An anti-theft alarm system and method for a power adapter based on the Internet of Things

Abstract

The present invention discloses an anti-theft alarm system and method for a power adapter based on the Internet of Things, which relates to the technical field of anti-theft alarm. The system of the present invention includes: a data acquisition and processing module, an abnormal feature recognition module, a deviation analysis module, and a real-time monitoring and alarm decision-making module; the data acquisition and processing module acquires historical data of the power adapter anti-theft alarm device and obtains a historical data feature vector; the abnormal feature recognition module identifies abnormal features of false alarm and missed alarm events according to historical data analysis; the deviation analysis module compares the deviation between normal alarm events and abnormal events and generates a deviation threshold interval; the real-time monitoring and alarm decision-making module determines alarm events according to real-time data and the deviation threshold and issues a notice in a timely manner; the present invention can effectively identify and process alarm events in the power adapter anti-theft alarm system to ensure the accuracy of alarms.

Description

Technical Field

[0001] The present invention relates to the technical field of anti-theft alarm, and particularly to an anti-theft alarm system and method for a power adapter based on the Internet of Things. Background Art

[0002] With the increasing emphasis on the security of digital products in the modern business environment, anti-theft alarm systems have been widely used in shopping malls, electronic product exhibitions, and other retail venues. In these venues, the displayed digital products are usually connected to the alarm through anti-theft alarm sensing wires, and the power adapter provides power support for the anti-theft alarm. Such anti-theft alarm devices are mainly used to prevent digital products from being taken away illegally, and trigger an alarm when the product is removed or the anti-theft alarm wire is unplugged, prompting the salesperson to take timely measures.

[0003] The existing power adapter anti-theft alarm devices usually adopt a connection method in which the alarm sensing wire is directly connected to the alarm host and rely on the power adapter to provide power. However, in some specific cases, for example, due to factors such as voltage fluctuations of the power adapter and changes in the external environment, the existing system may not be able to accurately determine whether the power adapter is supplying power normally, resulting in false alarms or missed alarms. In addition, when the connection between the power adapter and the alarm is disconnected, the existing anti-theft alarm system cannot effectively judge and alarm in real time, resulting in some theft behaviors not being discovered in time. Summary of the Invention

[0004] The purpose of the present invention is to provide an anti-theft alarm system and method for a power adapter based on the Internet of Things to solve the problems raised in the above background art.

[0005] To solve the above technical problems, the present invention provides the following technical solutions:

[0006] An anti-theft alarm method for a power adapter based on the Internet of Things includes the following steps:

[0007] Step S100. Through Internet of Things technology, data collection is carried out on the power adapter anti-theft alarm device, historical data of the power adapter anti-theft alarm device is obtained, the historical data is summarized and analyzed, so as to extract historical data features and form a historical data feature vector;

[0008] Step S200. According to the historical data feature vector, the change trend of each dimension of the historical data feature vector is analyzed to obtain the corresponding historical data feature trend curve; based on the historical data feature trend curve, the abnormal data features corresponding to false alarm and missed alarm events of the power adapter anti-theft alarm device are identified;

[0009] Step S300. Analyze the deviation relationship between the abnormal data characteristics of false alarm and missed alarm events and the normal data characteristics of normal alarm events in the power adapter anti-theft alarm device, so as to obtain the deviation threshold interval between the abnormal data characteristics corresponding to false alarm and missed alarm events and the normal data characteristics of normal alarm events;

[0010] Step S400. Obtain the real-time data of the power adapter anti-theft alarm device, analyze the real-time data, so as to obtain the real-time data characteristics; analyze the deviation relationship between the real-time data characteristics and the normal data characteristics of normal alarm events, and combine the deviation threshold interval to judge whether an alarm event occurs currently, and perform corresponding processing according to the judgment result.

[0011] The power adapter anti-theft alarm device mentioned in the present invention is common in the prior art, and its specific components are: an alarm sensing wire, a power adapter, a sensor monitoring module, and an alarm host; among them, the alarm sensing wire is connected between the anti-theft product and the alarm host, and triggers an alarm signal when the product is illegally removed or the sensing wire is unplugged; the power adapter is used to provide the required power support for the alarm, and is usually connected to a power socket to ensure the normal operation of the alarm device; the sensor monitoring module is responsible for real-time monitoring of the working state of the power adapter and the connection state of the alarm sensing wire; the alarm host receives signals from the alarm sensing wire and the sensor monitoring module, and includes a power adapter anti-theft alarm system, and judges whether to trigger an alarm through the power adapter anti-theft alarm system.

[0012] Further, step S100 includes:

[0013] S101. Through the Internet of Things technology, collect data on the power adapter anti-theft alarm device to obtain historical data of the power adapter anti-theft alarm device; the historical data includes various sensor data in the power adapter anti-theft alarm device and alarm event records; the alarm event records include normal alarm events, false alarm events, and missed alarm events, where the normal alarm event refers to the power adapter anti-theft alarm device emitting an alarm, and after being judged by relevant personnel, there is an alarm event; the false alarm event refers to the power adapter anti-theft alarm device emitting an alarm, and after being judged by relevant personnel, there is no alarm event; the missed alarm event refers to the power adapter anti-theft alarm device not emitting an alarm, and after being judged by relevant personnel, there is an alarm event; summarize various sensor data in the historical data, and preprocess the sensor data. The preprocessing includes preprocessing such as denoising, removing outliers, and filling in missing values to ensure the integrity and accuracy of the data; for the preprocessed various sensor data, use the dynamic time warping (DTW) algorithm to synchronize the timestamps of data from different sensors;

[0014] S102. Set a time window for various types of sensor data for timestamp synchronization, and divide the sensor data into several data blocks based on the set time window; for the sensor data within each time window, calculate the corresponding statistical features, and the statistical features include the mean, standard deviation, maximum value, minimum value, peak value, and skewness of the sensor data; where the mean is used to reflect the overall level of the sensor data unit, the standard deviation is used to measure the degree of data dispersion, the maximum and minimum values are used to capture the fluctuation range of the data, and the peak value is used to measure the sharpness of the data distribution; perform time-frequency feature extraction on the sensor data within each time window using wavelet transform, where the wavelet transform analyzes the time-frequency characteristics of the sensor data through the transform scale and displacement, and its formula is:

[0015] W a,b (t)=(1 / |a|)ψ[(t - b) / a];

[0016] where, W a,b (t) represents the wavelet transform result of the sensor data at the given time point t, given scale a, and displacement b, a represents the scale factor, b represents the translation factor, and t represents the time variable; ψ[(t - b) / a] represents the mother wavelet, which is the basic function for performing wavelet transform. By adjusting a and b, different time-frequency decompositions can be obtained; through wavelet transform, the time-frequency characteristics of the sensor data at different scales and time points are obtained, where these characteristics include energy, frequency band, peak value, skewness, maximum wavelet coefficient, etc.;

[0017] S103. For each data block of each type of sensor data, summarize the corresponding statistical features and time-frequency features as historical data features, and perform normalization processing to form the historical data feature vector L, and L = [l1, l2,..., ln], where l1 represents the feature value of the first dimension of the sensor data block, l2 represents the feature value of the second dimension of the sensor data block; and so on, ln represents the feature value of the nth dimension of the sensor data block, and n represents the total number of dimensions of the historical data feature vector corresponding to the sensor data block.

[0018] Furthermore, for the preprocessed various types of sensor data, use the dynamic time warping (DTW) algorithm to synchronize the timestamps of data from different sensors. The specific content of synchronizing the timestamps of data from different sensors is as follows:

[0019] For time series X = (x1, x2,..., xp) and Y = (y1, y2,..., yq) corresponding to different sensor data, where x1 represents the timestamp corresponding to the first data point in time series X, x2 represents the timestamp corresponding to the second data point in time series X, and so on, xp represents the timestamp corresponding to the p-th data point in time series X; similarly, y1 represents the timestamp corresponding to the first data point in time series Y, y2 represents the timestamp corresponding to the second data point in time series Y, and yq represents the timestamp corresponding to the q-th data point in time series Y; calculate the distance between each timestamp in time series X and time series Y, and the distance calculation formula is: d(u, v) = |xu - yv|, where d(u, v) represents the distance between xu in time series X and yv in time series Y, xu represents the timestamp corresponding to the u-th data point in time series X, and yv represents the timestamp corresponding to the v-th data point in time series Y;

[0020] Define an accumulated distance matrix D to store the cost of the shortest path, where the element D(u, v) of the accumulated distance matrix D represents the minimum cumulative distance from the u-th point of time series X to the v-th point of time series Y, and the calculation formula of the accumulated distance matrix is as follows:

[0021] D(u, v) = d(u, v) + min[D(u - 1, v), D(u, v - 1), D(u - 1, v - 1)];

[0022] Among them, D(u - 1, v) represents the minimum cumulative distance from the (u - 1)-th point of time series X to the v-th point of time series Y; D(u, v - 1) represents the minimum cumulative distance from the u-th point of time series X to the (v - 1)-th point of time series Y; D(u - 1, v - 1) represents the minimum cumulative distance from the (u - 1)-th point of time series X to the (v - 1)-th point of time series Y; for the boundary conditions of the accumulated distance matrix, assume D(0, 0) = 0, which means the starting point, and the cumulative distance is zero when there is no data; for other boundary elements: D(u, 0) = ∑ u k=1 d(k, 0), D(0, v) = ∑ v k=1 d(0, k);

[0023] According to the calculated cumulative distance matrix D, starting from D(p,q), select the one with the minimum value from D(u - 1,v - 1), D(u,v - 1), or D(u - 1,v) as the next step of the path until backtracking to D(0,0). Record the indices corresponding to (u,v) selected each time of backtracking to obtain the path with time stamps aligned in time series X and time series Y, thereby completing the time stamp synchronization of time series X and time series Y, and performing corresponding data processing on the data after time stamp synchronization. Among them, if the physical quantity values corresponding to the data points of the time series vary greatly at different time stamps, data points can be supplemented at the aligned time stamps through interpolation algorithms (such as linear interpolation, spline interpolation, etc.); for noisy sensor data, smoothing methods (such as moving average or Gaussian smoothing) can be used to suppress noise on the synchronized data.

[0024] Further, step S200 includes:

[0025] S201. Extract and summarize the eigenvalues corresponding to each dimension of the historical data feature vector L of various sensors, and draw the corresponding historical data feature trend curve Q in the plane rectangular coordinate system according to the time sequence of the eigenvalues corresponding to each dimension. The horizontal axis of the historical data feature trend curve Q is the time point, and the vertical axis is the eigenvalue; perform smoothing processing on each historical data feature trend curve Q to obtain the smoothed historical data feature trend curve Q1, and the corresponding calculation formula is:

[0026] Q1(t)=α·Q(t)+(1 - α)·Q1(t - 1);

[0027] where Q1(t) represents the historical data feature value corresponding to the t-th moment in the smoothed historical data feature trend curve, Q(t) represents the historical data feature value corresponding to the t-th moment in the original historical data feature trend curve Q; α represents the smoothing factor, and its value range is [0,1]; Q1(t - 1) represents the historical data feature value corresponding to the (t - 1)-th moment in the smoothed historical data feature trend curve;

[0028] S202. For each smoothed historical data feature trend curve Q1, calculate the standard deviation σ and the average value μ of the corresponding data points, calculate the coefficient of variation CV according to the standard deviation σ and the average value μ, and CV = σ / μ; calculate the rate of change ROC for the data points on each historical data feature trend curve Q1, and ROC(t)=[Q1(t)-Q1(t - 1)] / Q1(t - 1), and extract the maximum value M_ROC of the rate of change for the data points on each historical data feature trend curve Q1.

[0029] For false alarm events in the alarm event records of historical data, extract the timestamp t0 of the false alarm event, and obtain the characteristic trend curves Q1 of each dimension of the historical data feature vectors that overlap with the time period [t0-Δt, t0], where Δt represents the time window length of the historical data related to the false alarm event; based on the characteristic trend curve Q1, calculate the corresponding coefficient of variation CV and the maximum value of the instantaneous change rate M_ROC; similarly, for missed alarm events in the alarm event records of historical data, perform the same analysis as for false alarm events to obtain the corresponding coefficient of variation CV and the maximum value of the instantaneous change rate M_ROC; separately summarize the coefficient of variation CV and the maximum value of the instantaneous change rate M_ROC corresponding to false alarm events and missed alarm events, and use the coefficient of variation CV and the maximum value of the instantaneous change rate M_ROC as the abnormal data features F corresponding to false alarm events and missed alarm events, and F=(CV, M_ROC), so as to identify the abnormal data features corresponding to false alarm and missed alarm events of the power adapter anti-theft alarm device.

[0030] Among them, relying solely on alarm event records to identify false alarms and missed alarms may not be sufficient to capture all possible abnormal patterns. Alarm records are triggered by the device based on set thresholds and rules, but in actual situations, false alarms and missed alarms may occur due to various complex factors, such as environmental interference, fluctuations or errors in the sensors themselves. By combining the coefficient of variation CV and the instantaneous change rate ROC, more detailed features can be further extracted from the statistical characteristics and change patterns of the data, thereby improving the ability to identify false alarms and missed alarms. The coefficient of variation CV provides information about the stability of data changes. By calculating the ratio of the standard deviation to the mean of the data, the relative volatility of the data can be revealed, helping to distinguish normal fluctuations from abnormal fluctuations. The instantaneous change rate ROC describes the data change rate at each time point, which helps to capture sharp changes in the data in the short term and can help identify false alarms caused by sudden changes or missed alarms that are not triggered in time. Through the comprehensive analysis of these two indicators, more accurate judgments can be made on false alarms and missed alarms, avoiding the biases caused by simply relying on historical records.

[0031] Further, step S300 includes:

[0032] S301. Extract the timestamp t1 of the normal alarm events according to the alarm event records of historical data, and obtain the characteristic trend curves Q1 of each dimension of the historical data feature vector that overlaps with the time period [t1 - Δt, t1]; according to the calculation methods of the coefficient of variation CV and the maximum instantaneous change rate M_ROC corresponding to false alarm and missed alarm events, obtain the coefficient of variation CV1 and the maximum instantaneous change rate M1_ROC corresponding to the normal alarm events; summarize the coefficients of variation CV1 and the maximum instantaneous change rate M1_ROC corresponding to all normal events, and calculate the corresponding average value, so as to obtain the normal data feature F1 corresponding to the normal alarm events, and F1 = (CV1, M1_ROC);

[0033] S302. Perform a difference calculation on the normal data feature F1 corresponding to the normal alarm events and the abnormal data feature F corresponding to the false alarm and missed alarm events in turn, and the calculation formula is: R = |F1 - F|, so as to obtain the deviation data feature R corresponding to the normal alarm events and the false alarm and missed alarm events respectively, and R = (|r_cv|, |r_roc|), where |r_cv| = |CV1 - CV|, |r_roc| = |M1_ROC - M_ROC|; summarize the deviation data feature R corresponding to all normal alarm events and the false alarm and missed alarm events, and calculate the deviation distributions between the normal alarm events and the false alarm and missed alarm events respectively, so as to obtain the deviation threshold intervals P1 and P2 corresponding to the normal alarm events and the false alarm and missed alarm events, and there is no intersection between the deviation threshold intervals P1 and P2; among them, the deviation threshold interval P1 represents the deviation threshold interval corresponding to the normal alarm events and the false alarm events, and the deviation threshold interval P2 represents the deviation threshold interval corresponding to the normal alarm events and the missed alarm events.

[0034] Further, step S400 includes:

[0035] S401. Obtain the real-time data of the power adapter anti-theft alarm device, analyze the real-time data according to the analysis method of historical data, so as to obtain the real-time data feature vector S; analyze the real-time data feature vector S according to the analysis process of the historical data feature vector L, so as to obtain the real-time data characteristic trend curve Q2; based on the real-time data characteristic trend curve Q2, obtain the current corresponding coefficient of variation CV2 and the maximum instantaneous change rate M2_ROC, and form the real-time data feature F2, and F2 = (CV2, M2_ROC);

[0036] S402. Perform deviation analysis on the normal data feature F1 corresponding to the normal alarm event and the real-time data feature F2, so as to calculate the corresponding real-time deviation data feature R1, and compare the real-time deviation data feature R1 with the deviation threshold intervals P1 and P2 respectively; if the real-time deviation data feature R1 belongs to neither the deviation threshold interval P1 nor the deviation threshold interval P2, it indicates that the current is a normal event; if the real-time deviation data feature R1 only belongs to the deviation threshold interval P1, it indicates that the current is a false alarm event, and then output the notification information of the false alarm event to the relevant personnel for corresponding processing by the relevant personnel; if the real-time deviation data feature R1 only belongs to the deviation threshold interval P2, it indicates that the current is a missed alarm event, and then output the notification information of the missed alarm event to the relevant personnel for corresponding processing by the relevant personnel.

[0037] An anti-theft alarm system for power adapters based on the Internet of Things, comprising: a data acquisition and processing module, an abnormal feature recognition module, a deviation analysis module, and a real-time monitoring and alarm decision-making module;

[0038] The data acquisition and processing module collects data on the power adapter anti-theft alarm device through Internet of Things technology, obtains the historical data of the power adapter anti-theft alarm device, summarizes and analyzes the historical data, extracts the historical data features, and forms a historical data feature vector;

[0039] The abnormal feature recognition module analyzes the change trend of each dimension of the historical data feature vector according to the historical data feature vector, so as to obtain the corresponding historical data feature trend curve; based on the historical data feature trend curve, identify the abnormal data features corresponding to false alarm and missed alarm events of the power adapter anti-theft alarm device;

[0040] The deviation analysis module analyzes the deviation relationship between the abnormal data features of the power adapter anti-theft alarm device in false alarm and missed alarm events and the normal data features of normal alarm events, so as to obtain the deviation threshold intervals between the abnormal data features corresponding to false alarm and missed alarm events and the normal data features of normal alarm events;

[0041] The real-time monitoring and alarm decision-making module obtains the real-time data of the power adapter anti-theft alarm device, analyzes the real-time data, so as to obtain the real-time data features; analyzes the deviation relationship between the real-time data features and the normal data features of normal alarm events, and combines the deviation threshold intervals to judge whether an alarm event occurs currently, and performs corresponding processing according to the judgment result.

[0042] Furthermore, the data acquisition and processing module includes a data acquisition unit, a data preprocessing unit, and a feature extraction unit;

[0043] The data acquisition unit obtains various sensor data and alarm event records from the power adapter anti-theft alarm device through Internet of Things technology, including normal alarm events, false alarm events, and missed alarm events; the data preprocessing unit preprocesses the collected historical data, including cleaning, denoising, and timestamp synchronization of the sensor data; the feature extraction unit extracts statistical features and time-frequency features from the processed data and converts them into historical data feature vectors.

[0044] Furthermore, the abnormal feature recognition module includes a historical data analysis unit, an abnormal pattern recognition unit, and a feature matching unit;

[0045] The historical data analysis unit analyzes the change trends of each dimension based on the historical data feature vectors to obtain the historical data feature trend curves; the abnormal pattern recognition unit combines the historical data feature trends to identify the abnormal data features of the power adapter anti-theft alarm device during false alarms and missed alarms, so as to distinguish normal alarms from abnormal alarms; the feature matching unit identifies the feature patterns of false alarm and missed alarm events by comparing with the normal data features of normal alarm events and extracts the corresponding abnormal data features;

[0046] The deviation analysis module includes a deviation calculation unit and a deviation threshold interval generation unit;

[0047] The deviation calculation unit calculates the deviation relationship between them based on the historical data features of normal alarm events and the abnormal features of false alarm and missed alarm events, and extracts the deviation data features; the deviation threshold interval generation unit generates the deviation threshold intervals for false alarm and missed alarm events according to the calculated deviation data features.

[0048] Furthermore, the real-time monitoring and alarm decision-making module includes a real-time data acquisition and analysis unit, an alarm judgment unit, and an alarm decision-making unit;

[0049] The real-time data acquisition and analysis unit obtains the real-time data of the power adapter anti-theft alarm device, analyzes the real-time data features and generates real-time data feature vectors; the alarm judgment unit conducts deviation analysis on the real-time data features and the features of normal alarm events, and judges the current alarm event type; the alarm decision-making unit outputs an alarm notification according to the alarm judgment result and conducts relevant processing according to the false alarm or missed alarm situation.

[0050] Compared with the prior art, the beneficial effects of the present invention are as follows: By analyzing the historical data of the anti-theft alarm device for the power adapter, extracting data features and constructing a historical data feature vector, the present invention can effectively identify false alarms and missed alarms; by quantifying the abnormal data features of false alarms and missed alarms, it can more accurately judge whether an alarm event is real, significantly reducing the occurrence probability of false alarms and missed alarms. By using the DTW algorithm to synchronize the timestamps of data from different sensors, it can effectively handle the inconsistencies caused by time differences or fluctuations in sensor data, making data analysis more accurate; in addition, the smoothing and interpolation processing of noise data further improves the data quality and the accuracy of analysis. Through the analysis of real-time data, combined with the feature trend and deviation threshold interval of historical data, the present invention can make an immediate judgment and alarm processing when the anti-theft alarm device for the power adapter is abnormal; compared with the traditional system, it can detect the state changes of the power adapter and the alarm sensing wire in real time, especially in the case of power connection disconnection or voltage fluctuation, etc., it can accurately identify the abnormality and give an alarm in time, avoiding the problems of missed alarms or delayed responses. By calculating statistical features such as the smoothing of the historical data feature trend, the coefficient of variation CV, and the rate of change ROC of the instantaneous change, the present invention can identify the regular changes in the historical data and discover abnormalities in time through the changes in the data trend; this method based on data trend analysis enables the system to dynamically adapt to various environmental changes and avoid the defects of simply relying on fixed thresholds and rule judgments. By extracting historical data features and performing deviation analysis on false alarms and missed alarms, the present invention can establish a deviation threshold interval for false alarms and missed alarms and compare it with real-time data, thereby accurately judging the nature of the alarm event; the division of this deviation threshold interval can not only distinguish normal alarm events, but also accurately distinguish false alarms and missed alarms, avoiding the misjudgment problems that are prone to occur in traditional systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0052] Figure 1 is a schematic diagram of the modules of an anti-theft alarm system for a power adapter based on the Internet of Things according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0053] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0054] Please refer to Figure 1 , the present invention provides a technical solution:

[0055] An anti-theft alarm system for a power adapter based on the Internet of Things, comprising: a data acquisition and processing module, an abnormal feature recognition module, a deviation analysis module, and a real-time monitoring and alarm decision-making module;

[0056] The data acquisition and processing module collects data on the anti-theft alarm device for the power adapter through Internet of Things technology, obtains the historical data of the anti-theft alarm device for the power adapter, summarizes and analyzes the historical data, thereby extracting the historical data features and forming a historical data feature vector;

[0057] The abnormal feature recognition module analyzes the change trend of each dimension of the historical data feature vector according to the historical data feature vector, thereby obtaining the corresponding historical data feature trend curve; based on the historical data feature trend curve, it identifies the abnormal data features corresponding to false alarm and missed alarm events of the anti-theft alarm device for the power adapter;

[0058] The deviation analysis module analyzes the deviation relationship between the abnormal data features of the anti-theft alarm device for the power adapter in false alarm and missed alarm events and the normal data features of normal alarm events, thereby obtaining the deviation threshold interval between the abnormal data features corresponding to false alarm and missed alarm events and the normal data features of normal alarm events;

[0059] The real-time monitoring and alarm decision-making module obtains the real-time data of the anti-theft alarm device for the power adapter, analyzes the real-time data, thereby obtaining the real-time data features; analyzes the deviation relationship between the real-time data features and the normal data features of normal alarm events, and combines the deviation threshold interval to judge whether an alarm event occurs currently, and performs corresponding processing according to the judgment result.

[0060] The data acquisition and processing module includes a data acquisition unit, a data preprocessing unit, and a feature extraction unit;

[0061] The data acquisition unit obtains various sensor data and alarm event records from the anti-theft alarm device for the power adapter through Internet of Things technology, including normal alarm events, false alarm events, and missed alarm events; the data preprocessing unit preprocesses the collected historical data, including cleaning, denoising, and timestamp synchronization of the sensor data; the feature extraction unit extracts statistical features and time-frequency features from the processed data and converts them into a historical data feature vector.

[0062] The abnormal feature recognition module includes a historical data analysis unit, an abnormal pattern recognition unit, and a feature matching unit;

[0063] The historical data analysis unit analyzes the change trends of each dimension according to the historical data feature vector to obtain the historical data feature trend curve; the abnormal pattern recognition unit combines the historical data feature trend to identify the abnormal data features when the power adapter anti-theft alarm device has false alarms and missed alarms, so as to distinguish normal alarms from abnormal alarms; the feature matching unit identifies the feature patterns of false alarm and missed alarm events by comparing with the normal data features of normal alarm events, and extracts the corresponding abnormal data features;

[0064] The deviation analysis module includes a deviation calculation unit and a deviation threshold interval generation unit;

[0065] The deviation calculation unit calculates the deviation relationship between them according to the historical data features of normal alarm events and the abnormal features of false alarm and missed alarm events, and extracts the deviation data features; the deviation threshold interval generation unit generates the deviation threshold intervals for false alarm and missed alarm events according to the calculated deviation data features.

[0066] The real-time monitoring and alarm decision-making module includes a real-time data acquisition and analysis unit, an alarm judgment unit, and an alarm decision-making unit;

[0067] The real-time data acquisition and analysis unit obtains the real-time data of the power adapter anti-theft alarm device, analyzes the real-time data features and generates a real-time data feature vector; the alarm judgment unit performs deviation analysis on the real-time data features and the features of normal alarm events, and judges the current alarm event type; the alarm decision-making unit outputs an alarm notification according to the alarm judgment result, and performs relevant processing according to the false alarm or missed alarm situation.

[0068] An Internet of Things-based power adapter anti-theft alarm method includes the following steps:

[0069] Step S100. Through Internet of Things technology, data collection is carried out on the power adapter anti-theft alarm device to obtain the historical data of the power adapter anti-theft alarm device, the historical data is summarized and analyzed, so as to extract historical data features and form a historical data feature vector;

[0070] Step S200. According to the historical data feature vector, analyze the change trends of each dimension of the historical data feature vector, so as to obtain the corresponding historical data feature trend curve; based on the historical data feature trend curve, identify the abnormal data features corresponding to false alarm and missed alarm events of the power adapter anti-theft alarm device;

[0071] Step S300. Analyze the deviation relationship between the abnormal data features of false alarm and missed alarm events of the power adapter anti-theft alarm device and the normal data features of normal alarm events, so as to obtain the deviation threshold interval between the abnormal data features corresponding to false alarm and missed alarm events and the normal data features of normal alarm events;

[0072] Step S400. Obtain the real-time data of the power adapter anti-theft alarm device, analyze the real-time data, so as to obtain the real-time data characteristics; analyze the deviation relationship between the real-time data characteristics and the normal data characteristics of normal alarm events, and combine the deviation threshold interval to judge whether an alarm event occurs currently, and perform corresponding processing according to the judgment result.

[0073] The power adapter anti-theft alarm device mentioned in the present invention is common in the prior art, and its specific components are: an alarm sensing wire, a power adapter, a sensor monitoring module, and an alarm host; among them, the alarm sensing wire is connected between the anti-theft product and the alarm host, and triggers an alarm signal when the product is illegally removed or the sensing wire is unplugged; the power adapter is used to provide the required power support for the alarm, and is usually connected to a power socket to ensure the normal operation of the alarm device; the sensor monitoring module is responsible for real-time monitoring of the working state of the power adapter and the connection state of the alarm sensing wire; the alarm host receives signals from the alarm sensing wire and the sensor monitoring module, and includes a power adapter anti-theft alarm system, and judges whether to trigger an alarm through the power adapter anti-theft alarm system.

[0074] Step S100 includes:

[0075] S101. Through the Internet of Things technology, collect data on the power adapter anti-theft alarm device to obtain the historical data of the power adapter anti-theft alarm device; the historical data includes various sensor data in the power adapter anti-theft alarm device and alarm event records; the alarm event records include normal alarm events, false alarm events, and missed alarm events, where the normal alarm event refers to the power adapter anti-theft alarm device emitting an alarm, and after being judged by relevant personnel, there is an alarm event; the false alarm event refers to the power adapter anti-theft alarm device emitting an alarm, and after being judged by relevant personnel, there is no alarm event; the missed alarm event refers to the power adapter anti-theft alarm device not emitting an alarm, and after being judged by relevant personnel, there is an alarm event; summarize various sensor data in the historical data, and perform preprocessing on the sensor data. The preprocessing includes preprocessing such as denoising, removing outliers, and filling missing values to ensure the integrity and accuracy of the data; for the preprocessed various sensor data, use the dynamic time warping (DTW) algorithm to synchronize the timestamps of data from different sensors.

[0076] S102. Set a time window for various types of sensor data for timestamp synchronization, and divide the sensor data into several data blocks based on the set time window; for the sensor data within each time window, calculate the corresponding statistical features, and the statistical features include the mean, standard deviation, maximum value, minimum value, peak value, and skewness of the sensor data; where the mean is used to reflect the overall level of the sensor data unit, the standard deviation is used to measure the degree of data dispersion, the maximum and minimum values are used to capture the fluctuation range of the data, and the peak value is used to measure the sharpness of the data distribution; perform time-frequency feature extraction on the sensor data within each time window using wavelet transform, where the wavelet transform analyzes the time-frequency characteristics of the sensor data through the transform scale and displacement, and its formula is:

[0077] W a,b (t)=(1 / |a|)ψ[(t - b) / a];

[0078] where, W a,b (t) represents the wavelet transform result of the sensor data at the given time point t, given scale a, and displacement b, a represents the scale factor, b represents the translation factor, and t represents the time variable; ψ[(t - b) / a] represents the mother wavelet, which is the basic function for performing wavelet transform. By adjusting a and b, different time-frequency decompositions can be obtained; through wavelet transform, the time-frequency characteristics of the sensor data at different scales and time points are obtained, where these characteristics include energy, frequency band, peak value, skewness, maximum wavelet coefficient, etc.;

[0079] S103. For each data block of each type of sensor data, summarize the corresponding statistical features and time-frequency features as historical data features, and perform normalization processing to form a historical data feature vector L, and L = [l1, l2,..., ln], where l1 represents the feature value of the first dimension of the sensor data block, l2 represents the feature value of the second dimension of the sensor data block; and so on, ln represents the feature value of the nth dimension of the sensor data block, and n represents the total number of dimensions of the historical data feature vector corresponding to the sensor data block.

[0080] For the preprocessed various types of sensor data, use the dynamic time warping (DTW) algorithm to synchronize the timestamps of the data of different sensors. The specific content of synchronizing the timestamps of the data of different sensors is as follows:

[0081] For time series X = (x1, x2,..., xp) and Y = (y1, y2,..., yq) corresponding to different sensor data, where x1 represents the timestamp corresponding to the first data point in time series X, x2 represents the timestamp corresponding to the second data point in time series X, and so on, xp represents the timestamp corresponding to the p-th data point in time series X; similarly, y1 represents the timestamp corresponding to the first data point in time series Y, y2 represents the timestamp corresponding to the second data point in time series Y, and yq represents the timestamp corresponding to the q-th data point in time series Y; calculate the distance between each timestamp in time series X and time series Y, and the distance calculation formula is: d(u, v) = |xu - yv|, where d(u, v) represents the distance between xu in time series X and yv in time series Y, xu represents the timestamp corresponding to the u-th data point in time series X, and yv represents the timestamp corresponding to the v-th data point in time series Y;

[0082] Define an accumulated distance matrix D to store the cost of the shortest path, where the element D(u, v) of the accumulated distance matrix D represents the minimum cumulative distance from the u-th point of time series X to the v-th point of time series Y, and the calculation formula of the accumulated distance matrix is as follows:

[0083] D(u, v) = d(u, v) + min[D(u - 1, v), D(u, v - 1), D(u - 1, v - 1)];

[0084] Among them, D(u - 1, v) represents the minimum cumulative distance from the (u - 1)-th point of time series X to the v-th point of time series Y; D(u, v - 1) represents the minimum cumulative distance from the u-th point of time series X to the (v - 1)-th point of time series Y; D(u - 1, v - 1) represents the minimum cumulative distance from the (u - 1)-th point of time series X to the (v - 1)-th point of time series Y; for the boundary conditions of the accumulated distance matrix, assume D(0, 0) = 0, which means the starting point, and the cumulative distance is zero when there is no data; for other boundary elements: D(u, 0) = ∑ u k=1 d(k, 0), D(0, v) = ∑ v k=1 d(0, k);

[0085] According to the calculated cumulative distance matrix D, starting from D(p,q), select the one with the smallest value from D(u-1,v-1), D(u,v-1), or D(u-1,v) as the next step of the path until backtracking to D(0,0). Record the indices corresponding to (u,v) selected each time of backtracking to obtain the path with aligned timestamps in time series X and time series Y, thereby completing the timestamp synchronization of time series X and time series Y, and perform corresponding data processing on the data after timestamp synchronization. Among them, if the physical quantity values corresponding to the data points of the time series vary greatly at different timestamps, data points can be supplemented at the aligned timestamps through interpolation algorithms (such as linear interpolation, spline interpolation, etc.); for noisy sensor data, smoothing methods (such as moving average or Gaussian smoothing) can be used to suppress noise on the synchronized data.

[0086] In this embodiment, assume that the optimal path corresponding to time series X=(x1,x2,...,xp) and Y=(y1,y2,...,yq) is: (u1,v1)→(u2,v2)→...→(uk,vk). Then each pair of indices (ui,vi) is the aligned timestamp in the two time series. If p>q, it means that the length of time series X is greater than the length of time series Y. The situation where the lengths of the two time series are unequal can be processed by methods such as interpolation, resampling, dynamic time warping, or filling in missing values, and a suitable method is selected for alignment according to the actual problem.

[0087] Step S200 includes:

[0088] S201. Extract and summarize the eigenvalues corresponding to each dimension of the historical data feature vector L of various sensors, and plot the corresponding historical data feature trend curve Q in the plane rectangular coordinate system according to the time sequence for each dimension corresponding eigenvalue. The horizontal axis of the historical data feature trend curve Q is the time point, and the vertical axis is the eigenvalue; perform smoothing processing on each historical data feature trend curve Q to obtain the smoothed historical data feature trend curve Q1, and the corresponding calculation formula is:

[0089] Q1(t)=α·Q(t)+(1-α)·Q1(t-1);

[0090] where Q1(t) represents the historical data feature value corresponding to the t-th moment in the smoothed historical data feature trend curve, Q(t) represents the historical data feature value corresponding to the t-th moment in the original historical data feature trend curve Q; α represents the smoothing factor, and its value range is [0,1]; Q1(t-1) represents the historical data feature value corresponding to the (t-1)-th moment in the smoothed historical data feature trend curve;

[0091] S202. For each smoothed historical data feature trend curve Q1, calculate the standard deviation σ and the mean μ of the corresponding data points. Calculate the coefficient of variation CV based on the standard deviation σ and the mean μ, and CV = σ / μ. Calculate the rate of change (ROC) for each data point on each historical data feature trend curve Q1, and ROC(t) = [Q1(t) - Q1(t - 1)] / Q1(t - 1). Extract the maximum value of the rate of change (M_ROC) of each data point on each historical data feature trend curve Q1.

[0092] For false alarm events in the alarm event records of historical data, extract the timestamp t0 of the false alarm event, and obtain the feature trend curves Q1 of each dimension of the historical data feature vectors that overlap with the time period [t0 - △t, t0], where △t represents the time window length of the historical data related to the false alarm event. Based on the feature trend curve Q1, calculate the corresponding coefficient of variation CV and the maximum value of the rate of change (M_ROC). Similarly, for missed alarm events in the alarm event records of historical data, perform the same analysis as for false alarm events to obtain the corresponding coefficient of variation CV and the maximum value of the rate of change (M_ROC). Aggregate the coefficient of variation CV and the maximum value of the rate of change (M_ROC) corresponding to false alarm events and missed alarm events respectively. Use the coefficient of variation CV and the maximum value of the rate of change (M_ROC) as the abnormal data features F corresponding to false alarm events and missed alarm events, and F = (CV, M_ROC), so as to identify the abnormal data features corresponding to false alarm and missed alarm events of the power adapter anti-theft alarm device.

[0093] Among them, relying solely on alarm event records to identify false alarms and missed alarms may not be sufficient to capture all possible abnormal patterns. Alarm records are triggered by the device based on set thresholds and rules. However, in actual situations, false alarm and missed alarm events may occur due to various complex factors, such as environmental interference, fluctuations or errors in the sensors themselves, etc. By combining the coefficient of variation CV and the rate of change (ROC), more detailed features can be further extracted from the statistical characteristics and change patterns of the data, thereby improving the ability to identify false alarm and missed alarm events. The coefficient of variation CV provides information about the stability of data changes. By calculating the ratio of the standard deviation to the mean of the data, the relative volatility of the data can be revealed, helping to distinguish normal fluctuations from abnormal fluctuations. The rate of change (ROC) describes the rate of data change at each time point, which helps to capture sharp changes in the data in the short term and can help identify false alarms caused by sudden changes or missed alarms that are not triggered in a timely manner. Through the comprehensive analysis of these two indicators, more accurate judgments can be made on false alarms and missed alarms, avoiding the bias caused by simply relying on historical records.

[0094] In this embodiment, since the alarm events in the historical data records are usually triggered based on certain significant thresholds, but false alarms and missed alarms sometimes do not occur at these significant threshold points. The actual false alarm and missed alarm events may be just minor fluctuations in the data or some trends that have not been fully analyzed. Through the dual determination of CV and ROC, those minor fluctuations and trend changes near the thresholds can be identified, which is crucial for capturing potential false alarms or missed alarm events. For example, some alarm events may occur during the stable period of the data rather than the period of sudden drastic fluctuations. At this time, they cannot be effectively captured only through the historical alarm records. However, by combining CV and ROC, potential deviations can be identified within the normal fluctuation range of the data, thus effectively identifying missed alarms.

[0095] Step S300 includes:

[0096] S301. According to the alarm event records of the historical data, extract the time stamp t1 of the normal alarm event, and obtain the characteristic trend curves Q1 of each dimension of the historical data feature vector that overlaps with the time period [t1 - △t, t1]; according to the calculation methods of the coefficient of variation CV and the maximum value of the instantaneous change rate M_ROC corresponding to false alarms and missed alarms, obtain the coefficient of variation CV1 and the maximum value of the instantaneous change rate M1_ROC corresponding to the normal alarm event; summarize the coefficients of variation CV1 and the maximum values of the instantaneous change rate M1_ROC corresponding to all normal events, and calculate the corresponding average value, so as to obtain the normal data feature F1 corresponding to the normal alarm event, and F1 = (CV1, M1_ROC);

[0097] S302. Perform a difference calculation on the normal data feature F1 corresponding to the normal alarm event and the abnormal data feature F corresponding to false alarms and missed alarms in turn, and the calculation formula is: R = |F1 - F|, so as to obtain the deviation data features R corresponding to the normal alarm event and false alarms and missed alarms respectively, and R = (|r_cv|, |r_roc|), where |r_cv| = |CV1 - CV|, |r_roc| = |M1_ROC - M_ROC|; summarize the deviation data features R corresponding to all normal alarm events and false alarms and missed alarms, and calculate the deviation distributions between the normal alarm event and false alarms and missed alarms respectively, so as to obtain the deviation threshold intervals P1 and P2 corresponding to the normal alarm event and false alarms and missed alarms, and there is no intersection between the deviation threshold intervals P1 and P2; where the deviation threshold interval P1 represents the deviation threshold interval corresponding to the normal alarm event and false alarms, and the deviation threshold interval P2 represents the deviation threshold interval corresponding to the normal alarm event and missed alarms.

[0098] In this embodiment, taking the deviation threshold interval P1 corresponding to normal alarm events and false alarm events as an example, since R = (|r_cv|, |r_roc|), the deviation threshold interval P1 includes two parts. One is the deviation threshold interval corresponding to the coefficient of variation, and the other is the deviation threshold interval corresponding to the maximum value of the instantaneous change rate. The deviation distribution between normal alarm events and false alarm events is calculated using the mean and standard deviation.

[0099] The deviation threshold interval corresponding to the coefficient of variation is expressed as: [μ_|r_cv| - k1·σ_|r_cv|, μ_|r_cv| + k1·σ_|r_cv|]; the deviation threshold interval corresponding to the maximum value of the instantaneous change rate is expressed as: [μ_|r_roc| - k2·σ_|r_roc|, μ_|r_roc| + k2·σ_|r_roc|], where both k1 and k2 represent adjustment factors used to control the width of the error range.

[0100] Step S400 includes:

[0101] S401. Obtain the real-time data of the power adapter anti-theft alarm device, analyze the real-time data according to the analysis method of historical data to obtain the real-time data feature vector S; analyze the real-time data feature vector S according to the analysis process of the historical data feature vector L to obtain the real-time data feature trend curve Q2; based on the real-time data feature trend curve Q2, obtain the corresponding coefficient of variation CV2 and the maximum value of the instantaneous change rate M2_ROC at present, and form the real-time data feature F2, and F2 = (CV2, M2_ROC);

[0102] S402. Conduct a deviation analysis on the normal data feature F1 corresponding to the normal alarm event and the real-time data feature F2 to calculate the corresponding real-time deviation data feature R1, and compare the real-time deviation data feature R1 with the deviation threshold intervals P1 and P2 respectively; if the real-time deviation data feature R1 belongs to neither the deviation threshold interval P1 nor the deviation threshold interval P2, it indicates that the current is a normal event; if the real-time deviation data feature R1 only belongs to the deviation threshold interval P1, it indicates that the current is a false alarm event, and then output the notification information of the false alarm event to the relevant personnel for corresponding processing by the relevant personnel; if the real-time deviation data feature R1 only belongs to the deviation threshold interval P2, it indicates that the current is a missed alarm event, and then output the notification information of the missed alarm event to the relevant personnel for corresponding processing by the relevant personnel.

[0103] It should be noted that in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0104] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A power adapter anti-theft alarm method based on the Internet of Things, characterized in that: The method comprises the following steps: Step S100. Through the Internet of Things technology, data is collected on the power adapter anti-theft alarm device to obtain historical data of the power adapter anti-theft alarm device, and the historical data is summarized and analyzed to extract historical data features and form a historical data feature vector; Step S200. Analyze the change trend of each dimension of the historical data feature vector according to the historical data feature vector, so as to obtain the corresponding historical data feature trend curve; based on the historical data feature trend curve, identify the abnormal data features corresponding to the false alarm event and the missed alarm event of the power adapter anti-theft alarm device; Step S300. Analyze the deviation relationship between the abnormal data features of the power adapter anti-theft alarm device when false alarm events and missed alarm events occur and the normal data features of normal alarm events, so as to obtain the deviation threshold interval between the abnormal data features corresponding to the false alarm events and missed alarm events and the normal data features of normal alarm events; Step S400. Acquire real-time data of the power adapter anti-theft alarm device, analyze the real-time data, and obtain real-time data features; analyze the deviation relationship between the real-time data features and the normal data features of normal alarm events, and determine whether an alarm event currently occurs in combination with the deviation threshold interval, and perform corresponding processing according to the determination result; the normal alarm event refers to the power adapter anti-theft alarm device sounding an alarm, and after relevant personnel determine that an alarm event exists; the false alarm event refers to the power adapter anti-theft alarm device sounding an alarm, and after relevant personnel determine that no alarm event exists; the missed alarm event refers to the power adapter anti-theft alarm device not sounding an alarm, and after relevant personnel determine that an alarm event exists.

2. According to the method for anti-theft alarm of a power adapter based on the Internet of Things in claim 1, it is characterized in that: The step S100 includes: S101. Through the Internet of Things technology, data is collected on the power adapter anti-theft alarm device to obtain historical data of the power adapter anti-theft alarm device; the historical data includes various sensor data and alarm event records in the power adapter anti-theft alarm device; the alarm event records include normal alarm events, false alarm events and missed alarm events; various sensor data in the historical data are summarized and pre-processed; for various sensor data after pre-processing, the dynamic time warping (DTW) algorithm is used to synchronize the timestamps of different sensor data; S102. For various sensor data synchronized with timestamps, a time window is set, and the sensor data is divided into several data blocks based on the set time window; for the sensor data in each time window, the corresponding statistical features are calculated, and the statistical features include the mean, standard deviation, maximum value, minimum value, peak value and skewness of the sensor data; the wavelet transform is used to extract the time-frequency features of the sensor data in each time window, wherein the wavelet transform analyzes the time-frequency characteristics of the sensor data by transforming the scale and displacement, and the formula is: W a,b (t)=(1 / |a|)ψ[(t-b) / a]; Among them, W a,b (t) represents the wavelet transform result of the sensor data at a given time point t and a given scale a and displacement b, a represents the scale factor, b represents the translation factor, and t represents the time variable; ψ[(tb) / a] represents the mother wavelet; through wavelet transform, the time-frequency characteristics of the sensor data at different scales and time points are obtained; S103. For each data block of each type of sensor data, the corresponding statistical features and time-frequency features are summarized as historical data features, and normalized to form a historical data feature vector L, and L=[l1,l2,...,ln], where l1 represents the eigenvalue of the first dimension of the sensor data block, l2 represents the eigenvalue of the second dimension of the sensor data block; and so on, ln represents the eigenvalue of the nth dimension of the sensor data block, and n represents the total number of dimensions of the historical data feature vector corresponding to the sensor data block.

3. The power adapter anti-theft alarm method based on the Internet of Things according to claim 2 is characterized in that: The dynamic time warping (DTW) algorithm is used to synchronize the timestamps of the data of different sensors after preprocessing. The specific contents of synchronizing the timestamps of the data of different sensors are as follows: For the time series X=(x1,x2,...,xp) and Y=(y1,y2,...,yq) corresponding to different sensor data, x1 represents the timestamp corresponding to the first data point in the time series X, x2 represents the timestamp corresponding to the second data point in the time series X, and so on, xp represents the timestamp corresponding to the pth data point in the time series X; similarly, y1 represents the timestamp corresponding to the first data point in the time series Y, y2 represents the timestamp corresponding to the second data point in the time series Y, and yq represents the timestamp corresponding to the qth data point in the time series Y; calculate the distance between each timestamp in the time series X and the time series Y, and the distance calculation formula is: d(u,v)=|xu-yv|, where d(u,v) represents the distance between xu in the time series X and yv in the time series Y, xu represents the timestamp corresponding to the uth data point in the time series X, and yv represents the timestamp corresponding to the vth data point in the time series Y; Define a cumulative distance matrix D to store the cost of the shortest path, where the element D(u,v) of the cumulative distance matrix D represents the minimum cumulative distance from the u-th point of the time series X to the v-th point of the time series Y, and the calculation formula of the cumulative distance matrix is ​​as follows: D(u,v)=d(u,v)+min[D(u-1,v),D(u,v-1),D(u-1,v-1)]; Among them, D(u-1,v) represents the minimum cumulative distance from the u-1th point of time series X to the vth point of time series Y; D(u,v-1) represents the minimum cumulative distance from the u-1th point of time series X to the v-1th point of time series Y; D(u-1,v-1) represents the minimum cumulative distance from the u-1th point of time series X to the v-1th point of time series Y; for the boundary conditions of the cumulative distance matrix, assume that D(0,0)=0, indicating that it is the starting point, and the cumulative distance is zero when there is no data; for other boundary elements: D(u,0)=∑ u k=1 d(k,0),D(0,v)=∑ v k=1 d(0,k); According to the calculated cumulative distance matrix D, starting from D(p,q), select the smallest value from D(u-1,v-1), D(u,v-1) or D(u-1,v) as the next step of the path until backtracking to D(0,0), record the index corresponding to (u,v) selected each time backtracking, and obtain the path with aligned timestamps in time series X and time series Y, thereby completing the timestamp synchronization of time series X and time series Y, and the data after timestamp synchronization is processed accordingly.

4. The power adapter anti-theft alarm method based on the Internet of Things according to claim 2 is characterized in that: The step S200 includes: S201. Extract and summarize the eigenvalues ​​corresponding to each dimension of the historical data feature vector L of each type of sensor, and draw the corresponding historical data feature trend curve Q in the plane rectangular coordinate system according to the time sequence of the eigenvalues ​​corresponding to each dimension, and the horizontal axis of the historical data feature trend curve Q is the time point, and the vertical axis is the eigenvalue; perform smoothing on each historical data feature trend curve Q, so as to obtain the smoothed historical data feature trend curve Q1, and the corresponding calculation formula is: Q1(t)=α·Q(t)+(1-α)·Q1(t-1); Among them, Q1(t) represents the historical data characteristic value corresponding to time t in the smoothed historical data characteristic trend curve, Q(t) represents the historical data characteristic value corresponding to time t in the original historical data characteristic trend curve Q; α represents the smoothing factor, and its value range is [0,1]; Q1(t-1) represents the historical data characteristic value corresponding to time t-1 in the smoothed historical data characteristic trend curve; S202. For each smoothed historical data characteristic trend curve Q1, calculate the standard deviation σ and the average value μ of the corresponding data point, calculate the coefficient of variation CV according to the standard deviation σ and the average value μ, and CV=σ / μ; calculate the instantaneous rate of change ROC for each data point on the historical data characteristic trend curve Q1, and ROC(t)=[Q1(t)-Q1(t-1)] / Q1(t-1), extract the data points on each historical data characteristic trend curve Q1 to calculate the maximum instantaneous rate of change M_ROC; For the false alarm events in the alarm event records in the historical data, the timestamp t0 of the false alarm events is extracted, and the characteristic trend curve Q1 of each dimension of the historical data feature vector that overlaps with the time period [t0-△t, t0] is obtained, where △t represents the time window length of the historical data related to the false alarm events; based on the characteristic trend curve Q1, the corresponding coefficient of variation CV and the maximum instantaneous rate of change M_ROC are calculated; similarly, for the missed alarm events in the alarm event records in the historical data, the same analysis as the false alarm events is performed to obtain the corresponding coefficient of variation CV and the maximum instantaneous rate of change M_ROC; the coefficient of variation CV and the maximum instantaneous rate of change M_ROC corresponding to the false alarm events and the missed alarm events are respectively summarized, and the coefficient of variation CV and the maximum instantaneous rate of change M_ROC are used as the abnormal data features F corresponding to the false alarm events and the missed alarm events, and F=(CV,M_ROC), so as to identify the abnormal data features corresponding to the false alarm events and the missed alarm events of the power adapter anti-theft alarm device.

5. The power adapter anti-theft alarm method based on the Internet of Things according to claim 4 is characterized in that: The step S300 includes: S301. According to the alarm event records of historical data, extract the timestamp t1 of the normal alarm event, and obtain the characteristic trend curve Q1 of each dimension of the historical data feature vector that overlaps with the time period [t1-△t, t1]; according to the calculation method of the coefficient of variation CV and the maximum instantaneous change rate M_ROC corresponding to the false alarm event and the missed alarm event, obtain the coefficient of variation CV1 and the maximum instantaneous change rate M1_ROC corresponding to the normal alarm event; summarize the coefficient of variation CV1 and the maximum instantaneous change rate M1_ROC corresponding to all normal events, calculate the corresponding average value, and thus obtain the normal data feature F1 corresponding to the normal alarm event, and F1=(CV1,M1_ROC); S302. The normal data feature F1 corresponding to the normal alarm event is calculated in turn with the abnormal data feature F corresponding to the false alarm event and the missed alarm event, and the calculation formula is: R=|F1-F|, so as to obtain the deviation data features R corresponding to the normal alarm event and the false alarm event and the missed alarm event respectively, and R=(|r_cv|,|r_roc|), wherein |r_cv|=|CV1-CV|, |r_roc|=|M1_ROC-M_ROC|; the deviation data features R corresponding to all normal alarm events and the false alarm events and the missed alarm events are summarized, and the deviation distribution between the normal alarm events and the false alarm events and the missed alarm events is calculated respectively, so as to obtain the deviation threshold intervals P1 and P2 corresponding to the normal alarm events and the false alarm events and the missed alarm events, and there is no intersection between the deviation threshold intervals P1 and P2; wherein the deviation threshold interval P1 represents the deviation threshold interval corresponding to the normal alarm event and the false alarm event, and the deviation threshold interval P2 represents the deviation threshold interval corresponding to the normal alarm event and the missed alarm event.

6. The power adapter anti-theft alarm method based on the Internet of Things according to claim 5 is characterized in that: The step S400 includes: S401. Acquire the real-time data of the power adapter anti-theft alarm device, analyze the real-time data with reference to the analysis method of historical data, and obtain the real-time data feature vector S; analyze the real-time data feature vector S according to the analysis process of the historical data feature vector L, and obtain the real-time data feature trend curve Q2; based on the real-time data feature trend curve Q2, obtain the current corresponding coefficient of variation CV2 and the maximum instantaneous rate of change M2_ROC, and form the real-time data feature F2, and F2=(CV2,M2_ROC); S402. Perform deviation analysis on the normal data feature F1 and the real-time data feature F2 corresponding to the normal alarm event, so as to calculate the corresponding real-time deviation data feature R1, and compare the real-time deviation data feature R1 with the deviation threshold intervals P1 and P2 respectively; if the real-time deviation data feature R1 belongs neither to the deviation threshold interval P1 nor to the deviation threshold interval P2, it indicates that the current event is a normal event; if the real-time deviation data feature R1 only belongs to the deviation threshold interval P1, it indicates that the current event is a false alarm event, then the notification information of the false alarm event is output to the relevant personnel, and the relevant personnel perform corresponding processing; if the real-time deviation data feature R1 only belongs to the deviation threshold interval P2, it indicates that the current event is a missed alarm event, then the notification information of the missed alarm event is output to the relevant personnel, and the relevant personnel perform corresponding processing.

7. A power adapter anti-theft alarm system based on the Internet of Things, applied to a power adapter anti-theft alarm method based on the Internet of Things as claimed in any one of claims 1 to 6, characterized in that: The system includes: a data acquisition and processing module, an abnormal feature recognition module, a deviation analysis module, and a real-time monitoring and alarm decision module; The data collection and processing module collects data from the power adapter anti-theft alarm device through the Internet of Things technology, obtains historical data of the power adapter anti-theft alarm device, summarizes and analyzes the historical data, thereby extracting historical data features and forming a historical data feature vector; The abnormal feature recognition module analyzes the change trend of each dimension of the historical data feature vector according to the historical data feature vector, thereby obtaining a corresponding historical data feature trend curve; based on the historical data feature trend curve, the abnormal data features corresponding to the false alarm event and the missed alarm event of the power adapter anti-theft alarm device are identified; The deviation analysis module analyzes the deviation relationship between the abnormal data features of the power adapter anti-theft alarm device when false alarm events and missed alarm events occur and the normal data features of normal alarm events, thereby obtaining a deviation threshold interval between the abnormal data features corresponding to the false alarm events and missed alarm events and the normal data features of normal alarm events; The real-time monitoring and alarm decision module obtains the real-time data of the power adapter anti-theft alarm device, analyzes the real-time data, and thus obtains the real-time data features; analyzes the deviation relationship between the real-time data features and the normal data features of the normal alarm event, and determines whether an alarm event currently occurs in combination with the deviation threshold interval, and performs corresponding processing according to the judgment result; the normal alarm event refers to the power adapter anti-theft alarm device sending out an alarm, and after the relevant personnel judge, there is an alarm event; the false alarm event refers to the power adapter anti-theft alarm device sending out an alarm, and after the relevant personnel judge, there is no alarm event; the missed alarm event refers to the power adapter anti-theft alarm device not sending out an alarm, and after the relevant personnel judge, there is an alarm event.

8. The power adapter anti-theft alarm system based on the Internet of Things according to claim 7, characterized in that: The data acquisition and processing module includes a data acquisition unit, a data preprocessing unit and a feature extraction unit; The data acquisition unit obtains various sensor data and alarm event records from the power adapter anti-theft alarm device through the Internet of Things technology, including normal alarm events, false alarm events and missed alarm events; the data preprocessing unit preprocesses the collected historical data, including cleaning, denoising and timestamp synchronization of the sensor data; the feature extraction unit extracts statistical features and time-frequency features from the processed data and converts them into historical data feature vectors.

9. The power adapter anti-theft alarm system based on the Internet of Things according to claim 7, characterized in that: The abnormal feature recognition module includes a historical data analysis unit, an abnormal pattern recognition unit and a feature matching unit; The historical data analysis unit analyzes the change trend of each dimension according to the historical data feature vector to obtain the historical data feature trend curve; the abnormal pattern recognition unit identifies the abnormal data characteristics of the power adapter anti-theft alarm device when false alarms and missed alarms occur, thereby distinguishing normal alarms from abnormal alarms; The feature matching unit identifies the feature patterns of false alarm events and missed alarm events by comparing them with the normal data features of normal alarm events, and extracts the corresponding abnormal data features; The deviation analysis module includes a deviation calculation unit and a deviation threshold interval generation unit; The deviation calculation unit calculates the deviation relationship between the historical data characteristics of normal alarm events and the abnormal characteristics of false alarm and missed alarm events, and extracts the deviation data characteristics; The deviation threshold interval generating unit generates deviation threshold intervals for false alarm events and missed alarm events according to the calculated deviation data features.

10. The power adapter anti-theft alarm system based on the Internet of Things according to claim 7, characterized in that: The real-time monitoring and alarm decision module includes a real-time data acquisition and analysis unit, an alarm judgment unit and an alarm decision unit; The real-time data acquisition and analysis unit acquires the real-time data of the power adapter anti-theft alarm device, analyzes the real-time data features and generates a real-time data feature vector; the alarm judgment unit performs deviation analysis on the real-time data features and the features of normal alarm events, and judges the current alarm event type; the alarm decision unit outputs an alarm notification based on the alarm judgment result, and performs relevant processing based on the false alarm or missed alarm situation.

Images