An environmental detection and early warning method and system based on artificial intelligence

Abstract

The embodiment of the present invention provides an artificial intelligence-based environment detection and early warning method and system, which belongs to the field of data processing technology. The method includes: using a target sensor to collect initial sensor data corresponding to a target environment; performing abnormal data identification on the initial sensor data to obtain initial abnormal data corresponding to the target environment; using a data enhancement model to perform data expansion on the initial abnormal data to obtain target abnormal data; performing abnormal classification on the target abnormal data according to an abnormal classification model to obtain a target abnormal type corresponding to the target environment; and determining a target early warning result corresponding to the target environment according to the target abnormal type. The problem that the small amount of abnormal data in the related technology affects the accuracy of abnormal detection and the early warning effect is not ideal is solved, and the timeliness and accuracy of responding to environmental changes are improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an artificial intelligence-based environment detection and early warning method and system. Background Art

[0002] In the process of environmental detection, since the number of abnormal data is usually small, traditional classification algorithms tend to learn the characteristics of most normal samples, resulting in reduced sensitivity to a few abnormal samples. As a result, some situations that are actually abnormal may be misclassified as normal during anomaly identification, or the occurrence of abnormal events may not be detected at all. This not only affects the accuracy of anomaly detection, but also further causes the early warning effect to be less than ideal, affecting the timeliness and accuracy of responding to environmental changes. Summary of the invention

[0003] The main purpose of the embodiments of the present invention is to provide an artificial intelligence-based environmental detection and early warning method and system, aiming to solve the problem in related technologies that the small amount of abnormal data affects the accuracy of abnormal detection and the early warning effect is not ideal.

[0004] In a first aspect, an embodiment of the present invention provides an environment detection and early warning method based on artificial intelligence, comprising:

[0005] Using the target sensor to collect initial sensor data corresponding to the target environment;

[0006] Performing abnormal data identification on the initial sensor data to obtain initial abnormal data corresponding to the target environment;

[0007] Using a data enhancement model to perform data expansion on the initial abnormal data to obtain target abnormal data;

[0008] Performing abnormal classification on the target abnormal data according to the abnormal classification model to obtain the target abnormal type corresponding to the target environment;

[0009] A target warning result corresponding to the target environment is determined according to the target abnormality type.

[0010] In a second aspect, an embodiment of the present invention provides an artificial intelligence-based environment detection and early warning system, comprising:

[0011] A data acquisition module, used to collect initial sensor data corresponding to the target environment using the target sensor;

[0012] An abnormality identification module, used to identify abnormal data on the initial sensor data and obtain initial abnormal data corresponding to the target environment;

[0013] A data expansion module, used to expand the initial abnormal data using a data enhancement model to obtain target abnormal data;

[0014] An anomaly classification module, used to perform anomaly classification on the target anomaly data according to an anomaly classification model to obtain a target anomaly type corresponding to the target environment;

[0015] The abnormality warning module is used to determine the target warning result corresponding to the target environment according to the target abnormality type.

[0016] The embodiment of the present invention provides an artificial intelligence-based environment detection and early warning method and system, the method comprising: first, using a target sensor to collect initial sensor data of a target environment. The target sensor can accurately obtain key data related to the target environment, ensure the timeliness and accuracy of the collected initial sensor data, and thus lay a solid foundation for subsequent abnormal identification. Then, the initial sensor data is subjected to abnormal data identification to obtain initial abnormal data corresponding to the target environment. By implementing an efficient abnormal identification algorithm, potential environmental threats can be effectively identified, and the sensitivity of the system to abnormal events can be significantly improved. This sensitivity can not only detect deviations from the normal state in a timely manner, but also provide a reliable basis for early intervention, thereby reducing possible environmental risks. Subsequently, the initial abnormal data is expanded using a data enhancement model to obtain richer target abnormal data. After obtaining the target abnormal data, the abnormal classification model is used to analyze and process it to identify the corresponding target abnormal type in the target environment. Accurate classification provides an accurate basis for subsequent processing and countermeasures. Finally, a corresponding target early warning result is generated according to the accurately identified target abnormal type. This effectively solves the problem of low anomaly detection accuracy and unsatisfactory early warning effect caused by insufficient number of abnormal data samples in traditional technologies, and significantly improves the timeliness and accuracy of response to changes in the target environment, providing strong support for achieving sustainable environmental management and protecting the target environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A schematic diagram of a process flow of an artificial intelligence-based environment detection and early warning method provided by an embodiment of the present invention;

[0019] Figure 2A schematic diagram of the module structure of an artificial intelligence-based environmental detection and early warning system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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.

[0021] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0022] It should be understood that the terms used in this 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 the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0023] The embodiment of the present invention provides an artificial intelligence-based environment detection and early warning method and system. The artificial intelligence-based environment detection and early warning method can be applied to a terminal device, which can be an electronic device such as a tablet computer, a laptop computer, a desktop computer, a personal digital assistant, and a wearable device. The terminal device can be a server or a server cluster.

[0024] Some embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0025] Please refer to Figure 1 , Figure 1 A flowchart of an artificial intelligence-based environmental detection and early warning method provided in an embodiment of the present invention.

[0026] like Figure 1 As shown, the artificial intelligence-based environment detection and early warning method includes steps S101 to S105.

[0027] Step S101: using a target sensor to collect initial sensor data corresponding to a target environment.

[0028] Exemplarily, target sensors such as meteorological sensors, water quality sensors, air quality sensors, etc. are used to collect initial sensor data corresponding to the target environment, and the initial sensor data includes but is not limited to temperature data, humidity data, air pollutant content data, water body index data, etc. At the same time, satellite remote sensing technology can also be used to obtain geographic data and environmental data corresponding to a large range of target environments.

[0029] Step S102: performing abnormal data identification on the initial sensor data to obtain initial abnormal data corresponding to the target environment.

[0030] Exemplarily, normal sensor data collected under the target sensor is obtained, and then the normal sensor data is analyzed through statistical analysis to set a distinction threshold corresponding to when the data collected under the target sensor is normal, and then the initial sensor data is compared with the distinction threshold. When the initial sensor data is greater than or equal to the distinction threshold, the initial sensor data is determined as the initial abnormal data corresponding to the target environment.

[0031] In some embodiments, the abnormal data identification of the initial sensor data to obtain the initial abnormal data corresponding to the target environment includes: determining a first window and a first step length, and performing data segmentation on the initial sensor data according to the first window and the first step length to obtain multiple sub-sensor data; respectively calculating the change rate of adjacent data in the sub-sensor data to obtain change rate sequence data corresponding to the sub-sensor data; and performing abnormality identification on the change rate sequence data to obtain the initial abnormal data corresponding to the target environment.

[0032] Exemplarily, the first window is used to determine the amount of data to be inspected each time in the initial sensor data analysis. The first step length is the distance the first window moves forward when the next window movement is performed. The first step length affects the degree of overlap of the initial sensor data and the granularity of the segmentation.

[0033] Exemplarily, the first window and the first step length are set according to demand or expert experience, and then the initial sensor data is segmented according to the first window and the first step length to obtain a plurality of sub-sensor data.

[0034] For example, starting from the starting point of the initial sensor data, the first window is formed. The initial sensor data of the same size as the first window is extracted from the initial sensor data to form the first sub-sensor data, and then the first window is moved forward based on the first step length to continue to extract the data of the next first window, and then the above steps are repeated until all parts of the initial sensor data are covered. Thus, multiple sub-sensor data are obtained.

[0035] Exemplarily, the change rate calculation can generally be defined as the difference between the current value and the previous value, thereby calculating the change rate of each pair of adjacent data by stepping through the data points in the sub-sensor data, and then storing the change rate of each pair of adjacent data in sequence to obtain the change rate sequence data.

[0036] Exemplarily, statistical methods, machine learning algorithms or other specialized anomaly detection algorithms are used to analyze the change rate series data to screen out abnormal data points in the change rate series data, thereby obtaining sensor data corresponding to the abnormal data points in the initial sensor data, and then determining the sensor data corresponding to the abnormal data points as initial abnormal data.

[0037] Specifically, by determining the first window and the first step length, the initial sensor data can be divided into multiple sub-sensor data, making data analysis more orderly. Calculating the change rate of adjacent data in the sub-sensor data can not only reveal the trend and speed of data change, but also make it easier to identify obvious changes or mutations in the change rate sequence data, thereby helping to identify potential abnormal fluctuations and further improve the accuracy of analysis.

[0038] In some embodiments, the abnormality identification of the change rate sequence data to obtain the initial abnormality data corresponding to the target environment includes: performing data processing on the change rate sequence data to obtain a target mean and a target variance corresponding to the change rate sequence data; determining a data upper limit and a data lower limit corresponding to the change rate sequence data based on the target mean and the target variance; determining a target characterization value corresponding to the change rate sequence data based on the data upper limit and the data lower limit; and performing abnormality identification on the sub-sensor data corresponding to the change rate sequence data based on the target characterization value and a preset value to obtain the initial abnormality data corresponding to the target environment.

[0039] Exemplarily, all values ​​of the rate of change series data are obtained and their arithmetic mean is calculated, that is, all the rate of change values ​​are added and divided by the number of rate of change series data to obtain the target mean. The target mean can reflect the central tendency of the rate of change series data. The standard deviation formula is then used to calculate the target variance of the rate of change series data, which represents the discreteness and volatility of the rate of change series data.

[0040] Exemplarily, a random variable is determined according to a normal distribution, and then a square root operation is performed on the number of change rate sequence data to obtain a square root operation result, and the fluctuation value of the change rate sequence data is obtained by multiplying the target variance and the random variable and dividing by the square root operation result. Then, the target mean and the fluctuation value are added to obtain the data upper limit, and the target mean and the fluctuation value are subtracted to obtain the data lower limit.

[0041] Exemplarily, the preset value may be derived from historical abnormal data or defined by expert knowledge and should reflect the significance of the magnitude of the change rate.

[0042] Exemplarily, the average value of the difference between the data upper limit and the data lower limit is determined as the target characterization value corresponding to the change rate sequence data, and then the target characterization value is compared with the preset value. When the target characterization value is greater than the preset value, there is abnormal data in the sub-sensor data corresponding to the change rate sequence data, and the sub-sensor data is determined as the initial abnormal data corresponding to the target environment. In addition, since there is data overlap in the sub-sensor data, data merging is required after obtaining the initial abnormal data to reduce the impact of duplicate data on subsequent abnormality determination. When the target characterization value is less than or equal to the preset value, there is no abnormal data in the sub-sensor data corresponding to the change rate sequence data.

[0043] Specifically, by calculating the target mean and target variance, we can fully understand the central trend and dispersion of the rate of change series, which lays a solid foundation for subsequent anomaly identification. The upper and lower limits of the data determined based on the target mean and target variance can help accurately obtain the target representation value corresponding to the rate of change series data, thereby providing support for improving the sensitivity and accuracy of anomaly detection.

[0044] In some embodiments, the abnormality identification of the change rate sequence data to obtain the initial abnormality data corresponding to the target environment includes: determining a second window and a second step size, and segmenting the change rate sequence data according to the second window and the second step size to obtain multiple subsequence data; extracting statistical features of each subsequence data to obtain target statistical features corresponding to the subsequence data; clustering the target statistical features of all the subsequence data to obtain target clustering results; identifying abnormalities of the change rate sequence data corresponding to the subsequence data according to the target clustering results to obtain sequence abnormality data; and obtaining the initial abnormality data corresponding to the target environment according to the sequence abnormality data.

[0045] Exemplarily, the second window is used to determine the amount of data to be tested each time in the change rate series data analysis. The second step length is the distance the second window moves forward when the next window movement is performed. The second step length affects the overlap degree and segmentation granularity of the change rate series data.

[0046] Exemplarily, the second window and the second step size are set according to demand or expert experience, and then the change rate sequence data is segmented according to the second window and the second step size to obtain a plurality of subsequence data.

[0047] Exemplarily, statistical features are determined, and the statistical features may include mean, standard deviation, maximum value, minimum value, skewness, kurtosis, etc. Thus, the selected statistical features are calculated for each subsequence data to obtain the target statistical features.

[0048] Exemplarily, data clustering is performed based on the target statistical features corresponding to all subsequence data using clustering algorithms such as K-Means, hierarchical clustering, DBSCAN, etc. to obtain a target clustering result. If the number of a certain cluster in the target clustering result is significantly less than that of other clusters, or the distance between the data in each cluster and the center of the cluster is too far, the subsequence data corresponding to the cluster or the data in the cluster that is far from the center of the cluster is identified as sequence abnormal data.

[0049] Exemplarily, the sequence abnormal data are merged to obtain initial abnormal data corresponding to the target environment.

[0050] Specifically, dividing the rate of change sequence data into multiple subsequence data allows for focused analysis of local features of the rate of change sequence data, helping to reveal subtle changes that may be overlooked by global analysis, and then extracting statistical features from each subsequence data to obtain rich information that helps understand the basic behavior patterns of the subsequence data. Thus, by clustering the target statistical features, the similarities and differences between different subsequences can be identified, and the initial abnormal data can be accurately identified, providing a strong foundation for subsequent improvement of the accuracy of data analysis.

[0051] Step S103: using a data enhancement model to perform data expansion on the initial abnormal data to obtain target abnormal data.

[0052] Exemplarily, the data augmentation model is a generative adversarial network, which is then used to perform data augmentation on initial abnormal data to obtain target abnormal data.

[0053] In some embodiments, the use of the data enhancement model to perform data expansion on the initial abnormal data to obtain target abnormal data includes: using the data embedding layer of the data enhancement model to perform data mapping on the initial abnormal data to obtain target mapping data corresponding to the initial abnormal data; using the noise generation layer of the data enhancement model to randomly generate noise data corresponding to the initial abnormal data; using the spatial extraction layer of the data enhancement model to perform feature extraction on the target mapping data to obtain corresponding target feature data; using the data generation layer of the data enhancement model to perform data generation according to the target feature data and the noise data to obtain initial generated data; and performing data screening according to the initial generated data to obtain the target abnormal data.

[0054] Exemplarily, the data augmentation model includes a data embedding layer, a noise generation layer, a spatial extraction layer, and a data generation layer.

[0055] Exemplarily, the initial abnormal data is input into the data embedding layer of the data augmentation model. The neural network structure of this layer is designed to learn the potential representation of the initial abnormal data and map the initial abnormal data to a new space, thereby obtaining the target mapping data corresponding to the initial abnormal data. The target mapping data can be considered as a deep representation of the initial abnormal data.

[0056] For example, in the noise generation layer of the data enhancement model, a Gaussian distribution is used to generate random noise data corresponding to the initial abnormal data. The noise data can increase the diversity and complexity of the initial abnormal data, which helps to increase the generalization ability of the data enhancement model.

[0057] Exemplarily, the spatial extraction layer in the data enhancement model is used to perform feature extraction on the target mapping data. The extracted feature data represents important attributes of the initial abnormal data to obtain target feature data. The target feature data can help the data enhancement model better understand and generate data.

[0058] Exemplarily, the data generation layer in the data enhancement model generates new initial generated data using the extracted target feature data and noise data, and then selects qualified data as target abnormal data based on the distance or standard deviation between the initial generated data and the initial abnormal data.

[0059] In some embodiments, the data screening based on the initial generated data to obtain the target abnormal data includes: performing covariance calculation on the initial abnormal data to obtain a covariance matrix corresponding to the initial abnormal data; performing eigenvalue decomposition on the covariance matrix to obtain target-related features corresponding to the initial abnormal data; performing data characterization on the initial abnormal data using the target-related features to obtain first characterization data; performing data characterization on the initial generated data using the target-related features to obtain second characterization data; and performing data screening on the initial generated data based on the first characterization data and the second characterization data to obtain the target abnormal data.

[0060] Exemplarily, the initial abnormal data is feature calculated, and the covariance between the features is calculated using statistical methods to form a covariance matrix. The covariance matrix shows the relationship between different features and helps identify the correlation of features. The calculated covariance matrix is ​​then subjected to eigenvalue decomposition to obtain eigenvalues ​​and corresponding eigenvectors. The eigenvalue represents the variance of the feature, reflecting the importance of the feature to the data variation, so that the feature is screened according to the eigenvalue to obtain the target associated feature.

[0061] Exemplarily, the initial abnormal data is linearly transformed using the target-related features, i.e., important feature vectors, to obtain the first characterization data. This step is to retain the main information of the data by projecting the initial abnormal data into the feature space. Similarly, the initial generated data is transformed in the same way using the target-related features to obtain the second characterization data. This will enable the initial generated data to be represented in the same feature space as the initial abnormal data.

[0062] Exemplarily, a similarity measurement is performed based on the first characterization data and the second characterization data, so that when the similarity is greater than a preset similarity value, the initially generated data is determined as the target abnormal data; when the similarity is less than or equal to the preset similarity value, the initially generated data is eliminated.

[0063] Specifically, the covariance matrix and eigenvalue decomposition are used to obtain important features of the initial abnormal data, further characterize the initial abnormal data and the initial generated data, and then effective abnormal data screening is performed based on the characterization results, thereby improving the reliability of the target abnormal data.

[0064] In some embodiments, the data screening of the initial generated data according to the first characterization data and the second characterization data to obtain the target abnormal data includes: determining a target distance between the initial generated data and the initial generated data according to the first characterization data and the second characterization data; determining a maximum distance corresponding to the initial abnormal data according to the first characterization data; and data screening of the initial generated data according to the target distance and the maximum distance to obtain the target abnormal data; wherein the target distance is obtained according to the following formula:

[0065]

[0066] in, Represents the i-th initial generated data With the initial abnormal data The corresponding target distance, represents the second representation data corresponding to the i-th initially generated data, represents the first representation data corresponding to the j-th initial abnormal data, N represents the number of targets corresponding to the initial abnormal data, represents the target associated feature, Indicates transposition processing.

[0067] Exemplarily, the target distance is obtained according to the following formula:

[0068]

[0069] in, Represents the i-th initial generated data With initial abnormal data The corresponding target distance, represents the second representation data corresponding to the i-th initial generated data, represents the first representation data corresponding to the jth initial abnormal data, N represents the number of targets corresponding to the initial abnormal data, represents the target-related features, Indicates transposition processing.

[0070] Exemplarily, the first characterization data corresponding to each initial abnormal data is obtained, and then the first characterization data of all the initial abnormal data are averaged to obtain the average characterization data, thereby obtaining the i-th initial generated data The first difference between the initial generated data and the average characterization data is obtained, and then the first difference is transposed to obtain a second difference, and then the first difference and the inverse of the target associated feature are multiplied and then multiplied with the second difference and then squared to obtain the target distance between the initial generated data and the initial generated data.

[0071] Exemplarily, the first characterization data corresponding to each initial abnormal data is obtained, and the third difference between the first characterization data corresponding to each initial abnormal data and the average characterization data is obtained, and then the third difference is transposed to obtain the fourth difference, and then the third difference and the inverse of the target association feature are multiplied and then multiplied with the fourth difference and then squared to obtain the data distance between the initial abnormal data, and then the maximum value in the data distance is determined as the maximum distance corresponding to the initial abnormal data.

[0072] Exemplarily, when the target distance is greater than or equal to the maximum distance, the initial generated data corresponding to the target distance is eliminated; when the target distance is less than the maximum distance, the initial generated data corresponding to the target distance is determined as target abnormal data.

[0073] Step S104: classify the target abnormal data according to the abnormal classification model to obtain the target abnormal type corresponding to the target environment.

[0074] Exemplarily, the anomaly classification model is a support vector machine, a decision tree, a random forest, a neural network, etc. The anomaly classification model is trained according to the historical anomaly data to obtain a trained anomaly classification model. Then, the target anomaly data is classified according to the anomaly classification model to obtain the target anomaly type corresponding to the target environment.

[0075] In some embodiments, the method of performing abnormal classification on the target abnormal data according to the abnormal classification model to obtain the target abnormal type corresponding to the target environment includes: determining a third window and a third step size, and performing data segmentation on the target abnormal data according to the third window and the third step size to obtain multiple sub-abnormal data; performing abnormal classification on the sub-abnormal data according to the abnormal classification model to obtain abnormal distribution information; determining abnormal distribution similarity values ​​between the sub-abnormal data according to the abnormal distribution information; and determining the target abnormal type corresponding to the target environment according to the abnormal distribution similarity values.

[0076] Exemplarily, the third window is used to determine the amount of data to be inspected each time in the target abnormal data analysis. The third step length is the distance the third window moves forward when the next window movement is performed. The third step length affects the overlap degree of the target abnormal data and the granularity of the segmentation.

[0077] Exemplarily, the target abnormal data is sliced ​​according to the determined third window and third step length. The target abnormal data is segmented according to the third window and the third step length to obtain a plurality of sub-abnormal data.

[0078] Exemplarily, a pre-trained anomaly classification model is applied to each sub-anomaly data, and then the anomaly classification model analyzes the anomaly distribution information corresponding to each sub-anomaly data. The anomaly distribution information includes the distribution of each sub-anomaly data on each anomaly type, such as the probability of each anomaly type.

[0079] Exemplarily, according to the abnormal distribution information of each sub-abnormal data, the similarity value between the abnormal distribution information corresponding to the sub-abnormal data is calculated to obtain the abnormal distribution similarity value, and the relationship and similarity between the corresponding distributions of the sub-abnormal data are evaluated. Thus, the sub-abnormal data are clustered according to the abnormal distribution similarity value to obtain the abnormal clustering result.

[0080] Exemplarily, the anomaly distribution information corresponding to each cluster in the anomaly clustering result is summarized to obtain the anomaly type corresponding to the cluster, and then the anomaly type corresponding to each cluster is merged to obtain the target anomaly type corresponding to the target environment.

[0081] For example, the target clustering results include the first cluster, the second cluster and the third cluster, where the abnormal distribution information corresponding to each sub-abnormal data in the first cluster is voted to obtain the abnormal type corresponding to the first cluster as type 1, the abnormal distribution information corresponding to each sub-abnormal data in the second cluster is voted to obtain the abnormal type corresponding to the second cluster as type 2, and the abnormal distribution information corresponding to each sub-abnormal data in the third cluster is voted to obtain the abnormal type corresponding to the third cluster as type 2. Then, after merging the abnormal types corresponding to the first cluster, the second cluster and the third cluster, the target abnormal types can be obtained as type 1 and type 2.

[0082] Specifically, by determining the anomaly distribution similarity values ​​between sub-anomaly data, it is possible to identify potential relationships and patterns between different data segments. This similarity analysis can help discover potential common causes or correlations, making the analysis more comprehensive.

[0083] Step S105: Determine a target warning result corresponding to the target environment according to the target abnormality type.

[0084] Exemplarily, the frequency of occurrence of the target abnormal type within a preset time is obtained, and the corresponding warning threshold is set. For example, when the number of occurrences of the target abnormal type within a specified time exceeds the warning threshold, the corresponding target warning result will be generated. The target warning result includes information such as the time of recording the warning, the target abnormal type, and the frequency of occurrence. The target warning result is promptly notified to the relevant management personnel so that measures can be taken quickly.

[0085] In some embodiments, determining the target warning result corresponding to the target environment according to the target abnormality type includes: determining a strategy mapping table, and determining the target response strategy corresponding to the target environment according to the target warning result and the strategy mapping table; generating the target warning result corresponding to the target environment according to the target abnormality type and the target response strategy, and sending the target warning result to the target terminal.

[0086] Exemplarily, a policy mapping table is determined, and the policy mapping table matches the target anomaly type with the corresponding response strategy. The policy mapping table should clearly list the response strategies corresponding to each target anomaly type to ensure that it can take appropriate actions according to different warning situations. Therefore, the policy mapping table is used to find the corresponding target response strategy according to the target anomaly type monitored in the target environment. Therefore, the corresponding target warning result is generated by combining the identified target anomaly type and the determined target response strategy. And the target warning result is sent to the target terminal, and the method of sending the target warning result to the target terminal may include email, short message, push notification, system alarm, etc. Therefore, the generated target warning result is sent to the target terminal through a determined communication channel. Ensure that the information is clear, accurate and easy to understand so that the recipient can quickly understand the situation and take action.

[0087] The embodiment of the present invention provides an artificial intelligence-based environment detection and early warning method and system, the method comprising: first, using a target sensor to collect initial sensor data of a target environment. The target sensor can accurately obtain key data related to the target environment, ensure the timeliness and accuracy of the collected initial sensor data, and thus lay a solid foundation for subsequent abnormal identification. Then, the initial sensor data is subjected to abnormal data identification to obtain initial abnormal data corresponding to the target environment. By implementing an efficient abnormal identification algorithm, potential environmental threats can be effectively identified, and the sensitivity of the system to abnormal events can be significantly improved. This sensitivity can not only detect deviations from the normal state in a timely manner, but also provide a reliable basis for early intervention, thereby reducing possible environmental risks. Subsequently, the initial abnormal data is expanded using a data enhancement model to obtain richer target abnormal data. After obtaining the target abnormal data, the abnormal classification model is used to analyze and process it to identify the corresponding target abnormal type in the target environment. Accurate classification provides an accurate basis for subsequent processing and countermeasures. Finally, a corresponding target early warning result is generated according to the accurately identified target abnormal type. This effectively solves the problem of low anomaly detection accuracy and unsatisfactory early warning effect caused by insufficient number of abnormal data samples in traditional technologies, and significantly improves the timeliness and accuracy of response to changes in the target environment, providing strong support for achieving sustainable environmental management and protecting the target environment.

[0088] See also Figure 2 , Figure 2 An artificial intelligence-based environment detection and early warning system 200 is provided in an embodiment of the present application. The artificial intelligence-based environment detection and early warning system 200 includes a data acquisition module 201, an anomaly recognition module 202, a data expansion module 203, an anomaly classification module 204, and an anomaly warning module 205, wherein the data acquisition module 201 is used to use a target sensor to collect initial sensor data corresponding to a target environment; the anomaly recognition module 202 is used to perform abnormal data recognition on the initial sensor data to obtain initial abnormal data corresponding to the target environment; the data expansion module 203 is used to perform data expansion on the initial abnormal data using a data enhancement model to obtain target abnormal data; the anomaly classification module 204 is used to perform abnormal classification on the target abnormal data according to the anomaly classification model to obtain a target abnormal type corresponding to the target environment; the anomaly warning module 205 is used to determine a target warning result corresponding to the target environment according to the target abnormal type.

[0089] In some implementations, the abnormality identification module 202 performs, during the process of identifying abnormal data on the initial sensor data and obtaining initial abnormal data corresponding to the target environment:

[0090] Determine a first window and a first step length, and perform data segmentation on the initial sensor data according to the first window and the first step length to obtain a plurality of sub-sensor data;

[0091] Calculating the change rate of adjacent data in the sub-sensor data respectively to obtain the change rate sequence data corresponding to the sub-sensor data;

[0092] Anomalies are identified on the change rate sequence data to obtain the initial abnormal data corresponding to the target environment.

[0093] In some implementations, the anomaly identification module 202, in the process of identifying anomalies on the change rate sequence data and obtaining the initial anomaly data corresponding to the target environment, performs:

[0094] Performing data processing on the change rate sequence data to obtain a target mean and a target variance corresponding to the change rate sequence data;

[0095] Determine the data upper limit and data lower limit corresponding to the change rate sequence data according to the target mean and the target variance;

[0096] Determine a target characterization value corresponding to the change rate sequence data according to the data upper limit and the data lower limit;

[0097] The sub-sensor data corresponding to the change rate sequence data is subjected to abnormality identification according to the target characterization value and the preset value to obtain the initial abnormality data corresponding to the target environment.

[0098] In some implementations, the anomaly identification module 202, in the process of identifying anomalies on the change rate sequence data and obtaining the initial anomaly data corresponding to the target environment, performs:

[0099] Determine a second window and a second step length, and perform data segmentation on the change rate sequence data according to the second window and the second step length to obtain a plurality of subsequence data;

[0100] Extracting statistical features from each of the subsequence data to obtain target statistical features corresponding to the subsequence data;

[0101] Performing data clustering on the target statistical features of all the subsequence data to obtain a target clustering result;

[0102] performing abnormality identification on the change rate sequence data corresponding to the subsequence data according to the target clustering result to obtain sequence abnormality data;

[0103] The initial abnormal data corresponding to the target environment is obtained according to the sequence abnormal data.

[0104] In some implementations, the data expansion module 203 performs, during the process of using the data enhancement model to expand the initial abnormal data to obtain target abnormal data:

[0105] Using the data embedding layer of the data enhancement model to perform data mapping on the initial abnormal data to obtain target mapping data corresponding to the initial abnormal data;

[0106] Randomly generate noise data corresponding to the initial abnormal data using the noise generation layer of the data enhancement model;

[0107] Using the spatial extraction layer of the data enhancement model to perform feature extraction on the target mapping data to obtain corresponding target feature data;

[0108] Using the data generation layer of the data enhancement model to generate data according to the target feature data and the noise data to obtain initial generated data;

[0109] The target abnormal data is obtained by performing data screening according to the initially generated data.

[0110] In some implementations, the data expansion module 203 performs, during the process of performing data screening according to the initially generated data to obtain the target abnormal data:

[0111] Performing covariance calculation on the initial abnormal data to obtain a covariance matrix corresponding to the initial abnormal data;

[0112] Performing eigenvalue decomposition on the covariance matrix to obtain target correlation features corresponding to the initial abnormal data;

[0113] Using the target association feature to perform data characterization on the initial abnormal data to obtain first characterization data;

[0114] Using the target-related feature to characterize the initially generated data, obtaining second characterization data;

[0115] The target abnormal data is obtained by performing data screening on the initially generated data according to the first characterization data and the second characterization data.

[0116] In some implementations, the data expansion module 203 performs, during the process of filtering the initially generated data according to the first characterization data and the second characterization data to obtain the target abnormal data:

[0117] determining a target distance between the initial generated data and the initial generated data according to the first characterization data and the second characterization data;

[0118] Determine a maximum distance corresponding to the initial abnormal data according to the first characterization data;

[0119] Performing data screening on the initially generated data according to the target distance and the maximum distance to obtain the target abnormal data;

[0120] The target distance is obtained according to the following formula:

[0121]

[0122] in, Represents the i-th initial generated data With the initial abnormal data The corresponding target distance, represents the second representation data corresponding to the i-th initially generated data, represents the first representation data corresponding to the j-th initial abnormal data, N represents the number of targets corresponding to the initial abnormal data, represents the target associated feature, Indicates transposition processing.

[0123] In some implementations, the abnormality classification module 204 performs, during the process of performing abnormality classification on the target abnormal data according to the abnormality classification model to obtain the target abnormality type corresponding to the target environment:

[0124] Determine a third window and a third step length, and perform data segmentation on the target abnormal data according to the third window and the third step length to obtain a plurality of sub-abnormal data;

[0125] Performing abnormal classification on the sub-abnormal data according to the abnormal classification model to obtain abnormal distribution information;

[0126] Determine an abnormal distribution similarity value between the sub-abnormal data according to the abnormal distribution information;

[0127] The target anomaly type corresponding to the target environment is determined according to the anomaly distribution similarity value.

[0128] In some implementations, the abnormal warning module 205, in the process of determining the target warning result corresponding to the target environment according to the target abnormality type, performs:

[0129] Determine a strategy mapping table, and determine a target response strategy corresponding to the target environment according to the target warning result and the strategy mapping table;

[0130] The target warning result corresponding to the target environment is generated according to the target abnormality type and the target response strategy, and the target warning result is sent to the target terminal.

[0131] In some embodiments, the artificial intelligence-based environment detection and early warning system 200 can be applied to terminal devices.

[0132] It should be noted that technical personnel in the relevant field can clearly understand that, for the convenience and conciseness of description, the specific working process of the artificial intelligence-based environmental detection and early warning system 200 described above can refer to the corresponding process in the aforementioned artificial intelligence-based environmental detection and early warning method embodiment, and will not be repeated here.

[0133] An embodiment of the present invention also provides a storage medium for computer-readable storage, wherein 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 artificial intelligence-based environmental detection and early warning methods provided in the description of the embodiment of the present invention.

[0134] The storage medium may be an internal storage unit of the terminal device described in the foregoing embodiment, such as a hard disk or memory of the terminal device. The storage medium may also be an external storage device of the terminal device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc., equipped on the terminal device.

[0135] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In a hardware embodiment, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed 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 implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transient medium). As is known to those skilled 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 technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic 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 skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0136] It should be understood that the term "and / or" used in the present specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations. It should be noted that, in this article, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "including a..." does not exclude the presence of other identical elements in the process, method, article or system including the element.

[0137] The serial numbers of the embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. The above description is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. An artificial intelligence-based environmental detection and early warning method, characterized in that: The method comprises: Using the target sensor to collect initial sensor data corresponding to the target environment; Performing abnormal data identification on the initial sensor data to obtain initial abnormal data corresponding to the target environment; Using a data enhancement model to perform data expansion on the initial abnormal data to obtain target abnormal data; Performing abnormal classification on the target abnormal data according to the abnormal classification model to obtain the target abnormal type corresponding to the target environment; Determine a target warning result corresponding to the target environment according to the target abnormality type; The step of using a data enhancement model to perform data expansion on the initial abnormal data to obtain target abnormal data includes: Using the data embedding layer of the data enhancement model to perform data mapping on the initial abnormal data to obtain target mapping data corresponding to the initial abnormal data; Randomly generate noise data corresponding to the initial abnormal data using the noise generation layer of the data enhancement model; Using the spatial extraction layer of the data enhancement model to perform feature extraction on the target mapping data to obtain corresponding target feature data; Using the data generation layer of the data enhancement model to generate data according to the target feature data and the noise data to obtain initial generated data; Performing covariance calculation on the initial abnormal data to obtain a covariance matrix corresponding to the initial abnormal data; Performing eigenvalue decomposition on the covariance matrix to obtain target correlation features corresponding to the initial abnormal data; Using the target-related features, linearly transforming the initial abnormal data to obtain first representation data, where the first representation data is data corresponding to main information of the initial abnormal data retained by projecting the initial abnormal data into a feature space; Using the target-related features, linearly transforming the initial generated data to obtain second representation data, where the second representation data is data corresponding to main information of the initial generated data retained by projecting the initial generated data into a feature space; determining a target distance between the initial generated data and the initial abnormal data according to the first characterization data and the second characterization data; Determine a maximum distance corresponding to the initial abnormal data according to the first characterization data; Performing data screening on the initially generated data according to the target distance and the maximum distance to obtain the target abnormal data; The target distance is obtained according to the following formula: ； in, Represents the i-th initial generated data With the initial abnormal data The corresponding target distance, represents the second representation data corresponding to the i-th initially generated data, represents the first representation data corresponding to the j-th initial abnormal data, N represents the number of targets corresponding to the initial abnormal data, represents the target associated feature, Indicates transposition processing.

2. The method according to claim 1, characterized in that The performing abnormal data identification on the initial sensor data to obtain initial abnormal data corresponding to the target environment includes: Determine a first window and a first step length, and perform data segmentation on the initial sensor data according to the first window and the first step length to obtain a plurality of sub-sensor data; Calculating the change rate of adjacent data in the sub-sensor data respectively to obtain the change rate sequence data corresponding to the sub-sensor data; Anomalies are identified on the change rate sequence data to obtain the initial abnormal data corresponding to the target environment.

3. The method according to claim 2, characterized in that The performing abnormality identification on the change rate sequence data to obtain the initial abnormality data corresponding to the target environment includes: Performing data processing on the change rate sequence data to obtain a target mean and a target variance corresponding to the change rate sequence data; Determine the data upper limit and data lower limit corresponding to the change rate sequence data according to the target mean and the target variance; Obtaining an average value of the difference between the data upper limit and the data lower limit as a target characterization value corresponding to the change rate sequence data; The sub-sensor data corresponding to the change rate sequence data is subjected to abnormality identification according to the target characterization value and the preset value to obtain the initial abnormality data corresponding to the target environment.

4. The method according to claim 2, characterized in that: The performing abnormality identification on the change rate sequence data to obtain the initial abnormality data corresponding to the target environment includes: Determine a second window and a second step length, and perform data segmentation on the change rate sequence data according to the second window and the second step length to obtain a plurality of subsequence data; Extracting statistical features from each of the subsequence data to obtain target statistical features corresponding to the subsequence data; Performing data clustering on the target statistical features of all the subsequence data to obtain a target clustering result; performing abnormality identification on the change rate sequence data corresponding to the subsequence data according to the target clustering result to obtain sequence abnormality data; The initial abnormal data corresponding to the target environment is obtained according to the sequence abnormal data.

5. The method according to claim 1, characterized in that The performing abnormal classification on the target abnormal data according to the abnormal classification model to obtain the target abnormal type corresponding to the target environment includes: Determine a third window and a third step length, and perform data segmentation on the target abnormal data according to the third window and the third step length to obtain a plurality of sub-abnormal data; Performing abnormal classification on the sub-abnormal data according to the abnormal classification model to obtain abnormal distribution information; Determine an abnormal distribution similarity value between the sub-abnormal data according to the abnormal distribution information; The target anomaly type corresponding to the target environment is determined according to the anomaly distribution similarity value.

6. The method according to claim 1, characterized in that The determining, according to the target abnormality type, a target warning result corresponding to the target environment includes: Determine a strategy mapping table, and determine a target response strategy corresponding to the target environment according to the target anomaly type and the strategy mapping table; The target warning result corresponding to the target environment is generated according to the target abnormality type and the target response strategy, and the target warning result is sent to the target terminal.

7. An artificial intelligence-based environmental detection and early warning system, characterized in that: include: A data acquisition module, used to collect initial sensor data corresponding to the target environment using the target sensor; An abnormality identification module, used to identify abnormal data on the initial sensor data and obtain initial abnormal data corresponding to the target environment; A data expansion module is used to use a data enhancement model to perform data expansion on the initial abnormal data to obtain target abnormal data; wherein, the use of the data enhancement model to perform data expansion on the initial abnormal data to obtain target abnormal data includes: using the data embedding layer of the data enhancement model to perform data mapping on the initial abnormal data to obtain target mapping data corresponding to the initial abnormal data; using the noise generation layer of the data enhancement model to randomly generate noise data corresponding to the initial abnormal data; using the space extraction layer of the data enhancement model to perform feature extraction on the target mapping data to obtain corresponding target feature data; using the data generation layer of the data enhancement model to perform data generation according to the target feature data and the noise data to obtain initial generated data; performing covariance calculation on the initial abnormal data to obtain a covariance matrix corresponding to the initial abnormal data; performing eigenvalue decomposition on the covariance matrix to obtain The target association feature corresponding to the initial abnormal data; using the target association feature to linearly transform the initial abnormal data to obtain first characterization data, the first characterization data is the data corresponding to the main information of the initial abnormal data retained by projecting the initial abnormal data into the feature space; using the target association feature to linearly transform the initial generated data to obtain second characterization data, the second characterization data is the data corresponding to the main information of the initial generated data retained by projecting the initial generated data into the feature space; determining the target distance between the initial generated data and the initial abnormal data according to the first characterization data and the second characterization data; determining the maximum distance corresponding to the initial abnormal data according to the first characterization data; performing data screening on the initial generated data according to the target distance and the maximum distance to obtain the target abnormal data; wherein the target distance is obtained according to the following formula: ； in, Represents the i-th initial generated data With the initial abnormal data The corresponding target distance, represents the second representation data corresponding to the i-th initially generated data, represents the first representation data corresponding to the j-th initial abnormal data, N represents the number of targets corresponding to the initial abnormal data, represents the target associated feature, Indicates transposition processing; An anomaly classification module, used to perform anomaly classification on the target anomaly data according to an anomaly classification model to obtain a target anomaly type corresponding to the target environment; The abnormality warning module is used to determine the target warning result corresponding to the target environment according to the target abnormality type.

Images