Indoor Monitoring Data Security Early Warning Method Based on Deep Learning and Related Devices

Abstract

An embodiment of the present invention provides a method and related device for indoor monitoring data security warning based on deep learning. The method includes obtaining first monitoring data of a first LoRa sensor and second monitoring data of a second LoRa sensor, and determining first fusion data; determining a first target weight and a second target weight according to the first monitoring data, the second monitoring data and the first fusion data; obtaining third monitoring data of the first LoRa sensor and fourth monitoring data of the second LoRa sensor; extracting first feature information by performing feature extraction on the third monitoring data and extracting second feature information by performing feature extraction on the fourth monitoring data; performing feature fusion according to the first target weight, the first feature information, the second target weight and the second feature information to obtain second fusion data; performing anomaly recognition on the second fusion data by using a deep learning model to obtain an anomaly type; determining a security warning result according to the anomaly type, and performing a corresponding warning operation according to the security warning result.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular, to an indoor monitoring data security warning method based on deep learning and related devices. Background Art

[0002] With the wide application of the Internet of Things (IoT) in fields such as indoor monitoring, LoRa communication technology has become a key technology due to its characteristics such as low power consumption and long-distance transmission. However, the security of LoRa communication systems faces various threats, such as interference, spoofing, and replay attacks, which seriously affect the security and reliability of indoor monitoring systems. Traditional security protection methods have certain limitations in ensuring the security of LoRa systems. In the prior art, indoor monitoring systems based on LoRa communication mainly rely on traditional encryption and authentication means and cannot effectively cope with complex security threats and attacks. There is an urgent need for a method for quickly and accurately warning the security of indoor monitoring data. Summary of the Invention

[0003] The main purpose of the embodiments of the present invention is to provide an indoor monitoring data security warning method based on deep learning and related devices, aiming to solve the problem that indoor monitoring systems based on LoRa communication in the related art mainly rely on traditional encryption and authentication means and cannot effectively cope with complex security threats and attacks.

[0004] In a first aspect, the embodiments of the present invention provide an indoor monitoring data security warning method based on deep learning, including:

[0005] Obtain first monitoring data corresponding to a first LoRa sensor of a target object at a historical moment and second monitoring data corresponding to a second LoRa sensor of the target object at the historical moment, and determine first fusion data corresponding to the first LoRa sensor and the second LoRa sensor at the historical moment;

[0006] Determine a first target weight corresponding to the first LoRa sensor and a second target weight corresponding to the second LoRa sensor according to the first monitoring data and the second monitoring data in combination with the first fusion data;

[0007] Obtain third monitoring data corresponding to the first LoRa sensor of the target object at the current moment and fourth monitoring data corresponding to the second LoRa sensor of the target object at the current moment;

[0008] Extract first feature information by performing feature extraction on the third monitoring data and extract second feature information by performing feature extraction on the fourth monitoring data;

[0009] Feature fusion is performed based on the first target weight, the first feature information, the second target weight, and the second feature information to obtain second fusion data;

[0010] Based on the second fusion data, deep learning models are used for anomaly recognition to obtain anomaly types;

[0011] Based on the anomaly type, the safety warning result corresponding to the target object is determined, and corresponding warning operations are executed according to the safety warning result.

[0012] In a second aspect, an embodiment of the present invention provides an indoor monitoring data security warning device based on deep learning, including:

[0013] A data processing module, configured to obtain first monitoring data corresponding to a first LoRa sensor of a target object at a historical moment and second monitoring data corresponding to a second LoRa sensor of the target object at the historical moment, and determine first fusion data corresponding to the first LoRa sensor and the second LoRa sensor at the historical moment;

[0014] A weight determination module, configured to determine a first target weight corresponding to the first LoRa sensor and a second target weight corresponding to the second LoRa sensor according to the first monitoring data and the second monitoring data in combination with the first fusion data;

[0015] A data acquisition module, configured to obtain third monitoring data corresponding to the first LoRa sensor of the target object at the current moment and fourth monitoring data corresponding to the second LoRa sensor of the target object at the current moment;

[0016] A feature extraction module, configured to perform feature extraction on the third monitoring data to obtain first feature information and perform feature extraction on the fourth monitoring data to obtain second feature information;

[0017] A data fusion module, configured to perform feature fusion based on the first target weight, the first feature information, the second target weight, and the second feature information to obtain second fusion data;

[0018] An anomaly recognition module, configured to perform anomaly recognition using deep learning models based on the second fusion data to obtain anomaly types;

[0019] A warning processing module, configured to determine a safety warning result corresponding to the target object according to the anomaly type, and execute corresponding warning operations according to the safety warning result.

[0020] In a third aspect, an embodiment of the present invention further provides a terminal device, which includes a processor, a memory, a computer program stored on the memory and executable by the processor, and a data bus for implementing connection communication between the processor and the memory. When the computer program is executed by the processor, the steps of any one of the indoor monitoring data security warning methods based on deep learning provided in the specification of the present invention are implemented.

[0021] In a fourth aspect, an embodiment of the present invention further provides a storage medium for computer-readable storage, which is characterized in that the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any one of the indoor monitoring data security warning methods based on deep learning provided in the specification of the present invention.

[0022] An embodiment of the present invention provides a method and related device for indoor monitoring data security warning based on deep learning. The method includes: obtaining first monitoring data corresponding to a first LoRa sensor of a target object at a historical moment and second monitoring data corresponding to a second LoRa sensor of the target object at the historical moment, and determining first fusion data corresponding to the first LoRa sensor and the second LoRa sensor at the historical moment. Then, according to the first monitoring data, the second monitoring data and the first fusion data, determine a first target weight corresponding to the first LoRa sensor and a second target weight corresponding to the second LoRa sensor, so as to determine the contribution weights of different LoRa sensors to the target object, and further more accurately quantify the importance of different LoRa sensor data in security warning. Then, obtain third monitoring data corresponding to the first LoRa sensor of the target object at the current moment and fourth monitoring data corresponding to the second LoRa sensor of the target object at the current moment; extract first feature information from the third monitoring data and extract second feature information from the fourth monitoring data. Then, perform feature fusion according to the first target weight, the first feature information, the second target weight and the second feature information to obtain second fusion data, so as to be able to extract key information corresponding to the target object from the monitoring data. Then, use a deep learning model to perform anomaly recognition according to the second fusion data to obtain an anomaly type; finally, determine a security warning result corresponding to the target object according to the anomaly type, and execute corresponding warning operations according to the security warning result, so as to be able to detect unusual behavior patterns or security threats more quickly and accurately, and then take warning measures in time, and further be able to quickly respond to potential security risks and ensure the security of the indoor environment. And it solves the problem that the indoor monitoring system based on LoRa communication in the related technology mainly relies on traditional encryption and authentication means and cannot effectively cope with complex security threats and attacks. The present invention provides a LoRa communication data security warning method integrating advanced deep learning technology. Through enhanced security detection functions, it can effectively protect indoor monitoring data from security threats. It is not only applicable to general environmental monitoring, but also can be extended to fields such as cultural relic protection and ecological protection, and has broad application prospects and significant social and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained according to these drawings without creative efforts.

[0024] Figure 1 It is a schematic flow chart of a method for indoor monitoring data security warning based on deep learning provided by an embodiment of the present invention;

[0025] Figure 2 It is a schematic flow chart of a device for security warning of indoor monitoring data based on deep learning provided by an embodiment of the present invention;

[0026] Figure 3 It is a schematic structural diagram of a terminal device provided by an embodiment of the present invention. Detailed implementation manners

[0027] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0028] The flow chart shown in the accompanying drawings is only an example illustration, and does not necessarily include all contents and operations / steps, nor does it necessarily execute in the described order. For example, some operations / steps can be decomposed, combined or partially merged, so the actual execution order may change according to the actual situation.

[0029] It should be understood that the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless otherwise clearly specified in the context, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0030] An embodiment of the present invention provides a method for security warning of indoor monitoring data based on deep learning and related devices. Among them, the method for security warning of indoor monitoring data based on deep learning can be applied to a terminal device, and the terminal device can be an electronic device such as a tablet computer, a notebook computer, a desktop computer, a personal digital assistant, and a wearable device. The terminal device can be a server or a server cluster.

[0031] Next, some embodiments of the present invention will be described in detail in conjunction with the accompanying drawings. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0032] Please refer to Figure 1 , Figure 1 It is a schematic flow chart of a method for security warning of indoor monitoring data based on deep learning provided by an embodiment of the present invention.

[0033] As Figure 1 shown, the method for security warning of indoor monitoring data based on deep learning includes steps S101 to S107.

[0034] Step S101: Obtain the first monitoring data corresponding to the first LoRa sensor of the target object at a historical moment and the second monitoring data corresponding to the second LoRa sensor of the target object at the historical moment, and determine the first fusion data corresponding to the first LoRa sensor and the second LoRa sensor at the historical moment.

[0035] Exemplarily, the target object is an indoor environment to be monitored, which can be a scene room corresponding to cultural relic protection or an ecological environment corresponding to ecological protection, etc.

[0036] Exemplarily, multiple LoRa sensors are used to monitor the target object, and the first LoRa sensor or the second LoRa sensor is any one of the LoRa sensors such as a temperature sensor, a humidity sensor, a light intensity sensor, etc.

[0037] Exemplarily, obtain the first monitoring data corresponding to the first LoRa sensor and the second monitoring data corresponding to the second LoRa sensor at a historical moment from the database. And obtain the corresponding safety warning result at the historical moment from the database, and then infer the corresponding first fusion data at the historical moment according to the safety warning result. The first fusion data is used to represent that the safety warning result corresponding to the target object can be accurately obtained at the historical moment.

[0038] Step S102: Determine the first target weight corresponding to the first LoRa sensor and the second target weight corresponding to the second LoRa sensor according to the first monitoring data and the second monitoring data in combination with the first fusion data.

[0039] Exemplarily, by training a machine learning model (such as a regression model or a neural network), using the first monitoring data and the second monitoring data as inputs and the first fusion data as outputs to learn the mapping relationship of the LoRa sensor weights, so as to obtain the first target weight corresponding to the first LoRa sensor and the second target weight corresponding to the second LoRa sensor. Furthermore, the first target weight and the second target weight reflect the importance or contribution degree of different LoRa sensors in monitoring the target object.

[0040] In some embodiments, determining the first target weight corresponding to the first LoRa sensor and the second target weight corresponding to the second LoRa sensor based on the first monitoring data and the second monitoring data in combination with the first fusion data includes: determining a first initial weight corresponding to the first LoRa sensor and a second initial weight corresponding to the second LoRa sensor; fusing the first monitoring data and the second monitoring data according to the first initial weight and the second initial weight to obtain initial fusion data; obtaining a difference between the initial fusion data and the first fusion data, and when the difference is less than a preset value, determining a first accuracy corresponding to the first LoRa sensor according to the first monitoring data, and determining a second accuracy corresponding to the second LoRa sensor according to the second monitoring data; determining a third accuracy corresponding to the target object according to the first initial weight, the second initial weight, the first accuracy, and the second accuracy; when the third accuracy meets a preset condition, determining the first target weight and the second target weight according to the first accuracy and the second accuracy.

[0041] For example, based on prior knowledge, a first initial weight corresponding to the first LoRa sensor and a second initial weight corresponding to the second LoRa sensor are set. The first initial weight and the second initial weight can be evenly distributed, based on experience, or determined through preliminary data analysis.

[0042] Exemplarily, the first monitoring data of the first LoRa sensor and the second monitoring data of the second LoRa sensor are fused by weighted averaging in combination with the first initial weight and the second initial weight to obtain initial fused data, and then the difference between the initial fused data and the first fused data is compared.

[0043] Exemplarily, a preset value is determined. When the difference is less than the preset value, the initial fused data is considered sufficiently close to the first fused data, and a first accuracy corresponding to the first LoRa sensor is calculated based on the first monitoring data of the first LoRa sensor, and a second accuracy corresponding to the second LoRa sensor is determined based on the second monitoring data of the second LoRa sensor.

[0044] For example, a first variance corresponding to the first monitoring data and a second variance corresponding to the second monitoring data are calculated, and then the first variances are weighted and summed to obtain a first sum, and the second variances are weighted and summed to obtain a second sum. The first sum is then determined as the first accuracy, and the second sum is determined as the second accuracy.

[0045] Exemplarily, a first product is obtained by multiplying the first initial weight and the first accuracy, and a second product is obtained by multiplying the second initial weight and the second accuracy. Then, the first product and the second product are added to determine the third accuracy corresponding to the target object.

[0046] Exemplarily, it is evaluated whether the calculated third accuracy meets a preset condition. For example, a threshold of accuracy can be set to determine whether the required precision level is achieved. Then, when the third accuracy is less than the preset threshold, the reciprocal of the first accuracy is solved to obtain the first reciprocal, and the reciprocal of the second accuracy is solved to obtain the second reciprocal. Then, the sum of the first reciprocal and the second reciprocal is obtained to get the reciprocal sum value. Then, the first reciprocal is multiplied by the reciprocal sum value to obtain the third product, and the reciprocal value corresponding to the third product is determined as the first target weight. Also, the second reciprocal is multiplied by the reciprocal sum value to obtain the fourth product, and the reciprocal value corresponding to the fourth product is determined as the second target weight.

[0047] Specifically, by calculating the difference between the initial fusion data and the first fusion data and adjusting the accuracy of the LoRa sensor according to the difference, the possible errors of the LoRa sensor during the monitoring process can be dynamically corrected, improving the stability and reliability of the monitoring method. Thus, the accurate weights corresponding to the first LoRa sensor and the second LoRa sensor during data fusion can be effectively obtained, providing good support for subsequent safety warnings.

[0048] In some embodiments, the method further includes: when the difference is greater than or equal to the preset value, or the third accuracy does not meet the preset condition, the first initial weight corresponding to the first LoRa sensor is adjusted to obtain a third initial weight, and the second initial weight corresponding to the second LoRa sensor is adjusted to obtain a fourth initial weight, until the adjusted difference is less than the preset value, or the adjusted third accuracy meets the preset condition.

[0049] Exemplarily, the difference between the initial fusion data and the first fusion data is compared. According to the difference, it is judged whether it meets the preset value. If the difference is greater than or equal to the preset value, or the third accuracy does not meet the preset condition, the weight adjustment stage is entered. The first initial weight is adjusted to obtain a third initial weight, and the second initial weight is adjusted to obtain a fourth initial weight. Then, the above steps are repeatedly executed until the set preset value and preset condition are reached.

[0050] Specifically, by achieving the goal of dynamically adjusting the initial weights of the LoRa sensors, the fusion data and accuracy are ensured to meet the target requirements, thereby optimizing the performance and stability of the monitoring method and providing good support for subsequent safety warnings.

[0051] Step S103: Obtain the third monitoring data corresponding to the first LoRa sensor of the target object at the current moment and the fourth monitoring data corresponding to the second LoRa sensor of the target object at the current moment.

[0052] Exemplarily, the third monitoring data is obtained by monitoring the target object using the first LoRa sensor according to the current moment. Similarly, the fourth monitoring data is obtained by monitoring the target object using the second LoRa sensor according to the current moment.

[0053] Step S104: Extract features from the third monitoring data to obtain first feature information and extract features from the fourth monitoring data to obtain second feature information.

[0054] Exemplarily, obtain the first maximum value and the first minimum value corresponding to the third monitoring data, and then normalize the third monitoring data according to the first maximum value and the first minimum value to obtain the first processed data. And obtain the second maximum value and the second minimum value corresponding to the fourth monitoring data, and then normalize the fourth monitoring data according to the second maximum value and the second minimum value to obtain the second processed data.

[0055] For example, the first processed data is obtained according to the following formula:

[0056] ;

[0057] represents the first processed data, represents the third monitoring data, represents the first minimum value, represents the first maximum value.

[0058] Similarly, the second processed data is obtained according to the following formula:

[0059] ;

[0060] represents the second processed data, represents the fourth monitoring data, represents the second minimum value, represents the second maximum value.

[0061] Exemplarily, after obtaining the first processed data and the second processed data, decompose the first processed data and the second processed data into various frequency components according to the discrete wavelet transform (DWT), so as to isolate and remove noise to obtain the first denoised data corresponding to the first processed data and the second denoised data corresponding to the second processed data.

[0062] Exemplarily, a comprehensive feature set of time-domain and frequency-domain features is extracted from the first denoised data to obtain first feature information. And a comprehensive feature set of time-domain and frequency-domain features is extracted from the second denoised data to obtain second feature information.

[0063] In some embodiments, the obtaining of the first feature information by performing feature extraction on the third monitoring data and the obtaining of the second feature information by performing feature extraction on the fourth monitoring data include: performing filtering processing on the third monitoring data to obtain first filtered data and performing filtering processing on the fourth monitoring data to obtain second filtered data; determining a first feature sequence corresponding to the first filtered data and a second feature sequence corresponding to the second filtered data; setting a first abnormal frequency corresponding to the first LoRa sensor and a second abnormal frequency corresponding to the second LoRa sensor; obtaining a first value corresponding to the first abnormal frequency according to the first feature sequence and obtaining a second value corresponding to the second abnormal frequency according to the second feature sequence; determining the first feature information according to the first value and the first feature sequence, and determining the second feature information according to the second value and the second feature sequence.

[0064] Exemplarily, an appropriate filtering algorithm is selected according to the nature of the monitoring data, such as low-pass filtering, high-pass filtering or other digital filtering methods, to eliminate noise or unwanted signal components in the data. Thus, the selected filtering algorithm is applied to process the third monitoring data obtained by the first LoRa sensor to obtain first filtered data. Similarly, the same filtering algorithm is applied to the fourth monitoring data obtained by the second LoRa sensor to obtain second filtered data.

[0065] Exemplarily, a feature sequence is extracted from the filtered data, which may involve specific feature extraction methods such as time series analysis, spectrum analysis or other signal processing techniques, so as to perform feature extraction from the first filtered data and the second filtered data respectively to obtain a first feature sequence and a second feature sequence.

[0066] Exemplarily, a first abnormal frequency corresponding to the first LoRa sensor and a second abnormal frequency corresponding to the second LoRa sensor are obtained according to historical data. The first abnormal frequency and the second abnormal frequency are used to represent the frequencies of unusual events or specific abnormal patterns in different monitoring data.

[0067] Exemplarily, a first value corresponding to the first feature sequence at the first abnormal frequency is obtained and a second value corresponding to the second feature sequence at the second abnormal frequency is obtained. The first value and the second value are used to reflect the data characteristics corresponding to the current moment at the abnormal frequency.

[0068] Exemplarily, obtain a first ratio of a first numerical value in a first feature sequence, and then determine the first ratio as first feature information, and obtain a second ratio of a second numerical value in a second feature sequence, and then determine the second ratio as second feature information.

[0069] Specifically, the filtering process can effectively remove noise and unnecessary interference signals in the LoRa sensor data, thereby improving the quality and reliability of the data. By extracting the feature sequence from the filtered data, the key feature information of the target object can be captured more accurately, which helps subsequent data analysis and decision-making. Furthermore, by setting the first abnormal frequency and the second abnormal frequency of the first LoRa sensor and the second LoRa sensor, and determining the numerical values according to the feature sequence and the abnormal frequency, real-time detection and early warning of abnormal events or important data features can be achieved, thereby obtaining the first feature information and the second feature information, and then the corresponding characterization features under different LoRa sensors can be effectively obtained, which provides good support for subsequent safety early warning.

[0070] In some embodiments, the obtaining the first numerical value corresponding to the first abnormal frequency according to the first feature sequence and the obtaining the second numerical value corresponding to the second abnormal frequency according to the second feature sequence include: determining a first quantity corresponding to the first abnormal frequency and determining a second quantity corresponding to the second abnormal frequency; obtaining a first integer sequence corresponding to the first quantity and obtaining a second integer sequence corresponding to the second quantity; multiplying each first integer in the first integer sequence by the first abnormal frequency to obtain a first frequency, and multiplying each second integer in the second integer sequence by the second abnormal frequency to obtain a second frequency; obtaining the first numerical value corresponding to the first frequency according to the first feature sequence and obtaining the second numerical value corresponding to the second frequency according to the second feature sequence.

[0071] Exemplarily, determine a first quantity required to be collected at the first abnormal frequency and a second quantity required to be collected at the second abnormal frequency, and then determine the integers from 1 to the first quantity as the first integer sequence and the integers from 1 to the second quantity as the second integer sequence.

[0072] Exemplarily, multiply each integer in the first integer sequence by the first abnormal frequency to obtain a first frequency, and multiply each integer in the second integer sequence by the second abnormal frequency to obtain a second frequency.

[0073] Exemplarily, use the existing first feature sequence to obtain the first numerical value corresponding to the first frequency, and use the existing second feature sequence to obtain the second numerical value corresponding to the second frequency.

[0074] Specifically, by accurately determining the specific values corresponding to the first abnormal frequency and the second abnormal frequency, abnormal events or important data features in the monitoring data can be precisely captured, providing an important basis for subsequent analysis and decision-making. Thus, obtaining the integer sequences corresponding to the first quantity and the second quantity is helpful for further calculation and processing. Furthermore, by multiplying each integer in the integer sequence by the abnormal frequency, the first frequency and the second frequency can be calculated. Thus, according to the first feature sequence and the second feature sequence, the specific values corresponding to the first frequency and the second frequency are determined. These values may be used to evaluate data security and further provide detailed data support for subsequent decision-making.

[0075] In some embodiments, the determining the first feature information according to the first value and the first feature sequence, and the determining the second feature information according to the second value and the second feature sequence include: obtaining a first sum value by summing the first values, and obtaining a second sum value by summing the second values; obtaining a third sum value by summing the first feature sequence, and obtaining a fourth sum value by summing the second feature sequence; determining the first feature information according to the first sum value, the third sum value, and the first quantity; determining the second feature information according to the second sum value, the fourth sum value, and the second quantity; wherein, the first feature information and the second feature information are obtained according to the following formula:

[0076] ;

[0077] ;

[0078] wherein, represents the first feature information, represents the second feature information, X1 represents the first quantity, X2 represents the second quantity, f1 represents the first abnormal frequency, f2 represents the second abnormal frequency, i* 1 represents the first frequency, i* 2 represents the second frequency, represents the first value, represents the second value, represents the value corresponding to the j-th position in the first feature sequence, represents the value corresponding to the j-th position in the second feature sequence, represents the quantity corresponding to the first feature sequence, represents the quantity corresponding to the second feature sequence.

[0079] Exemplarily, summing all the first values to obtain a first sum value, and summing all the second values to obtain a second sum value.

[0080] Exemplarily, the sum of all features in the first feature sequence is calculated to obtain a third sum value, and the sum of all features in the second feature sequence is calculated to obtain a fourth sum value.

[0081] Exemplarily, the first feature information is determined according to the first sum value, the third sum value, and the first quantity using the following formula:

[0082] ;

[0083] where represents the first feature information, X1 represents the first quantity, f1 represents the first abnormal frequency, i* 1 represents the first frequency, represents the first numerical value, represents the numerical value corresponding to the j-th position in the first feature sequence, represents the quantity corresponding to the first feature sequence.

[0084] Exemplarily, according to the above formula, the prominence degree of the feature value corresponding to the first abnormal frequency in the first feature sequence is measured, so as to obtain the first feature information corresponding to the first LoRa sensor.

[0085] Exemplarily, the second feature information is determined according to the second sum value, the fourth sum value, and the second quantity using the following formula:

[0086] ;

[0087] where represents the second feature information, X2 represents the second quantity, f2 represents the second abnormal frequency, i* 2 represents the second frequency, represents the second numerical value, represents the numerical value corresponding to the j-th position in the second feature sequence, represents the quantity corresponding to the second feature sequence.

[0088] Exemplarily, according to the above formula, the prominence degree of the feature value corresponding to the second abnormal frequency in the second feature sequence is measured, so as to obtain the second feature information corresponding to the second LoRa sensor.

[0089] Specifically, obtaining the first feature information by measuring the prominence degree of the feature value corresponding to the first abnormal frequency in the first feature sequence, and obtaining the second feature information by measuring the prominence degree of the feature value corresponding to the second abnormal frequency in the second feature sequence provide good support for subsequent security warnings.

[0090] Step S105: Perform feature fusion based on the first target weight, the first feature information, the second target weight, and the second feature information to obtain second fusion data.

[0091] Exemplarily, the first target weight and the first feature information, as well as the second target weight and the second feature information, are combined using weighted averaging to obtain the second fusion data.

[0092] Step S106: Use a deep learning model to perform anomaly recognition based on the second fusion data to obtain the anomaly type.

[0093] Exemplarily, according to specific anomaly recognition requirements, a suitable deep learning model is selected, such as a convolutional neural network (CNN), a recurrent neural network (RNN), or a Transformer model, etc. If the existing model is not suitable, model adjustment or retraining can be considered to meet specific data and task requirements, and then the trained deep learning model is used to perform anomaly recognition on the second fusion data. The result output by the model can be the probability of an anomaly or a direct anomaly type classification result.

[0094] In some embodiments, the deep learning model can adopt the LoRaSec deep learning framework, combine CNN, RNN, and an attention mechanism to perform anomaly recognition on the second fusion data to obtain the anomaly type. Exemplarily, the deep learning model includes a feature extraction layer, a feature attention layer, and an autoencoder network layer. The performing anomaly recognition on the second fusion data using the deep learning model to obtain the anomaly type includes: using the feature extraction layer to perform feature extraction on the third monitoring data to obtain third feature information and using the feature extraction layer to perform feature extraction on the fourth monitoring data to obtain fourth feature information; using the feature attention layer to perform feature fusion on the third feature information, the fourth feature information, and the second fusion data to obtain target fusion data; using the autoencoder network layer to perform anomaly recognition on the target fusion data to obtain the anomaly type (for example, the anomaly type includes interference, spoofing, or replay attack).

[0095] Exemplarily, for the third monitoring data and the fourth monitoring data, a feature extraction layer is designed and implemented. This may include using a convolutional neural network (CNN), a recurrent neural network (RNN), or other feature extraction models suitable for the task to obtain the third feature information corresponding to the third monitoring data and the fourth feature information corresponding to the fourth monitoring data.

[0096] Exemplarily, a layer that can dynamically focus on different input features is designed through an attention mechanism or other weighting strategies, which combines the third feature information, the fourth feature information, and the second fusion data to generate target fusion data, so as to ensure that the target fusion data can make full use of the input information and the prominence corresponding to the abnormal frequency.

[0097] Exemplarily, an autoencoder network layer is implemented for anomaly recognition of the target fusion data. The autoencoder network can learn the stable representation of the data through unsupervised learning and identify the anomaly types that do not conform to the normal data.

[0098] Specifically, the third feature information and the fourth feature information are obtained by extracting features from the third monitoring data and the fourth monitoring data, then feature fusion is performed using the attention mechanism, and then an autoencoder network is used for anomaly recognition and classification. Such a process can effectively process complex monitoring data, achieve efficient anomaly detection and type recognition, thereby improving the intelligence and response ability of the monitoring method.

[0099] Step S107: Determine the safety warning result corresponding to the target object according to the anomaly type, and perform corresponding warning operations according to the safety warning result.

[0100] Exemplarily, a mapping table or rule set from anomaly types to safety warning results is established. These mappings can be predefined. For example, different types of anomalies are mapped to different levels of safety warnings (such as low, medium, high), or directly mapped to specific warning operations (such as sending an alarm, automatically stopping an operation, etc.). Furthermore, corresponding warning rules and operations are defined according to each safety warning result. These rules can include triggering an alarm to notify relevant personnel, recording anomaly events, performing automated emergency operations, etc. Ensure that each warning result has a clear response measure.

[0101] Exemplarily, the safety warning result corresponding to the target object is determined from the mapping table according to the anomaly type, and then the corresponding warning operation is performed according to the safety warning result, so as to effectively protect the safety of the target object and reduce potential risks and losses.

[0102] An embodiment of the present invention provides a method and related device for indoor monitoring data security warning based on deep learning. The method includes: obtaining first monitoring data corresponding to a first LoRa sensor of a target object at a historical moment and second monitoring data corresponding to a second LoRa sensor of the target object at the historical moment, and determining first fusion data corresponding to the first LoRa sensor and the second LoRa sensor at the historical moment. Then, according to the first monitoring data, the second monitoring data, and the first fusion data, determine a first target weight corresponding to the first LoRa sensor and a second target weight corresponding to the second LoRa sensor, so as to determine the contribution weights of different LoRa sensors to the target object, and further more accurately quantify the importance of different LoRa sensor data in security warning. Furthermore, obtain third monitoring data corresponding to the first LoRa sensor of the target object at the current moment and fourth monitoring data corresponding to the second LoRa sensor of the target object at the current moment; then extract first feature information by performing feature extraction on the third monitoring data and extract second feature information by performing feature extraction on the fourth monitoring data. Then, perform feature fusion according to the first target weight, the first feature information, the second target weight, and the second feature information to obtain second fusion data, so as to be able to extract key information corresponding to the target object from the monitoring data. Furthermore, use a deep learning model to perform anomaly recognition according to the second fusion data to obtain an anomaly type; finally, determine a security warning result corresponding to the target object according to the anomaly type, and perform corresponding warning operations according to the security warning result, so as to be able to detect unusual behavior patterns or security threats more quickly and accurately, and thus take warning measures in a timely manner, and further be able to quickly respond to potential security risks and ensure the security of the indoor environment. And it solves the problem that in the related art, the indoor monitoring system based on LoRa communication mainly relies on traditional encryption and authentication means and cannot effectively cope with complex security threats and attacks. The present invention provides a LoRa communication data security warning method integrating advanced deep learning technology. Through enhanced security detection functions, it can effectively protect indoor monitoring data from being affected by security threats. It is not only applicable to general environmental monitoring, but also can be extended to fields such as cultural relic protection and ecological protection, and has broad application prospects and significant social and economic benefits.

[0103] Please refer to Figure 2 , Figure 2An indoor monitoring data security early warning device 200 provided by an embodiment of the present application. The indoor monitoring data security early warning device 200 based on deep learning includes a data processing module 201, a weight determination module 202, a data acquisition module 203, a feature extraction module 204, a data fusion module 205, an anomaly recognition module 206, and an early warning processing module 207. Among them, the data processing module 201 is used to obtain first monitoring data corresponding to a first LoRa sensor of a target object at a historical moment and second monitoring data corresponding to a second LoRa sensor of the target object at the historical moment, and determine first fusion data corresponding to the first LoRa sensor and the second LoRa sensor at the historical moment; the weight determination module 202 is used to determine a first target weight corresponding to the first LoRa sensor and a second target weight corresponding to the second LoRa sensor according to the first monitoring data and the second monitoring data in combination with the first fusion data; the data acquisition module 203 is used to obtain third monitoring data corresponding to the first LoRa sensor of the target object at the current moment and fourth monitoring data corresponding to the second LoRa sensor of the target object at the current moment; the feature extraction module 204 is used to perform feature extraction on the third monitoring data to obtain first feature information and perform feature extraction on the fourth monitoring data to obtain second feature information; the data fusion module 205 is used to perform feature fusion according to the first target weight, the first feature information, the second target weight, and the second feature information to obtain second fusion data; the anomaly recognition module 206 is used to perform anomaly recognition using a deep learning model according to the second fusion data to obtain an anomaly type; the early warning processing module 207 is used to determine a security early warning result corresponding to the target object according to the anomaly type, and perform a corresponding early warning operation according to the security early warning result.

[0104] In some embodiments, during the process of determining the first target weight corresponding to the first LoRa sensor and the second target weight corresponding to the second LoRa sensor according to the first monitoring data and the second monitoring data in combination with the first fusion data, the weight determination module 202 performs:

[0105] Determine a first initial weight corresponding to the first LoRa sensor and a second initial weight corresponding to the second LoRa sensor;

[0106] Fuse the first monitoring data and the second monitoring data according to the first initial weight and the second initial weight to obtain initial fusion data;

[0107] Obtain the difference between the obtained initial fusion data and the first fusion data. When the difference is less than a preset value, determine the first accuracy corresponding to the first LoRa sensor according to the first monitoring data, and determine the second accuracy corresponding to the second LoRa sensor according to the second monitoring data;

[0108] Determine the third accuracy corresponding to the target object according to the first initial weight, the second initial weight, the first accuracy, and the second accuracy;

[0109] When the third accuracy meets a preset condition, determine the first target weight and the second target weight according to the first accuracy and the second accuracy.

[0110] In some embodiments, the weight determination module 202 also performs:

[0111] When the difference is greater than or equal to the preset value, or the third accuracy does not meet the preset condition, adjust the first initial weight corresponding to the first LoRa sensor to obtain a third initial weight and adjust the second initial weight corresponding to the second LoRa sensor to obtain a fourth initial weight until the adjusted difference is less than the preset value, or the adjusted third accuracy meets the preset condition.

[0112] In some embodiments, during the process of extracting first feature information from the third monitoring data and extracting second feature information from the fourth monitoring data, the feature extraction module 204 performs:

[0113] Perform filtering processing on the third monitoring data to obtain first filtered data and perform filtering processing on the fourth monitoring data to obtain second filtered data;

[0114] Determine a first feature sequence corresponding to the first filtered data and a second feature sequence corresponding to the second filtered data;

[0115] Set a first abnormal frequency corresponding to the first LoRa sensor and a second abnormal frequency corresponding to the second LoRa sensor;

[0116] Obtain a first value corresponding to the first abnormal frequency according to the first feature sequence and obtain a second value corresponding to the second abnormal frequency according to the second feature sequence;

[0117] Determine the first feature information according to the first value and the first feature sequence, and determine the second feature information according to the second value and the second feature sequence.

[0118] In some embodiments, during the process of obtaining the first value corresponding to the first abnormal frequency according to the first feature sequence and obtaining the second value corresponding to the second abnormal frequency according to the second feature sequence, the feature extraction module 204 performs:

[0119] Determine the first quantity corresponding to the first abnormal frequency and determine the second quantity corresponding to the second abnormal frequency;

[0120] Obtain the first integer sequence corresponding to the first quantity and obtain the second integer sequence corresponding to the second quantity;

[0121] Multiply each first integer in the first integer sequence by the first abnormal frequency to obtain a first frequency, and multiply each second integer in the second integer sequence by the second abnormal frequency to obtain a second frequency;

[0122] Obtain the first value corresponding to the first frequency according to the first feature sequence and obtain the second value corresponding to the second frequency according to the second feature sequence.

[0123] In some embodiments, during the process of determining the first feature information according to the first value and the first feature sequence, and determining the second feature information according to the second value and the second feature sequence, the feature extraction module 204 performs:

[0124] Sum the first values to obtain a first sum value, and sum the second values to obtain a second sum value;

[0125] Sum the first feature sequence to obtain a third sum value, and sum the second feature sequence to obtain a fourth sum value;

[0126] Determine the first feature information according to the first sum value, the third sum value, and the first quantity;

[0127] Determine the second feature information according to the second sum value, the fourth sum value, and the second quantity;

[0128] Wherein, the first feature information and the second feature information are obtained according to the following formula:

[0129] ;

[0130] ;

[0131] Wherein, represents the first feature information, represents the second feature information, X1 represents the first quantity, X2 represents the second quantity, f1 represents the first abnormal frequency, f2 represents the second abnormal frequency, i* 1 represents the first frequency, i* 2 represents the second frequency, represents the first numerical value, represents the second numerical value, represents the numerical value corresponding to the j-th bit in the first feature sequence, represents the numerical value corresponding to the j-th bit in the second feature sequence, represents the quantity corresponding to the first feature sequence, represents the quantity corresponding to the second feature sequence.

[0132] In some embodiments, the deep learning model includes a feature extraction layer, a feature attention layer, and an autoencoder network layer. When the anomaly recognition module 206 performs anomaly recognition using the deep learning model based on the second fusion data to obtain the anomaly type, it executes:

[0133] Use the feature extraction layer to extract features from the third monitoring data to obtain third feature information and use the feature extraction layer to extract features from the fourth monitoring data to obtain fourth feature information;

[0134] Use the feature attention layer to perform feature fusion on the third feature information, the fourth feature information, and the second fusion data to obtain target fusion data;

[0135] Use the autoencoder network layer to perform anomaly recognition on the target fusion data to obtain the anomaly type.

[0136] In some embodiments, the indoor monitoring data security early warning device 200 based on deep learning can be applied to a terminal device.

[0137] It should be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the above-described indoor monitoring data security early warning device 200 based on deep learning can refer to the corresponding process in the foregoing embodiment of the indoor monitoring data security early warning method based on deep learning, and will not be elaborated here.

[0138] Please refer to Figure 3 , Figure 3 which is a schematic block diagram of the structure of a terminal device provided by an embodiment of the present invention.

[0139] As Figure 3As shown, the terminal device 300 includes a processor 301 and a memory 302. The processor 301 and the memory 302 are connected through a bus 303, which is, for example, an I2C (Inter - integrated Circuit) bus.

[0140] Specifically, the processor 301 is used to provide computing and control capabilities to support the operation of the entire terminal device. The processor 301 can be a central processing unit (CPU), or it can also be other general - purpose processors, digital signal processors (DSPs), application - specific integrated circuits (ASICs), field - programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general - purpose processor can be a microprocessor or any conventional processor, etc.

[0141] Specifically, the memory 302 can be a Flash chip, read - only memory (ROM), magnetic disk, optical disc, USB flash drive, or mobile hard disk, etc.

[0142] Those skilled in the art can understand that Figure 3 the structure shown in is only a block diagram of some structures related to the solution of the embodiment of the present invention, and does not constitute a limitation on the terminal device to which the solution of the embodiment of the present invention is applied. A specific server may include more or fewer components than those shown in the figure, or combine some components, or have a different component layout.

[0143] The processor is used to run a computer program stored in the memory and, when executing the computer program, implement any one of the indoor monitoring data security warning methods based on deep learning provided by the embodiments of the present invention.

[0144] In one embodiment, the processor is used to run a computer program stored in the memory and, when executing the computer program, implement the following steps:

[0145] Obtain the first monitoring data corresponding to the first LoRa sensor of the target object at a historical moment and the second monitoring data corresponding to the second LoRa sensor of the target object at the historical moment, and determine the first fusion data corresponding to the first LoRa sensor and the second LoRa sensor at the historical moment;

[0146] Determine the first target weight corresponding to the first LoRa sensor and the second target weight corresponding to the second LoRa sensor based on the first monitoring data and the second monitoring data in combination with the first fusion data;

[0147] Obtain the third monitoring data corresponding to the first LoRa sensor of the target object at the current moment and the fourth monitoring data corresponding to the second LoRa sensor of the target object at the current moment;

[0148] Extract features from the third monitoring data to obtain first feature information and extract features from the fourth monitoring data to obtain second feature information;

[0149] Perform feature fusion based on the first target weight, the first feature information, the second target weight, and the second feature information to obtain second fusion data;

[0150] Perform anomaly recognition on the second fusion data using a deep learning model to obtain the anomaly type;

[0151] Determine the safety warning result corresponding to the target object according to the anomaly type, and perform corresponding warning operations according to the safety warning result.

[0152] In some embodiments, when the processor 301 determines the first target weight corresponding to the first LoRa sensor and the second target weight corresponding to the second LoRa sensor based on the first monitoring data and the second monitoring data in combination with the first fusion data, it performs:

[0153] Determine the first initial weight corresponding to the first LoRa sensor and the second initial weight corresponding to the second LoRa sensor;

[0154] Fuse the first monitoring data and the second monitoring data according to the first initial weight and the second initial weight to obtain initial fusion data;

[0155] Obtain the difference between the initial fusion data and the first fusion data. When the difference is less than a preset value, determine the first accuracy corresponding to the first LoRa sensor according to the first monitoring data, and determine the second accuracy corresponding to the second LoRa sensor according to the second monitoring data;

[0156] Determine the third accuracy corresponding to the target object according to the first initial weight, the second initial weight, the first accuracy, and the second accuracy;

[0157] When the third accuracy meets the preset condition, the first target weight and the second target weight are determined according to the first accuracy and the second accuracy.

[0158] In some embodiments, the processor 301 further executes:

[0159] When the difference is greater than or equal to the preset value, or the third accuracy does not meet the preset condition, the first initial weight corresponding to the first LoRa sensor is adjusted to obtain a third initial weight, and the second initial weight corresponding to the second LoRa sensor is adjusted to obtain a fourth initial weight, until the adjusted difference is less than the preset value, or the adjusted third accuracy meets the preset condition.

[0160] In some embodiments, during the process of extracting first feature information from the third monitoring data and extracting second feature information from the fourth monitoring data, the processor 301 executes:

[0161] Perform filtering processing on the third monitoring data to obtain first filtered data and perform filtering processing on the fourth monitoring data to obtain second filtered data;

[0162] Determine a first feature sequence corresponding to the first filtered data and a second feature sequence corresponding to the second filtered data;

[0163] Set a first abnormal frequency corresponding to the first LoRa sensor and a second abnormal frequency corresponding to the second LoRa sensor;

[0164] Obtain a first value corresponding to the first abnormal frequency according to the first feature sequence and obtain a second value corresponding to the second abnormal frequency according to the second feature sequence;

[0165] Determine the first feature information according to the first value and the first feature sequence, and determine the second feature information according to the second value and the second feature sequence.

[0166] In some embodiments, during the process of obtaining a first value corresponding to the first abnormal frequency according to the first feature sequence and obtaining a second value corresponding to the second abnormal frequency according to the second feature sequence, the processor 301 executes:

[0167] Determine a first quantity corresponding to the first abnormal frequency and determine a second quantity corresponding to the second abnormal frequency;

[0168] Obtain a first integer sequence corresponding to the first quantity and obtain a second integer sequence corresponding to the second quantity;

[0169] Multiply each first integer in the first integer sequence by the first abnormal frequency to obtain a first frequency, and multiply each second integer in the second integer sequence by the second abnormal frequency to obtain a second frequency;

[0170] Obtain the first value corresponding to the first frequency according to the first feature sequence and obtain the second value corresponding to the second frequency according to the second feature sequence.

[0171] In some embodiments, when the processor 301 determines the first feature information according to the first value and the first feature sequence and determines the second feature information according to the second value and the second feature sequence, it performs:

[0172] Sum the first values to obtain a first sum value, and sum the second values to obtain a second sum value;

[0173] Sum the first feature sequence to obtain a third sum value, and sum the second feature sequence to obtain a fourth sum value;

[0174] Determine the first feature information according to the first sum value, the third sum value, and the first quantity;

[0175] Determine the second feature information according to the second sum value, the fourth sum value, and the second quantity;

[0176] Wherein, the first feature information and the second feature information are obtained according to the following formula:

[0177] ;

[0178] ;

[0179] Wherein, represents the first feature information, represents the second feature information, X1 represents the first quantity, X2 represents the second quantity, f1 represents the first abnormal frequency, f2 represents the second abnormal frequency, i* 1 represents the first frequency, i* 2 represents the second frequency, represents the first value, represents the second value, represents the value corresponding to the j-th bit in the first feature sequence, represents the value corresponding to the j-th bit in the second feature sequence, represents the quantity corresponding to the first feature sequence, represents the quantity corresponding to the second feature sequence.

[0180] In some embodiments, the deep learning model includes a feature extraction layer, a feature attention layer, and an autoencoder network layer. When the processor 301 uses the deep learning model to perform anomaly recognition based on the second fusion data to obtain the anomaly type, the following operations are executed:

[0181] Using the feature extraction layer to extract features from the third monitoring data to obtain third feature information and using the feature extraction layer to extract features from the fourth monitoring data to obtain fourth feature information;

[0182] Using the feature attention layer to perform feature fusion on the third feature information, the fourth feature information, and the second fusion data to obtain target fusion data;

[0183] Using the autoencoder network layer to perform anomaly recognition on the target fusion data to obtain the anomaly type.

[0184] It should be noted that those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working process of the above-described terminal device can refer to the corresponding process in the embodiment of the indoor monitoring data security warning method based on deep learning described above, and will not be repeated here.

[0185] The embodiment of the present invention further provides a storage medium for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any one of the indoor monitoring data security warning methods based on deep learning provided in the specification of the embodiment of the present invention.

[0186] Among them, the storage medium can be the internal storage unit of the terminal device in the foregoing embodiment, such as the hard disk or memory of the terminal device. The storage medium can also be an external storage device of the terminal device, such as a plug-in hard disk equipped on the terminal device, a smart media card (SMC), a secure digital (SD) card, a flash card, etc.

[0187] Those of ordinary skill in the art will understand that all or some of the steps in the methods disclosed above, and the functional modules / units in systems and devices, can be implemented as software, firmware, hardware, and appropriate combinations thereof. In a hardware embodiment, the division of functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be executed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory, or other memory technologies, CD-ROM, digital versatile disks (DVDs), or other optical disk storage, magnetic cassettes, tapes, magnetic disk storage, or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that a communication medium typically contains computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and may include any information delivery medium.

[0188] It should be understood that the term "and / or" used in the specification and appended claims of the present invention refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations. It should be noted that in this document, the terms "include", "comprise", or any other variation thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that includes a series of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article, or system. Without further limitation, an element defined by the statement "including one..." does not exclude the presence of additional identical elements in the process, method, article, or system that includes the element.

[0189] The serial numbers of the embodiments of the present invention above are only for description and do not represent the superiority or inferiority of the embodiments. The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. An indoor monitoring data security warning method based on deep learning, characterized in that, The method includes: Obtaining first monitoring data corresponding to a first LoRa sensor of a target object at a historical moment and second monitoring data corresponding to a second LoRa sensor of the target object at the historical moment, and determining first fusion data corresponding to the first LoRa sensor and the second LoRa sensor at the historical moment; wherein, obtaining a safety warning result corresponding to the historical moment from a database, and then inferring the first fusion data corresponding to the historical moment according to the safety warning result, and the first fusion data is used to represent that the safety warning result corresponding to the target object can be accurately obtained at the historical moment; Determining a first target weight corresponding to the first LoRa sensor and a second target weight corresponding to the second LoRa sensor according to the first monitoring data, the second monitoring data and the first fusion data; Obtaining third monitoring data corresponding to the first LoRa sensor of the target object at the current moment and fourth monitoring data corresponding to the second LoRa sensor of the target object at the current moment; Performing feature extraction on the third monitoring data to obtain first feature information and performing feature extraction on the fourth monitoring data to obtain second feature information; Performing feature fusion on the first target weight, the first feature information, the second target weight and the second feature information to obtain second fusion data; Performing anomaly recognition on the second fusion data by using a deep learning model to obtain an anomaly type; Determining a safety warning result corresponding to the target object according to the anomaly type, and performing a corresponding warning operation according to the safety warning result; Wherein, the determining the first target weight corresponding to the first LoRa sensor and the second target weight corresponding to the second LoRa sensor according to the first monitoring data, the second monitoring data and the first fusion data includes: Determining a first initial weight corresponding to the first LoRa sensor and a second initial weight corresponding to the second LoRa sensor; Fusing the first monitoring data and the second monitoring data according to the first initial weight and the second initial weight to obtain initial fusion data; Obtaining a difference between the initial fusion data and the first fusion data. When the difference is less than a preset value, determining a first accuracy corresponding to the first LoRa sensor according to the first monitoring data, and determining a second accuracy corresponding to the second LoRa sensor according to the second monitoring data; wherein, calculating a first variance corresponding to the first monitoring data and a second variance corresponding to the second monitoring data, then performing weighted summation on the first variance to obtain a first sum, and performing weighted summation on the second variance to obtain a second sum; and then determining the first sum as the first accuracy and determining the second sum as the second accuracy; Determine the third accuracy corresponding to the target object according to the first initial weight, the second initial weight, the first accuracy, and the second accuracy; wherein, multiply the first initial weight by the first accuracy to obtain a first product, and multiply the second initial weight by the second accuracy to obtain a second product, and then add the first product and the second product to determine the third accuracy corresponding to the target object; When the third accuracy meets a preset condition, determine the first target weight and the second target weight according to the first accuracy and the second accuracy; wherein, when the third accuracy is less than a preset threshold, solve the reciprocal of the first accuracy to obtain a first reciprocal, and solve the reciprocal of the second accuracy to obtain a second reciprocal, then sum the first reciprocal and the second reciprocal to obtain a reciprocal sum value, and then multiply the first reciprocal by the reciprocal sum value to obtain a third product, so as to determine the reciprocal value corresponding to the third product as the first target weight, and multiply the second reciprocal by the reciprocal sum value to obtain a fourth product, so as to determine the reciprocal value corresponding to the fourth product as the second target weight.

2. The method according to claim 1, characterized in that, The method further includes: When the difference is greater than or equal to the preset value, or the third accuracy does not meet the preset condition, adjust the first initial weight corresponding to the first LoRa sensor to obtain a third initial weight and adjust the second initial weight corresponding to the second LoRa sensor to obtain a fourth initial weight until the adjusted difference is less than the preset value, or the adjusted third accuracy meets the preset condition.

3. The method according to claim 1, wherein The obtaining the first feature information by performing feature extraction on the third monitoring data and the obtaining the second feature information by performing feature extraction on the fourth monitoring data include: Perform filtering processing on the third monitoring data to obtain first filtered data and perform filtering processing on the fourth monitoring data to obtain second filtered data; Determine a first feature sequence corresponding to the first filtered data and a second feature sequence corresponding to the second filtered data; Set a first abnormal frequency corresponding to the first LoRa sensor and a second abnormal frequency corresponding to the second LoRa sensor; Obtain a first value corresponding to the first abnormal frequency according to the first feature sequence and obtain a second value corresponding to the second abnormal frequency according to the second feature sequence; Determine the first feature information according to the first value and the first feature sequence, and determine the second feature information according to the second value and the second feature sequence.

4. The method according to claim 3, wherein The obtaining the first value corresponding to the first abnormal frequency according to the first feature sequence and the obtaining the second value corresponding to the second abnormal frequency according to the second feature sequence include: Determine a first quantity corresponding to the first abnormal frequency and determine a second quantity corresponding to the second abnormal frequency; Obtain a first integer sequence corresponding to the first quantity and obtain a second integer sequence corresponding to the second quantity; Multiply each first integer in the first integer sequence by the first abnormal frequency to obtain a first frequency, and multiply each second integer in the second integer sequence by the second abnormal frequency to obtain a second frequency; Obtain the first value corresponding to the first frequency according to the first feature sequence and obtain the second value corresponding to the second frequency according to the second feature sequence.

5. The method according to claim 4, characterized in that, The determining the first feature information according to the first value and the first feature sequence, and determining the second feature information according to the second value and the second feature sequence includes: Sum the first values to obtain a first sum value, and sum the second values to obtain a second sum value; Sum the first feature sequence to obtain a third sum value, and sum the second feature sequence to obtain a fourth sum value; Determine the first feature information according to the first sum value, the third sum value and the first quantity; Determine the second feature information according to the second sum value, the fourth sum value and the second quantity; Wherein, the first feature information and the second feature information are obtained according to the following formula: ; ; Among them, represents the first feature information, represents the second feature information, X1 represents the first quantity, X2 represents the second quantity, f1 represents the first abnormal frequency, f2 represents the second abnormal frequency, i* 1 represents the first frequency, i* 2 represents the second frequency, represents the first numerical value, represents the second numerical value, represents the numerical value corresponding to the j-th position in the first feature sequence, represents the numerical value corresponding to the j-th position in the second feature sequence, represents the quantity corresponding to the first feature sequence, represents the quantity corresponding to the second feature sequence.

6. The method according to claim 1, wherein The deep learning model includes a feature extraction layer, a feature attention layer and an autoencoder network layer. The using the deep learning model to perform anomaly recognition on the second fusion data to obtain an anomaly type includes: Use the feature extraction layer to extract features from the third monitoring data to obtain third feature information and use the feature extraction layer to extract features from the fourth monitoring data to obtain fourth feature information; Use the feature attention layer to perform feature fusion on the third feature information, the fourth feature information and the second fusion data to obtain target fusion data; Use the autoencoder network layer to perform anomaly recognition on the target fusion data to obtain the anomaly type.

7. An indoor monitoring data security warning device based on deep learning, characterized in that, Including: A data processing module, configured to obtain first monitoring data corresponding to a first LoRa sensor of a target object at a historical moment and second monitoring data corresponding to a second LoRa sensor of the target object at the historical moment, and determine first fusion data corresponding to the first LoRa sensor and the second LoRa sensor at the historical moment; wherein, obtain a safety warning result corresponding to the historical moment from a database, and then infer the first fusion data corresponding to the historical moment according to the safety warning result, and the first fusion data is used to represent that the safety warning result corresponding to the target object can be accurately obtained at the historical moment; A weight determination module, configured to determine a first target weight corresponding to the first LoRa sensor and a second target weight corresponding to the second LoRa sensor according to the first monitoring data and the second monitoring data in combination with the first fusion data; A data acquisition module, configured to obtain third monitoring data corresponding to the first LoRa sensor of the target object at the current moment and fourth monitoring data corresponding to the second LoRa sensor of the target object at the current moment; A feature extraction module, configured to extract features from the third monitoring data to obtain first feature information and extract features from the fourth monitoring data to obtain second feature information; A data fusion module, configured to perform feature fusion based on the first target weight, the first feature information, the second target weight, and the second feature information to obtain second fusion data; An anomaly recognition module, configured to use a deep learning model to perform anomaly recognition based on the second fusion data to obtain an anomaly type; An early warning processing module, configured to determine a safety early warning result corresponding to the target object according to the anomaly type, and perform corresponding early warning operations according to the safety early warning result; Wherein, determining the first target weight corresponding to the first LoRa sensor and the second target weight corresponding to the second LoRa sensor according to the first monitoring data, the second monitoring data, and the first fusion data includes: Determining a first initial weight corresponding to the first LoRa sensor and a second initial weight corresponding to the second LoRa sensor; Fusing the first monitoring data and the second monitoring data according to the first initial weight and the second initial weight to obtain initial fusion data; Obtaining a difference between the initial fusion data and the first fusion data. When the difference is less than a preset value, determining a first accuracy corresponding to the first LoRa sensor according to the first monitoring data, and determining a second accuracy corresponding to the second LoRa sensor according to the second monitoring data; wherein, calculating a first variance corresponding to the first monitoring data and a second variance corresponding to the second monitoring data, then performing weighted summation on the first variance to obtain a first sum, and performing weighted summation on the second variance to obtain a second sum; then determining the first sum as the first accuracy and determining the second sum as the second accuracy; Determining a third accuracy corresponding to the target object according to the first initial weight, the second initial weight, the first accuracy, and the second accuracy; wherein, multiplying the first initial weight and the first accuracy to obtain a first product, and multiplying the second initial weight and the second accuracy to obtain a second product, then adding the first product and the second product to determine the third accuracy corresponding to the target object; When the third accuracy meets the preset condition, the first target weight and the second target weight are determined according to the first accuracy and the second accuracy; wherein, when the third accuracy is less than the preset threshold, the reciprocal of the first accuracy is obtained by reciprocal solving, and the reciprocal of the second accuracy is obtained by reciprocal solving, and then the sum of the first reciprocal and the second reciprocal is obtained to get the reciprocal sum value, and then the first reciprocal is multiplied by the reciprocal sum value to obtain a third product, so as to determine the reciprocal value corresponding to the third product as the first target weight, and the second reciprocal is multiplied by the reciprocal sum value to obtain a fourth product, so as to determine the reciprocal value corresponding to the fourth product as the second target weight.

8. A terminal device, characterized in that, The terminal device includes a processor and a memory; The memory is used for storing a computer program; The processor is used for executing the computer program and implementing the indoor monitoring data security early warning method based on deep learning according to any one of claims 1 to 6 when executing the computer program.

9. A computer storage medium for computer storage, characterized in that, The computer storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the indoor monitoring data security early warning method based on deep learning according to any one of claims 1 to 6.

Images