P A3 +γ 31 +F s2 ×F AV1 )]+JCert1;

[0202] If A1=A2, the joint certificate is reviewed and the received image data Pdata is confirmed i 、Tdata i 、Humidity data Hdata i And vibration data Vdata i the completeness and availability of

[0203] S32, for the received Pdata i For preprocessing, a two-dimensional Gaussian function is used to smooth the image and remove high-frequency noise. The Gaussian filter formula is:

[0204]

[0205] In the formula, G(x,y) represents the response value of the Gaussian filter at the point (x,y), indicating the degree of smoothness of the point; σ represents the standard deviation, which controls the width of the Gaussian distribution and determines the degree of smoothness of the filter; x and y represent the two-dimensional spatial coordinates of the image, which are used to locate the position of the pixel;

[0206] S33, pre-process the received temperature data and humidity data, use the sliding average method to smooth the curve and remove the noise of short-term fluctuations. The sliding average method formula is:

[0207]

[0208] In the formula, Represents the sliding average of the temperature data or humidity data at time t; X t-i Represents the original temperature data value or humidity data value at time ti time units; N represents the size of the sliding window, which determines the number of historical data points required to calculate the average value;

[0209] S34, pre-processing the received vibration data, using discrete wavelet transform to remove low-frequency noise, the formula of discrete wavelet transform is:

[0210]

[0211] Where f(t) represents the original signal of vibration data; c j,k represents the wavelet coefficient, which indicates the projection degree of the signal on the wavelet basis function and reflects the characteristics of the signal; ψ j,k(t) represents the j-th layer and the k-th wavelet basis function, which represents the shape of the wavelet in the time domain; j represents the scale parameter of the wavelet, which represents the frequency of the wavelet basis function. The larger the value, the lower the frequency of the wavelet basis function and the larger the scale; k represents the displacement parameter, which represents the position of the wavelet basis function on the time axis; t represents the time variable, which represents the time point of the vibration data input signal; ψ represents the wavelet mother function, which is the basic wavelet shape and determines the characteristics of the wavelet. The Haar wavelet is used in this embodiment;

[0212] S35. Use the Z-score standardization formula to standardize the denoised image data, temperature data, humidity data, and vibration data. The Z-score standardization formula is:

[0213]

[0214] X' represents the standardized image data, temperature data, humidity data or vibration data; X represents the original image data, temperature data, humidity data or vibration data; μ represents the mean of the image data, temperature data, humidity data or vibration data; τ represents the standard deviation of the image data, temperature data, humidity data or vibration data;

[0215] S36, the pre-processed image data Pdata i 、Tdata i 、Humidity data Hdata i And vibration data Vdata i Transmitted to the data fusion and analysis module and the system visualization module.

[0216] S4, the data fusion and analysis module receives the pre-processed image data Pdata i 、Tdata i 、Humidity data Hdata i And vibration data Vdata i , and perform the following operations:

[0217] S41, with image data Pdata i Input parameters for the model into the trained visual fault detection model, detect the 5G infrastructure, identify whether there is a fault, and if a fault is detected, transmit the detected image fault information to the intelligent decision module and the system visualization module;

[0218] It should be noted that the visual fault detection model integrates the coordinate attention module in the backbone feature extraction network based on the YOLOv8 network structure. The coordinate attention module greatly enhances the model's perception of spatial information by introducing directional weights.

[0219] The implementation steps of coordinate attention are as follows:

[0220] S4101, perform average pooling operations on the input feature map F in the horizontal and vertical directions respectively, compressing the features into information in the height direction and compressing the features into information in the width direction respectively:

[0221]

[0222] In the formula, F X and F Y Respectively represent the compressed horizontal feature representation and the compressed vertical feature representation; AvgPool W () means that the feature map of each channel is averaged in the width direction (i.e. each row) to generate features in the height direction; AvgPool H () indicates that the feature map of each channel is averaged in the height direction (i.e., each column) to generate features in the width direction; F represents the input feature map;

[0223] S4102, through the convolution operation, the horizontal feature representation F X and the longitudinal feature representation F Y Mapped to the attention weight matrix respectively, we get the horizontal attention weight matrix W X and the vertical attention weight matrix W Y :

[0224]

[0225] Wherein, σ represents the activation function, and this embodiment adopts the Sigmoid activation function; Conv 1×1 () represents a 1×1 convolution kernel, which is used to generate attention weights; W X and W Y Represent the horizontal attention weight matrix and the vertical attention weight matrix respectively;

[0226] S4103, the generated directional weights and the weights of the input feature map F are readjusted in the horizontal and vertical directions respectively, thereby generating a feature representation with spatial perception enhancement:

[0227] Multiply the horizontal weights by the input feature map F element by element to obtain the horizontal weighted feature F′ X :

[0228] F′ X =F⊙W X;

[0229] Multiply the vertical weight by the input feature map F element by element to obtain the vertical weighted feature F′ Y :

[0230] F′ Y =F⊙W Y ;

[0231] In the formula, F′ X and F′ Y Respectively represent horizontal weighted features and vertical weighted features; ⊙ represents element-by-element multiplication operation;

[0232] S4104, the horizontal weighted feature F' X and the longitudinal weighted feature F′ Y After fusion, we get the final feature F' with spatial attention enhancement:

[0233] F'=F' X +F′ Y ;

[0234] S42, fusing and analyzing the temperature data Tdata, humidity data Hdata and vibration data Vdata, the specific implementation process is as follows:

[0235] S4201, define state variable x i : Indicates the current health status of the system;

[0236] Define the measurement variable z i : Indicates the real-time collected temperature, humidity, and vibration data, including temperature data Tdata i 、Humidity data Hdata i And vibration data Vdata i , that is, z i =[Tdata i , Hdata i , Vdata i ] T ;

[0237] S4202, predict the state variables at the next moment And corresponds to the uncertainty of the predicted state, using the covariance matrix P i|i-1 Indicates the error range of the predicted state:

[0238]

[0239] In the formula, Indicates time T i The predicted state, that is, at the last moment T i-1 Under the observation conditions, for the current moment T iThe predicted value of the health status of the 5G infrastructure; A represents the state transfer matrix, which describes the dynamic characteristics of the system state transfer over time. In this embodiment, the unit matrix is ​​used; Indicates the last moment T i-1 The estimated state, that is, based on the last moment T i-1 The estimated value of health status after data processing; w i represents process noise, which indicates the change of system state caused by unknown interference factors; P i|i-1 The covariance matrix of the predicted state reflects the confidence of the system in the state prediction. The larger the value, the lower the confidence in the prediction. i-1|i-1 Indicates the last moment T i-1 The state estimation covariance matrix represents the prediction uncertainty of the previous moment; Q represents the process noise covariance matrix, which measures the process noise w i The uncertainty of is estimated through a priori data or system identification methods;

[0240] S4203, calculate Kalman gain K i , and use the measured variable z i Update the system state and get the state estimate Then update the covariance matrix P of the state estimate i|i , reflecting the system's confidence in the current state estimate:

[0241] K i =P i|i-1 H T (HP i|i-1 H T +R) -1 ;

[0242]

[0243] P i|i =(IK i H)P i|i-1 ;

[0244] In the formula, K i represents the Kalman gain matrix, that is, the weight of the measurement data in the current state estimation; H represents the measurement matrix, which is used to describe the relationship between the measurement data and the state variables. In this embodiment, the measurement data is the temperature data Tdata i 、Humidity data Hdata i And vibration data Vdata i , using the unit matrix; R represents the measurement noise covariance matrix, reflecting the degree of measurement error of the sensor; Indicates the current time T i The state estimation of , i.e., the health state estimation after combining the prediction and the current measurement; Represents the residual (innovation), that is, the difference between prediction and measurement. The larger the residual, the greater the gap between prediction and measurement. The Kalman gain will increase, which will increase the influence of measurement data on the final estimate. i|i Indicates the current time T i The covariance matrix of represents the updated uncertainty range; I represents the identity matrix, which is used to keep the dimensions of matrix operations consistent;

[0245] S4204, construct the health status time series data Y of 5G infrastructure = {y1, y2, ..., y i , ..., y n};

[0246] Among them, y i Indicates time T i health status data;

[0247] S4205: Build a fault risk prediction model. The construction process is as follows:

[0248] (1) Collect time series data related to the health status of the equipment, including temperature, humidity, and vibration data, ensure that the frequency of the data is consistent (including but not limited to every minute, every hour, or every day), and merge them into a time series, which is defined as training data;

[0249] (2) Use the Augmented Dickey-Fuller (ADF) test to check whether the time series is stationary. If the series is not stationary, perform difference processing to eliminate the trend, and the number of differences is d;

[0250] (3) Draw ACF and PACF graphs and observe the significance (with 95% confidence interval as reference) to select appropriate p' and q' values;

[0251] Among them, the autocorrelation function (ACF) is used to identify the order q' of the moving average (MA) part; the partial autocorrelation function (PACF) is used to identify the order p' of the autoregressive (AR) part;

[0252] (4) Define the ARIMA (Auto-Regressive Integrated Moving Average) model: ARIMA(p', d, q');

[0253] (5) Train the ARIMA model, use the training data to fit the ARIMA model, and estimate the model parameters: autoregressive coefficients and moving average coefficients by minimizing the error, so that the model can best fit the existing training data;

[0254] (6) Repeat steps (2) to (5), use the Akaike Information Criterion (AIC) to evaluate the performance of different ARIMA models, and select the model with the lowest AIC value as the fault risk prediction model;

[0255] It should be noted that the Akaike Information Criterion (AIC) is a criterion for model selection, and the formula is as follows:

[0256] AIC = -2log(L) + 2k;

[0257] In the formula, L represents the likelihood function value of the model, which means the fitness of the model parameters under given data; k represents the number of parameters of the model;

[0258] S4206, based on the trained fault risk prediction model, the health status time series data Y = {y1, y2, ..., y i , ..., y n} is the input parameter, predicting the future T i+h The health of 5G infrastructure at all times The prediction formula is as follows:

[0259]

[0260] In the formula, Represents the predicted future T i+h Keep track of the health of your 5G infrastructure. represents the predicted value of past health status; ε i+h-j represents past forecast errors;

[0261] S43, the future T i+h The health of 5G infrastructure at all times Transmitted to the intelligent decision-making module and system visualization module.

[0262] S5, intelligent decision module receives future T i+h The health of 5G infrastructure at all times Based on the decision fusion rule base, intelligent decision-making is carried out, and the first-level risk threshold T is set based on the historical health status data and the normal working range of the equipment. task1 , Secondary risk threshold T task2 and the third-level risk threshold T task3 , the future T i+h The health of 5G infrastructure at all times With the first-level risk threshold T task1 , Secondary risk threshold T task2 and the third-level risk threshold Ttask3 Make a comparison;

[0263] Among them, the decision fusion rule base is as follows:

[0264] like When the 5G infrastructure is in normal state, the level is 0 and the system continues to work.

[0265] like The 5G infrastructure is deemed to be in a low-risk state, triggering a level 1 response, performing self-diagnosis operations, reviewing the temperature, humidity, and vibration values, and determining whether it is a sensor failure or a short-term fluctuation. After the self-diagnosis is completed, if the fluctuation disappears, the normal state is restored, otherwise the recording continues;

[0266] like If the 5G infrastructure is deemed to be at medium risk, a secondary response will be triggered, which will automatically record fault information and immediately send an emergency alert to the visualization platform to notify maintenance personnel.

[0267] like Or if image failure information is received, the 5G infrastructure status is considered to be at high risk, triggering a level 3 response, immediately sending an emergency alert to the visualization platform and triggering an emergency plan. For example, in the case of severe temperature failure, the power supply of related equipment is automatically cut off to prevent overheating or other damage to the equipment;

[0268] The 5G infrastructure status level is then transmitted to the system visualization module.

[0269] S6. The system visualization module displays the collected temperature data, humidity data, vibration data, image data, 5G infrastructure status level and intelligent scheduling decisions in real time through a graphical interface.

[0270] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A multi-sensor fusion intelligent detection method for 5G infrastructure, characterized in that: The specific implementation steps include the following: S1. Collect appearance image data, temperature data, humidity data and vibration data of 5G infrastructure, and review the integrity and availability of the image data, temperature data, humidity data and vibration data, and then pre-process the collected image data, temperature data, humidity data and vibration data; S2. Build a visual fault detection model, use image data as model input parameters, input it into the trained visual fault detection model, detect the 5G infrastructure, identify whether there is a fault, and output image fault information if a fault is detected; S3: Fusion analysis of temperature data, humidity data, and vibration data to build a fault risk prediction model and predict the health status of 5G infrastructure; S4. Combine image fault information and predicted health status of 5G infrastructure, automatically classify the 5G infrastructure status according to the decision fusion rule base, and make automatic intelligent scheduling decisions based on the 5G infrastructure status level; S5, real-time display of temperature data, humidity data, vibration data, image data, 5G infrastructure status level and intelligent scheduling decisions.

2. According to claim 1, a multi-sensor fusion intelligent detection method for 5G infrastructure is characterized in that: The analysis process of integrating temperature data, humidity data and vibration data is as follows: S21. Define the state variable x representing the current health status i and the measured variables z representing the real-time collected temperature, humidity, and vibration data i =[Tdata i , Hdata i , Vdata i ] T ; Among them, Tdata i 、Hdata i 、Vdata i Respectively represent temperature data, humidity data and vibration data; S22. Predict the state variables at the next moment And corresponds to the uncertainty of the predicted state, using the covariance matrix P i|i-1 Indicates the error range of the predicted state: In the formula, Indicates time T i The predicted state, that is, at the last moment T i-1 Under the observation conditions, for the current moment T i The predicted value of the health status of the 5G infrastructure; A represents the state transition matrix; Indicates the last moment T i-1 The estimated state, that is, based on the last moment T i-1 The estimated value of health status after data processing; w i represents process noise, i.e., state changes caused by unknown interference factors; P i|i-1 represents the covariance matrix of the predicted state; P i-1|i-1 Indicates the last moment T i-1 The state estimation covariance matrix is ​​the prediction uncertainty of the previous moment; Q represents the process noise covariance matrix, which measures the process noise w i uncertainty; S23, calculate Kalman gain K i , and use the measured variable z i Update 5G infrastructure status and get status estimate Then update the covariance matrix P of the state estimate i|i , reflecting the confidence of the current state estimate: K i =P i|i-1 H T (HP i|i-1 H T +R) -1 ; P i|i =(I-K i H)P i|i-1 ; In the formula, K i represents the Kalman gain matrix, which represents the weight of the measurement data in the current state estimation; H represents the measurement matrix, which is used to describe the temperature data Tdata i 、Humidity data Hdata i And vibration data Vdata i The relationship between and state variables is represented by the unit matrix; R represents the measurement noise covariance matrix, which reflects the degree of measurement error of the sensor; Indicates the current time T i The state estimation of , i.e., the health state estimation after combining the prediction and the current measurement; represents the residual, that is, the difference between the prediction and the measurement; P i|i Indicates the current time T i The covariance matrix of is the updated uncertainty range; I represents the identity matrix; S24, construct the health status time series data Y of 5G infrastructure = {y1, y2, ..., y i , ..., y n }; Among them, y i Indicates time T i Health status data of 5G infrastructure; S25, based on the trained fault risk prediction model, the health status time series data of 5G infrastructure Y = {y1, y2, ..., y i , ..., y n } is the input parameter, predicting the future T i+h The health of 5G infrastructure at all times The prediction formula is as follows: In the formula, Represents the predicted future T i+h Keep track of the health of your 5G infrastructure. represents the predicted value of past health status; ε i+h-j represents the past forecast error.

3. A multi-sensor fusion intelligent detection method for 5G infrastructure according to claim 2, characterized in that: The training process of the fault risk prediction model is as follows: S31. Collect time series data related to the health status of the equipment, including temperature, humidity and vibration data, ensure that the frequency of the data is consistent, and merge the temperature, humidity and vibration data into a time series, which is defined as training data; S32. Use the Augmented Dickey-Fuller test to check whether the time series is stationary. If the series is not stationary, perform difference processing to eliminate the trend. The number of differences is d. S33, draw the autocorrelation function and partial autocorrelation function graphs, take the 95% confidence interval as a reference, observe the significance, and select appropriate p' and q' values; Among them, the autocorrelation function is used to identify the order q' of the moving average part; the partial autocorrelation function is used to identify the order p' of the autoregressive part; S34. Define the ARIMA model: ARIMA(p', d, q'); S35, train the ARIMA model, use the training data to fit the ARIMA model, and estimate the model parameters: autoregressive coefficient and moving average coefficient by minimizing the error, so that the model fits the existing training data; S36. Repeat steps S32 to S35, use the Akaike Information Criterion to evaluate the performance of different ARIMA models, and select the model with the lowest AIC value as the fault risk prediction model.

4. A multi-sensor fusion intelligent detection method for 5G infrastructure according to claim 1, characterized in that: The visual fault detection model integrates the coordinate attention module in the backbone feature extraction network based on the YOLOv8 network structure.

5. A multi-sensor fusion intelligent detection method for 5G infrastructure according to claim 4, characterized in that: The implementation steps of the coordinate attention module are as follows: S51, perform average pooling operations on the input feature map F in the horizontal and vertical directions respectively, compressing the features into information in the height direction and compressing the features into information in the width direction respectively: In the formula, F X and F Y Respectively represent the compressed horizontal feature representation and the compressed vertical feature representation; AvgPool W () means that the feature map of each channel is averaged in the width direction (i.e. each row) to generate features in the height direction; AvgPool H () indicates that the feature map of each channel is averaged in the height direction (i.e., each column) to generate features in the width direction; F represents the input feature map; S52, through the convolution operation, the horizontal feature representation F X and the longitudinal feature representation F Y Mapped to the attention weight matrix respectively, we get the horizontal attention weight matrix W X and the vertical attention weight matrix W Y : Where σ represents the activation function; Conv 1×1 () represents a 1×1 convolution kernel, which is used to generate attention weights; W X and W Y Represent the horizontal attention weight matrix and the vertical attention weight matrix respectively; S53, generated horizontal attention weight matrix W X and the vertical attention weight matrix W Y The weights of the input feature map F are rescaled horizontally and vertically respectively, resulting in a spatially enhanced feature representation: Multiply the horizontal weights by the input feature map F element by element to obtain the horizontal weighted feature F X ': F X '=F⊙W X ; Multiply the vertical weight by the input feature map F element by element to get the vertical weighted feature F Y ': F Y '=F⊙W Y ; In the formula, F X ' and F Y ' represents the horizontal weighted features and the vertical weighted features respectively; ⊙ represents the element-by-element multiplication operation; S54, the horizontal weighted feature F X ' and the longitudinal weighted feature F Y 'Fuse and obtain the final feature F with spatial attention enhancement': F'=F X '+F Y '。 6. A multi-sensor fusion intelligent detection system for 5G infrastructure, which is used to execute a multi-sensor fusion intelligent detection method for 5G infrastructure according to any one of claims 1 to 5, characterized in that: include: The sensor data acquisition module is used to collect and transmit the status data of the 5G infrastructure, including image sensors, temperature and humidity sensors, vibration sensors, and data acquisition and transmission units; Image sensors for real-time monitoring of appearance image data of 5G infrastructure; Temperature and humidity sensors, used to collect temperature and humidity data of 5G infrastructure; Vibration sensors to collect vibration data from 5G infrastructure; A data acquisition and transmission unit, used to obtain real-time image data, temperature data, humidity data and vibration data from the image sensor, the temperature and humidity sensor and the vibration sensor, and generate a joint certificate, and transmit {joint certificate, image data, temperature data, humidity data, vibration data} to the data processing module; The data processing module is used to build a data interaction protocol with the data acquisition and transmission unit in the system initialization phase, and to review the received joint certificate in the multi-sensor fusion intelligent detection phase, verify the integrity and availability of image data, temperature data, humidity data, and vibration data, and perform denoising and standardization processing on the image data, temperature data, humidity data, and vibration data; The data fusion and analysis module is used to build a fault risk prediction model, fuse and analyze temperature data, humidity data, and vibration data, predict the health status of 5G infrastructure, and build a visual fault detection model to detect 5G infrastructure and identify whether there is a fault. If a fault is detected, the image fault information is output and the predicted health status of the 5G infrastructure and the image fault information are transmitted to the intelligent decision-making module; An intelligent decision-making module, which is used to classify the status of 5G infrastructure and automatically generate maintenance recommendations based on the predicted health status and image fault information of 5G infrastructure; The system visualization module is used to display the collected temperature data, humidity data, vibration data, image data, 5G infrastructure status level and intelligent scheduling decisions in real time through a graphical interface.

7. A multi-sensor fusion intelligent detection system for 5G infrastructure according to claim 6, characterized in that: The specific implementation process of the data interaction protocol is as follows: S71, the data processing module implements the following operations: S7101. Define a cyclic group G of order q, where P is a generator of the cyclic group G; Where q>2 k , k is an arbitrarily selected security parameter; S7102. Define the following secure hash function: Wherein, L1 is the length of the binary string of the set image sensor, temperature and humidity sensor or vibration sensor device identifier, and L2 is the length of the binary string of the specified collected image data, temperature data, humidity data or vibration data; is the multiplicative group: S7103, Randomly select the first-level audit factor Calculate the secondary audit factor F s2 =F s1 ×P; S7104, save the first level audit factor F s1 and common parameters {q, P, G, F s2 , H1, H2}, and disclose the public parameters {q, P, G, F s2 , H1, H2}; S72, the data acquisition and transmission unit performs the following operations: S7201, select a unique random number for the image sensor Select a unique random number for the temperature and humidity sensor Select a unique random number for the vibration sensor And calculate the permission application parameter P A1 =α1×P, temperature and humidity sensor permission application parameter P A2 =α2×P, vibration sensor permission application parameter P A3 =α3×P; S7202, obtain the device identification of the image sensor, temperature and humidity sensor, and vibration sensor, and establish an associated group <ID1,P A1 >, Association Group <ID2,P A2 >, Association Group <ID3,P A3 > Among them, ID1, ID2 and ID3 are the device identifiers of the image sensor, temperature and humidity sensor and vibration sensor respectively; S7203, save the image sensor exclusive random number α1, the temperature and humidity sensor exclusive random number α2, the vibration sensor exclusive random number α3, and the association group <ID1,P A1 >, Association Group <ID2,P A2 >, Association Group <ID3,P A3 > and associate the group <ID1,P A1 >, Association Group <ID2,P A2 >, Association Group <ID3,P A3 >Transmit to data processing module; S73, the data processing module implements the following operations: S7301. Select auxiliary authority control factor for image sensor Select auxiliary permission control factors for temperature and humidity sensors Select auxiliary permission control factors for vibration sensors S7302. Calculate the primary data collection authority parameter γ of the image sensor 11 =β1×P, secondary data collection authority parameter γ 12 =β1+F s1 ×H1(ID1,P A1 , γ 11 ); Calculate the primary data collection authority parameter γ of the temperature and humidity sensor 21 =β2×P, secondary data collection authority parameter γ 22 =β2+F s1 ×H1(ID2,P A2 , γ 21 ); Calculate the primary data collection authority parameter γ of the vibration sensor 31 =β3×P, secondary data collection authority parameter γ 32 =β3+F s1 ×H1(ID3,P A3 , γ 31 ); S7303, create and save association group <ID1,P A1 , γ 11 , γ 12 >, Association Group <ID2,P A2 , γ 21 , γ 22 > and associated groups <ID3,P A3 , γ 31 , γ 32 >, then associate the group <ID1,P A1 , γ 11 , γ 12 >, Association Group <ID2,P A2 , γ 21 , γ 22 > and associated groups <ID3,P A3 , γ 31 , γ 32 >Transmit to the data acquisition and transmission unit.

8. The multi-sensor fusion intelligent detection system for 5G infrastructure according to claim 6, characterized in that: The process of generating a joint certificate is as follows: S81, image data Pdata i Generate a data certificate. The generation process is as follows: S8101, select random number S8102. Calculate the following certificate generation factors: CG P1 =κ1×P; CG P2 =H2(ID1,Pdata i ,CG P1 ); CG P3 =κ1+(α1+γ 12 )×CG P2 ; S8103, generate image data Pdata i Data certificate Cert1 = (CG P1 , C.G. P3 ); S82, temperature data Pdata i Generate a data certificate. The generation process is as follows: S8201, select random number S8202. Calculate the following certificate generation factors: CG T1 =κ2×P; CG T2 =H2(ID2,Tdata i ,CG T1 ); CG T3 =κ2+(α2+γ 22 )×CG T2 ; S8203, generate temperature data Tdata i Data certificate Cert2 = (CG T1 , C.G. T3 ); S83, humidity data Hdata i Generate a data certificate. The generation process is as follows: S8301, select random number S8302. Calculate the following certificate generation factors: CG H1 =κ3×P; CG H2 =H2(ID2,Hdata i ,CG H1 ); CG H3 =κ3+(α2+γ 22 )×CG H2 ; S8303, generate humidity data Hdata i Data certificate Cert3 = (CG H1 , C.G. H3 ); S84, vibration data Vdata i Generate a data certificate. The generation process is as follows: S8401, select random number S8402. Calculate the following certificate generation factors: CG V1 =κ4×P; CG V2 =H2(ID3,Vdata i ,CG V1 ); CG V3 =κ4+(α3+γ 32 )×CG V2 ; S8403, generate vibration data Vdata i Data certificate Cert3 = (CG V1 , C.G. V3 ); S85, Computing Joint Certificate Level 1 JCert1 = CG P1 +CG T1 +CG H1 +CG V1 , calculate the second level joint certificate JCert2 = CG P3 +CG T3 +CG H3 +CG V3 ; S86. Generate a joint certificate JCert={JCert1, JCert2}.

9. A multi-sensor fusion intelligent detection system for 5G infrastructure according to any one of claims 6, 7 or 8, characterized in that: The audit process for the received joint certificates is as follows: S9101. Calculate the following audit factors: F AP1 =H2(ID1,P A1 ,c 11 ); F AP2 =H2(ID1,Pdata i ,CG P1 , F AT1 =H2(ID2,P A2 ,c 21 ); F AT2 =H2(ID2,Tdata i ,CG T1 ); F AH1 =H2(ID2,P A2 ,c 21 ); F AH2 =H2(ID2,Hdata i ,CG H1 ); F AV1 =H2(ID3,P A3 ,c 31 ); F AV2 =H2(ID3,Vdata i ,CG V1 ); S9102, calculate the first-level audit parameter A1 = JCert2 × P, calculate the second-level audit parameter A2 = [F AP2 ×(P A1 +γ 11 +F s2 ×F AP1 )]+[F AT2 ×(P A2 +γ 21 +F s2 ×F AT1 )]+[F AH2 ×(P A2 +γ 21 +F s2 ×F AH1 )]+[F AV2 ×(P A3 +γ 31 +F s2 ×F AV1 )]+JCert1; If A1=A2, the joint certificate is approved, which means the received image data Pdata is confirmed. i 、Tdata i 、Humidity data Hdata i And vibration data Vdata i completeness and availability.

Images