Intelligent security monitoring system and monitoring method
By using deep learning models and statistical methods to extract and analyze environmental data in the security monitoring system, a multi-model system is built, which solves the problem of low processing efficiency and accuracy of traditional security monitoring systems, and realizes intelligent analysis and automatic alarm Function, improve monitoring efficiency and security.
Patent Information
- Application Number
- CN202410987594.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-23
- Publication Date
- 2024-11-19
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional security monitoring systems lack intelligent analysis and processing capabilities, cannot effectively deal with complex and changeable monitoring environments, and have low processing efficiency and accuracy.
Using data acquisition module, data processing module, feature extraction module, model building module, multi-model fusion module, alarm strategy module and real-time monitoring module, feature extraction and analysis of environmental data through deep learning models and statistical methods, and construct convolution based on convolution A multi-model system of neural networks, recurrent neural networks and long-term memory networks realizes real-time monitoring and automatic alarms.
It improves the intelligent analysis capabilities of the monitoring system, can accurately identify abnormal events and automatically alarm, improves monitoring efficiency and security, and is suitable for various security monitoring scenarios.
Smart Images

Figure CN118984322A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of security monitoring, and in particular to an intelligent security monitoring system and a monitoring method. Background Art
[0002] The security monitoring system uses optical fiber, coaxial cable or microwave to transmit video signals in a closed loop, and constitutes an independent and complete system from camera to image display and recording. It can reflect the monitored object in real time, vividly and truthfully, and can replace manual long-term monitoring in harsh environments and record it through a video recorder. With the advancement of science and technology and the development of society, the application of security monitoring systems is becoming more and more extensive.
[0003] However, traditional security monitoring systems often lack intelligent analysis and processing capabilities and cannot effectively cope with complex and changing monitoring environments. At the same time, existing security monitoring systems often rely on simple video surveillance and basic sensor data, making it difficult to achieve efficient real-time monitoring and intelligent analysis. In addition, existing systems have low processing efficiency and accuracy when faced with complex environments and abnormal situations.
[0004] Therefore, it is of great significance to develop an intelligent security monitoring system that can intelligently analyze, accurately identify abnormal events and automatically alarm. Summary of the invention
[0005] The purpose of the present invention is to provide an intelligent security monitoring system and a monitoring method to solve the problems raised in the above background technology.
[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:
[0007] In a first aspect, an intelligent security monitoring method comprises the following steps:
[0008] Step 1, data acquisition module: collect environmental data based on various sensor devices, including video stream, audio stream, temperature, humidity and smoke concentration data information;
[0009] Step 2: Data processing module performs cleaning and normalization preprocessing operations on the collected data to ensure the quality and consistency of the data;
[0010] Step 3: Feature extraction module, which extracts and selects features from the preprocessed data based on deep learning models and statistical methods, and screens out key features related to security events;
[0011] Step 4: Model building module, building a multi-model system based on complex functions, including convolutional neural networks, recursive neural networks, and long short-term memory networks, to train and evaluate the extracted features;
[0012] The model training process consists of repeated cycles, as follows:
[0013] S1: Initialize parameters, first initialize the parameters of the model;
[0014] S2: Data batch processing, dividing the training data into multiple batches, each batch contains several samples;
[0015] S3: For each batch of data, forward propagation is performed to calculate the output value;
[0016] S4: Calculate the loss function: Calculate the loss function using the result of forward propagation;
[0017] S5: Calculate the gradient of the loss function with respect to the model parameters;
[0018] S6: Update parameters using gradient descent or its variants;
[0019] S7: loop through the above steps S1-S6 until the loss function converges to a certain threshold;
[0020] Step 5: The multi-model fusion module scores the monitoring data in real time based on multiple trained models, and determines whether to trigger an alarm based on the scoring results;
[0021] Step 6, alarm strategy module, sets alarm strategy based on the scoring results, optimizes the alarm strategy through algorithms, and ensures the accuracy and real-time nature of the alarm;
[0022] Step 7: Real-time monitoring module monitors environmental data in real time, automatically adjusts monitoring parameters according to alarm strategies, and optimizes model performance through feedback mechanisms;
[0023] Step 8: The result analysis module performs statistical analysis based on the monitoring data, generates a monitoring report, and provides optimization suggestions.
[0024] A further improvement of the technical solution of the present invention is that the data acquisition module includes a storage, a video sensor, a sound sensor, a temperature sensor, a humidity sensor and a smoke sensor to respectively collect the environment including video stream, audio stream, temperature, humidity and smoke concentration data information, and express it as: and collect the video stream V = {V1, V2..., V m}, where V m represents the mth data, m is a positive integer, and the audio stream and the collected data is marked as P = {P1, P2..., P n}, where P n represents the nth data, n is a positive integer; the temperature is marked as T = {T1, T2..., T r}, where T rrepresents the rth data, r is a positive integer, and the humidity is S = {S1, S2..., S a}, where S a represents the a-th data, a is a positive integer; the smoke concentration is marked as Q = {Q1, Q2..., Q l}, where Q l It is represented as the lth data, where l is a positive integer, and the collected data is stored in the memory.
[0025] A further improvement of the technical solution of the present invention is that the data processing module includes a processor, and the processor performs cleaning and normalization preprocessing operations on the collected data. The specific processing steps are as follows:
[0026] The data cleaning formula is as follows:
[0027] D clean ={d i ||d i -μ|≤3σ}
[0028] Where μ is the mean of feature d, σ is the standard deviation of d, and D is the value of the above-mentioned video stream V, audio stream P, temperature T, humidity V, and smoke concentration data Q.
[0029] Normalization formula:
[0030]
[0031] A further improvement of the technical solution of the present invention is that: the feature extraction module extracts features from the preprocessed data, and the feature extraction formula is:
[0032] H=Φ(W·X+b)
[0033] Where Φ is the activation function, W is the weight matrix, b is the bias, and X is the value of the above-mentioned video stream V, audio stream P, temperature T, humidity V and smoke concentration data Q after feature processing.
[0034] A further improvement of the technical solution of the present invention is that the model building module includes initialization parameters as follows:
[0035] Initialize the parameters based on the normal distribution. The formula for the normal distribution is:
[0036]
[0037] in, To represent the mean as μ and the variance as σ 2 The normal distribution of .
[0038] The formula for the forward propagation is:
[0039] h i =ζ(G c ·X+a z )
[0040]
[0041] Among them, ζ is the activation function, τ is the activation function of the output layer G c 、a z The model's weight and bias parameters, h i is the forward propagation function, where i is a positive integer.
[0042] The loss function is calculated based on the result of forward propagation, and the specific analysis using the cross entropy loss function is as follows:
[0043]
[0044] where y i is the actual value, is the predicted value, θ is the set of all trainable parameters in the neural network, and B is the number of samples in the batch.
[0045] The gradient of the loss function to the model parameters is calculated to update the parameters. The back propagation formula is as follows:
[0046]
[0047] in is the loss function L for the i-th weight W i The gradient of is the loss function L for the i-th weight b i The gradient of , the number of samples in batch B, y i is the actual value, is the predicted value.
[0048] The gradient descent updates the parameters, and the specific formula is as follows:
[0049]
[0050] in is the learning rate, is the gradient of the loss function, θ is the set of all trainable parameters in the neural network, including weights and biases, θ e Represents the parameter value at the e-th iteration.
[0051] The loop executes the above steps until
[0052]
[0053] At this point the loop terminates and the output is:
[0054] Where θ is the set of all trainable parameters in the neural network, including weights and biases, θ e represents the parameter value at the eth iteration, the number of samples in the B batch, and y i is the actual value, is the predicted value, and L(θ) is the loss function.
[0055] The multi-model fusion module uses multiple trained models to score the monitoring data in real time. The specific formula is:
[0056]
[0057] Among them v is the model weight, f v (X) is the output of the v-th model, A(X) represents the optimized score of the monitoring feature set X, and X is the value of the video stream V, audio stream P, temperature T, humidity V and smoke concentration data Q after feature processing, and R is the number of models involved in the scoring.
[0058] The alarm strategy module includes a warning light. The alarm strategy module sets the alarm strategy based on the scoring result. Through the cyclic optimization algorithm, the specific formula is as follows:
[0059]
[0060] Where X is the value of the video stream V, audio stream P, temperature T, humidity V and smoke concentration data Q after feature processing, N is the number of evaluation models, κ i is the weight of the i-th evaluation model, Score i (X) is the evaluation result of the i-th evaluation model on the feature set X. When F(X)>F 阀值 (X) Warning light gives warning, otherwise no warning is given, where F 阀值 (X) is the threshold of the scoring result.
[0061] The intelligent security monitoring method is characterized in that the real-time monitoring module monitors the environmental data in real time and automatically adjusts the monitoring parameters based on the alarm strategy. The monitoring formula is:
[0062]
[0063] Where K(g) represents the monitoring parameter at time g, Ω is the number of monitoring parameters, ∫ g is the weight of the γth evaluation model, is the function of the γth monitoring parameter, is the environmental data at time g.
[0064] The result analysis module performs statistical analysis based on the monitoring data and proposes specific suggestions for adjusting the monitoring strategy, improving the system design or optimizing the data processing flow according to the problems found.
[0065] An intelligent security monitoring system, used to implement an intelligent security monitoring method as described in any one of claims 1-13, characterized in that it includes a data acquisition module, a preprocessing module, a feature extraction module, a model building module, a multi-model fusion module, an alarm strategy module, a real-time monitoring module and a result analysis module, wherein each module is electrically connected.
[0066] Including data acquisition module: collecting environmental values through the data acquisition module;
[0067] The preprocessing module: improves data quality through data cleaning and normalization;
[0068] The feature extraction module uses complex functions and deep learning models to extract features;
[0069] The model building module includes a convolutional neural network, a recursive neural network and a long short-term memory network, and optimizes model parameters through a loop step;
[0070] The multi-model fusion module: calculates the comprehensive score by weighted average or other fusion strategies;
[0071] The alarm strategy module: adjusts the alarm strategy through an optimization algorithm to improve the accuracy of the alarm;
[0072] The real-time monitoring module: monitors environmental data in real time and automatically adjusts monitoring parameters according to the alarm strategy;
[0073] The result analysis module analyzes the monitoring results, generates a detailed monitoring report and provides optimization suggestions.
[0074] Due to the adoption of the above technical solution, the present invention has the following technical advances compared with the prior art:
[0075] 1. The present invention provides an intelligent security monitoring system and a monitoring method, which collects environmental values through a data acquisition module, cleans and normalizes the data through a preprocessing module to improve data quality, uses complex functions and deep learning models to extract features, and uses a model construction module: including a convolutional neural network, a recursive neural network, and a long short-term memory network, and optimizes model parameters through a loop step, which is conducive to improving data processing efficiency;
[0076] 2. The present invention provides an intelligent security monitoring system and monitoring method, which uses a multi-model fusion module: uses weighted average or other fusion strategies to calculate the comprehensive score and uses an alarm strategy module to adjust the alarm strategy through an optimization algorithm to improve the accuracy of the alarm;
[0077] 3. The present invention provides an intelligent security monitoring system and monitoring method, which realizes environmental data through a real-time monitoring module and automatically adjusts monitoring parameters according to the alarm strategy; at the same time, the result analysis module is used to analyze the monitoring results, generate detailed monitoring reports and provide optimization suggestions, further improving the purpose of accurate monitoring of the system;
[0078] 4. The present invention provides an intelligent security monitoring system and a monitoring method. The system realizes high-precision intelligent monitoring of the monitoring area by constructing a complex function model, combining sophisticated formula calculations, and setting repeated detection cycle steps. The system has the functions of intelligent analysis, accurate identification of abnormal events and automatic alarm, and can be widely used in various security monitoring scenarios to improve monitoring efficiency and safety. BRIEF DESCRIPTION OF THE DRAWINGS
[0079] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.
[0080] Figure 1 A flow chart of an intelligent security monitoring system and monitoring method of the present invention;
[0081] Figure 2 A model building module diagram of an intelligent security monitoring system and monitoring method of the present invention; DETAILED DESCRIPTION
[0082] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0083] Embodiment 1, as Figure 1 and Figure 2 As shown, the present invention provides an intelligent security monitoring method, comprising the following steps:
[0084] Step 1, data acquisition module: collect environmental data based on various sensor devices, including video stream, audio stream, temperature, humidity and smoke concentration data information;
[0085] Step 2: Data processing module cleans and normalizes the collected data to ensure data quality and consistency;
[0086] Step 3: Feature extraction module extracts and selects features from preprocessed data based on deep learning models and statistical methods to screen out key features related to security events.
[0087] Step 4: Model building module, building a multi-model system based on complex functions, including convolutional neural networks, recursive neural networks, and long short-term memory networks, to train and evaluate the extracted features;
[0088] The model training process consists of repeated cycles, as follows:
[0089] S1: Initialize parameters, first initialize the parameters of the model;
[0090] S2: Data batch processing, dividing the training data into multiple batches, each batch contains several samples;
[0091] S3: For each batch of data, forward propagation is performed to calculate the output value;
[0092] S4: Calculate the loss function: Calculate the loss function using the result of forward propagation;
[0093] S5: Calculate the gradient of the loss function with respect to the model parameters;
[0094] S6: Update parameters using gradient descent or its variants;
[0095] S7: loop through the above steps S1-S6 until the loss function converges to a certain threshold;
[0096] Step 5: The multi-model fusion module scores the monitoring data in real time based on multiple trained models, and determines whether to trigger an alarm based on the scoring results;
[0097] Step 6, alarm strategy module, sets alarm strategy based on the scoring results, optimizes the alarm strategy through algorithms, and ensures the accuracy and real-time nature of the alarm;
[0098] Step 7: Real-time monitoring module monitors environmental data in real time, automatically adjusts monitoring parameters according to alarm strategies, and optimizes model performance through feedback mechanisms;
[0099] Step 8: The result analysis module performs statistical analysis based on the monitoring data, generates a monitoring report, and provides optimization suggestions.
[0100] Embodiment 2, as Figure 1 and Figure 2 As shown, the data acquisition module includes storage, video sensor, sound sensor, temperature sensor, humidity sensor and smoke sensor to collect environment data information including video stream, audio stream, temperature, humidity and smoke concentration, and is expressed as: and collects the video stream V = {V1, V2..., V m}, where V m represents the mth data, m is a positive integer, and the audio stream and the collected data is marked as P = {P1, P2..., P n}, where P n represents the nth data, n is a positive integer; the temperature is marked as T = {T1, T2..., T r}, where T r represents the rth data, r is a positive integer, and the humidity is S = {S1, S2..., S a}, where S a represents the a-th data, a is a positive integer; the smoke concentration is marked as Q = {Q1, Q2..., Q l}, where Q l It is represented as the lth data, where l is a positive integer, and the collected data is stored in the memory.
[0101] The data processing module includes a processor, which performs cleaning and normalization preprocessing operations on the collected data. The specific processing steps are as follows:
[0102] The data cleaning formula is as follows:
[0103] D clean ={d i ||d i -μ|≤3σ}
[0104] Where μ is the mean of feature d, σ is the standard deviation of d, and d is the value of video stream V, audio stream P, temperature T, humidity V, and smoke concentration data Q after feature processing.
[0105] Normalization formula:
[0106]
[0107] The feature extraction module extracts features from the preprocessed data. The feature extraction formula is:
[0108] H=Φ(W·X+b)
[0109] Where Φ is the activation function, W is the weight matrix, b is the bias, and X is the value of the above-mentioned video stream V, audio stream P, temperature T, humidity V and smoke concentration data Q after feature processing.
[0110] The model building module includes the following initialization parameters:
[0111] Initialize the parameters based on the normal distribution. The formula for the normal distribution is:
[0112]
[0113] in, To represent the mean as μ and the variance as σ 2 The normal distribution of .
[0114] The formula for forward propagation is:
[0115] h i =ζ(G c ·X+a z )
[0116]
[0117] Among them, ζ is the activation function, τ is the activation function of the output layer G c 、a z The model's weight and bias parameters, h i is the forward propagation function, where i is a positive integer.
[0118] The loss function is calculated based on the result of forward propagation, and the specific analysis using the cross entropy loss function is as follows:
[0119]
[0120] where y i is the actual value, is the predicted value, θ is the set of all trainable parameters in the neural network, and B is the number of samples in the batch.
[0121] Calculate the gradient of the loss function to the model parameters to update the parameters. The back propagation formula is as follows:
[0122]
[0123] in is the loss function L for the i-th weight W i The gradient of is the loss function L for the i-th weight b i The gradient of , the number of samples in batch B, y i is the actual value, is the predicted value.
[0124] Gradient descent updates the parameters. The specific formula is as follows:
[0125]
[0126] θ is the set of all trainable parameters in the neural network, including weights and biases. e represents the parameter value at the e-th iteration, L(θ e ) is the loss function.
[0127] The learning rate is a hyperparameter that controls the parameter update step size, which determines the adjustment amplitude of the parameters at each update. At the same time, in each iteration, the parameters of the model are adjusted according to the gradient and learning rate to reduce the value of the loss function and optimize the model performance. The back propagation algorithm calculates the gradient of the loss function for each parameter and uses these gradients to update the parameters of the model. This can gradually reduce the value of the loss function, thereby improving the accuracy and performance of the model.
[0128] Repeat the above steps until
[0129]
[0130] Where B is the number of samples in a batch, that is, the total number of samples contained in a batch, L(θ) is the value of the loss function, which measures the accuracy of the model prediction, and y i is the actual label value of the i-th sample, is the predicted probability value of a sample, indicating the probability that the model predicts that the sample belongs to the positive class, that is, y i =1 probability.
[0131] At this point the loop terminates and the result is output;
[0132] Multi-model fusion module,The multi-model fusion module uses multiple trained models to score the monitoring data in real time. The specific formula is:
[0133]
[0134] Among them v is the model weight, f v (X) is the output of the vth model, R is the number of models involved in the scoring, X is the value of the video stream V, audio stream P, temperature T, humidity V and smoke concentration data Q after feature processing,
[0135] The alarm strategy module includes a warning light. The alarm strategy module sets the alarm strategy based on the scoring results. Through the cyclic optimization algorithm, the specific formula is as follows:
[0136]
[0137] Where X is the value of the video stream V, audio stream P, temperature T, humidity V and smoke concentration data Q after feature processing, N is the number of evaluation models, κ i is the weight of the i-th evaluation model, Score i (X) is the evaluation result of the i-th evaluation model on the feature set X. When F(X)>F 阀值 (X) Warning light gives warning, otherwise no warning is given, where F 阀值 (X) is the threshold of the scoring result.
[0138] The real-time monitoring module monitors environmental data in real time and automatically adjusts monitoring parameters based on the alarm strategy. The monitoring formula is:
[0139]
[0140] Where K(g) represents the monitoring parameter at time g, Ω is the number of monitoring parameters, ∫ g is the weight of the γth evaluation model, is the function of the γth monitoring parameter, is the environmental data at time g.
[0141] The result analysis module performs statistical analysis based on monitoring data and proposes specific suggestions for adjusting monitoring strategies, improving system design, or optimizing data processing procedures based on the problems found.
[0142] Embodiment 3, as Figure 1 and Figure 2 As shown, an intelligent security monitoring system is used to implement an intelligent security monitoring method according to any one of claims 1 to 13, comprising a data acquisition module, a preprocessing module, a feature extraction module, a model building module, a multi-model fusion module, an alarm strategy module, a real-time monitoring module and a result analysis module, wherein each module is electrically connected.
[0143] Data acquisition module: collect environmental values through the data acquisition module;
[0144] Preprocessing module: improve data quality through data cleaning and normalization;
[0145] Feature extraction module: uses complex functions and deep learning models to extract features;
[0146] Model building modules: including convolutional neural networks, recurrent neural networks and long short-term memory networks, and optimizing model parameters through loop steps;
[0147] Multi-model fusion module: Calculate the comprehensive score using weighted average or other fusion strategies;
[0148] Alarm strategy module: adjusts the alarm strategy through optimization algorithm to improve alarm accuracy;
[0149] Real-time monitoring module: monitors environmental data in real time and automatically adjusts monitoring parameters based on alarm strategies;
[0150] Result analysis module: Analyzes the monitoring results, generates detailed monitoring reports and provides optimization suggestions.
[0151] Working principle: Using data acquisition module: Collect environmental data according to various sensor devices, including video stream, audio stream, temperature, humidity and smoke concentration data information. For example, video sensor, sound sensor, temperature sensor, humidity sensor and smoke sensor collect environmental data including video stream, audio stream, temperature, humidity and smoke concentration data information respectively, and express it as: and collect it into video stream V = {V1, V2..., V m}, where V m represents the mth data, m is a positive integer, and the audio stream and the collected data is marked as P = {P1, P2..., P n}, where P n represents the nth data, n is a positive integer; the temperature is marked as T = {T1, T2..., T r}, where T r represents the rth data, r is a positive integer, and the humidity is S = {S1, S2..., S a}, where S a represents the a-th data, a is a positive integer; the smoke concentration is marked as Q = {Q1, Q2..., Q l}, where Q l It is represented as the lth data, where l is a positive integer, and the collected data is stored in the storage; then the data processing module is used to clean and normalize the collected data to ensure the quality and consistency of the data; the specific processing steps are as follows:
[0152] The data cleaning formula is as follows:
[0153] D clean ={d i ||d i -μ|≤3σ}
[0154] Where μ is the mean of feature d, σ is the standard deviation of d, and d is the above-mentioned video stream, audio stream, temperature, humidity and smoke concentration data.
[0155] Normalization formula:
[0156]
[0157] Then, the feature extraction module is used to extract and select features from the preprocessed data based on the deep learning model and statistical methods to screen out key features related to security events. The feature extraction module extracts features from the preprocessed data. The feature extraction formula is:
[0158] H=Φ(W·X+b)
[0159] Where Φ is the activation function, W is the weight matrix, b is the bias, and X is the value of the above-mentioned video stream V, audio stream P, temperature T, humidity V and smoke concentration data Q after feature processing.
[0160] At the same time, the model building module is used to loop the data, including initialization parameters as follows:
[0161] Initialize the parameters based on the normal distribution. The formula for the normal distribution is:
[0162]
[0163] in, To represent the mean as μ and the variance as σ 2 Normal distribution
[0164] The normal distribution is used to initialize the parameters for the video stream V, audio stream P, temperature T, humidity V and smoke concentration data Q after feature processing.
[0165] Data batch processing: In order to improve training efficiency and stability, the training data is divided into multiple batches, each batch contains a number of samples. Assuming that the size of each batch is B and the total amount of training data is N, the number of batches is
[0166] For each batch of data, forward propagation is performed to calculate the output value. Using a simple neural network model, the forward propagation formula is:
[0167] h i =ζ(G c ·X+a z )
[0168]
[0169] Among them, ζ is the activation function, τ is the activation function of the output layer, G c 、a z The weight and bias parameters of the model, X is the value of the above video stream V, audio stream P, temperature T, humidity V and smoke concentration data Q after feature processing.
[0170] The loss function is calculated based on the result of forward propagation, and the specific analysis using the cross entropy loss function is as follows:
[0171]
[0172] where y i The data processed are actual values. The data processed is the predicted value, B is the size of each batch, and L(θ) is the loss function calculated based on the result of forward propagation.
[0173] Calculate the gradient of the loss function to the model parameters to update the parameters. The back propagation formula is as follows:
[0174]
[0175] The specific formula is as follows:
[0176]
[0177] in is the learning rate, is the gradient of the loss function, and B is the size of each batch.
[0178] Repeat the above steps until
[0179]
[0180] At this point the loop terminates and the result is output.
[0181] Then, the multi-model fusion module is used. The multi-model fusion module uses multiple trained models to score the monitoring data in real time. The specific formula is:
[0182]
[0183] Among them i is the model weight, f v (X) Output of the vth model.
[0184] At the same time, the alarm strategy module includes a warning light. The alarm strategy module sets the alarm strategy based on the scoring results. Through the cyclic optimization algorithm, the specific formula is as follows:
[0185]
[0186] Where X represents the monitoring feature set, N is the number of evaluation models, and κ i is the weight of the i-th evaluation model, Score i (X) is the evaluation result of the i-th evaluation model on the feature set X,
[0187] When F(X)>F 阀值 (X) Warning light gives warning, otherwise no warning is given, where F 阀值(X) is the threshold of the scoring result.
[0188] At the same time, the real-time monitoring module is used to monitor the environmental data in real time, and the monitoring parameters are automatically adjusted based on the alarm strategy. The monitoring formula is:
[0189]
[0190] Where K(g) represents the monitoring parameter at time g, Ω is the number of monitoring parameters, ∫ g is the weight of the γth evaluation model, is the function of the γth monitoring parameter, is the environmental data at time g.
[0191] The result analysis module is then used to conduct statistical analysis based on the monitoring data, and specific suggestions are made to adjust the monitoring strategy, improve the system design, or optimize the data processing process based on the problems found.
[0192] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
Claims
1. An intelligent security monitoring method, characterized in that: The following steps are involved: Step 1, data acquisition module: collect environmental data based on various sensor devices, including video stream, audio stream, temperature, humidity and smoke concentration data information; Step 2: Data processing module cleans and normalizes the collected data to ensure data quality and consistency; Step 3: Feature extraction module extracts and selects features from preprocessed data based on deep learning models and statistical methods to screen out key features related to security events. Step 4: Model building module, building a multi-model system based on complex functions, including convolutional neural networks, recursive neural networks, and long short-term memory networks, to train and evaluate the extracted features; The model training process consists of repeated cycles, as follows: S1: Initialize parameters, first initialize the parameters of the model; S2: Data batch processing, dividing the training data into multiple batches, each batch contains several samples; S3: For each batch of data, forward propagation is performed to calculate the output value; S4: Calculate the loss function: Calculate the loss function using the result of forward propagation; S5: Calculate the gradient of the loss function with respect to the model parameters; S6: Update parameters using gradient descent or its variants; S7: loop through the above steps S1-S6 until the loss function converges to a certain threshold; Step 5: The multi-model fusion module scores the monitoring data in real time based on multiple trained models, and determines whether to trigger an alarm based on the scoring results; Step 6, alarm strategy module, sets alarm strategy based on the scoring results, optimizes the alarm strategy through algorithms, and ensures the accuracy and real-time nature of the alarm; Step 7: Real-time monitoring module monitors environmental data in real time, automatically adjusts monitoring parameters according to alarm strategies, and optimizes model performance through feedback mechanisms; Step 8: The result analysis module performs statistical analysis based on the monitoring data, generates a monitoring report, and provides optimization suggestions.
2. The intelligent security monitoring method according to claim 1, characterized in that: The data acquisition module includes a storage, a video sensor, a sound sensor, a temperature sensor, a humidity sensor and a smoke sensor to respectively collect the environment including video stream, audio stream, temperature, humidity and smoke concentration data information, and is expressed as: and the collected video stream V = {V1, V2..., V m }, where V m Represents the mth data, m is a positive integer, audio stream and the collected data is marked as P= {P1, P2..., P n }, where P n represents the nth data, n is a positive integer; the temperature is marked as T = {T1, T2..., T r }, where T r represents the rth data, r is a positive integer, and the humidity is S = {S1, S2..., S a }, where S a represents the a-th data, a is a positive integer; the smoke concentration is marked as Q = {Q1, Q2..., Q l }, where Q l It is represented as the lth data, where l is a positive integer, and the collected data is stored in the memory.
3. The intelligent security monitoring method according to claim 2, characterized in that: The data processing module includes a processor, which performs cleaning and normalization preprocessing operations on the collected data. The specific processing steps are as follows: The data cleaning formula is as follows: D clean ={d i ||d i -μ|≤3σ} Where μ is the mean of feature d, σ is the standard deviation of d, and D is the value of the above-mentioned video stream V, audio stream P, temperature T, humidity V, and smoke concentration data Q. Normalization formula:
4. The intelligent security monitoring method according to claim 3, characterized in that: The feature extraction module extracts features from the preprocessed data, and the feature extraction formula is: H=Φ(W·X+b) Where Φ is the activation function, W is the weight matrix, b is the bias, and X is the value of the above-mentioned video stream V, audio stream P, temperature T, humidity V and smoke concentration data Q after feature processing.
5. The intelligent security monitoring method according to claim 4, characterized in that: The model building module includes initialization parameters as follows: Initialize the parameters based on the normal distribution. The formula for the normal distribution is: in, To represent the mean as μ and the variance as σ 2 The normal distribution of .
6. The intelligent security monitoring method according to claim 5, characterized in that: The formula for the forward propagation is: h i =ζ(G c ·X+a z ) Among them, z is the activation function, τ is the activation function of the output layer G c 、a z The model's weight and bias parameters, h i is the forward propagation function, where i is a positive integer.
7. The intelligent security monitoring method according to claim 6, characterized in that: The loss function is calculated based on the result of forward propagation, and the specific analysis using the cross entropy loss function is as follows: where y i is the actual value, is the predicted value, θ is the set of all trainable parameters in the neural network, and B is the number of samples in the batch. The gradient of the loss function to the model parameters is calculated to update the parameters. The back propagation formula is as follows: in is the loss function L for the i-th weight W i The gradient of is the loss function L for the i-th weight b i The gradient of , the number of samples in batch B, y i is the actual value, For the predicted value The gradient descent updates the parameters, and the specific formula is as follows: in is the learning rate, is the gradient of the loss function, θ is the set of all trainable parameters in the neural network, including weights and biases, θ e Represents the parameter value at the e-th iteration. The loop executes the above steps until At this point the loop terminates and the result is output; Where θ is the set of all trainable parameters in the neural network, including weights and biases, θ e represents the parameter value at the eth iteration, the number of samples in the B batch, and y i is the actual value, is the predicted value, and L(θ) is the loss function.
8. The intelligent security monitoring method according to claim 7, characterized in that: The multi-model fusion module uses multiple trained models to score the monitoring data in real time. The specific formula is: Among them v is the model weight, f v (X) is the output of the v-th model, A(X) represents the optimized score of the monitoring feature set X, and X is the value of the video stream V, audio stream P, temperature T, humidity V and smoke concentration data Q after feature processing, and R is the number of models involved in the scoring. The alarm strategy module includes a warning light. The alarm strategy module sets the alarm strategy based on the scoring result. Through the cyclic optimization algorithm, the specific formula is as follows: Where X is the value of the video stream V, audio stream P, temperature T, humidity V and smoke concentration data Q after feature processing, N is the number of evaluation models, κ i is the weight of the i-th evaluation model, Score i (X) is the evaluation result of the i-th evaluation model on the feature set X. When F(X)>F 阀值 (X) Warning light gives warning, otherwise no warning is given, where F 阀值 (X) is the threshold of the scoring result.
9. The intelligent security monitoring method according to claim 8, characterized in that: The real-time monitoring module monitors the environmental data in real time and automatically adjusts the monitoring parameters based on the alarm strategy. The monitoring formula is: Where K(g) represents the monitoring parameter at time g, Ω is the number of monitoring parameters, ∫ g is the weight of the γth evaluation model, is the function of the γth monitoring parameter, is the environmental data at time g. The result analysis module performs statistical analysis based on the monitoring data and proposes specific suggestions for adjusting the monitoring strategy, improving the system design or optimizing the data processing flow according to the problems found.
10. An intelligent security monitoring system, used to implement an intelligent security monitoring method according to any one of claims 1 to 9, characterized in that: It includes a data acquisition module, a preprocessing module, a feature extraction module, a model building module, a multi-model fusion module, an alarm strategy module, a real-time monitoring module and a result analysis module, wherein each module is electrically connected. The data acquisition module is used to collect environmental values; The preprocessing module: improves data quality through data cleaning and normalization; The feature extraction module uses complex functions and deep learning models to extract features; The model building module includes a convolutional neural network, a recursive neural network and a long short-term memory network, and optimizes model parameters through a loop step; The multi-model fusion module: calculates the comprehensive score by weighted average or other fusion strategies; The alarm strategy module: adjusts the alarm strategy through an optimization algorithm to improve the accuracy of the alarm; The real-time monitoring module: monitors environmental data in real time and automatically adjusts monitoring parameters according to the alarm strategy; The result analysis module analyzes the monitoring results, generates a detailed monitoring report and provides optimization suggestions.
Citation Information
Patent Citations
Deep learning fusion model-based transient stability estimation method of power system
CN109921414A
Oil delivery pump rolling bearing state evaluation method based on convolutional neural network and long-term and short-term memory network
CN111767521A
Urination and defecation early warning system and method based on convolutional network and a long short-term memory network
CN113786205A
Mobile application sustainable trust evaluation method and device based on fusion model
CN115460097A
Fire safety monitoring system and method based on alarm area model
CN117095506A
Cited By
Intelligent analysis system based on artificial intelligence
CN120578945A
